MTNS – 2002 1 Final Program

Fifteenth International Symposium on

Mathematical Theory of Networks and Systems

University of Notre Dame, South Bend, Indiana, USA August 12-16, 2002.

www.nd.edu/~mtns/ MTNS – 2002 2 Organizing Committee of MTNS 2002

SYMPOSIUM CHAIR Victor Vinnikov (Israel) Joachim Rosenthal (USA) Xiaochang Wang (USA) Shigeru Yamamoto (Japan) PUBLICATION CHAIR Sandro Zampieri (Italy) David S. Gilliam (USA) STEERING COMMITTEE PROGRAM COMMITTEE V. Blondel (Belgium) Mark Alber (USA) C.I. Byrnes (USA) Joe Ball (USA) R. Curtain (The Netherlands) Vincent Blondel (Belgium) B.N. Datta (USA) Tyrone Duncan (USA) P. Dewilde (The Netherlands) Avraham Feintuch (Israel) P. Van Dooren (Belgium) David Forney (USA) H. Dym (Israel) Krzysztof Galkowski (Poland) A. El Jay (France) Tryphon Georgiou (USA) M. Fliess (France) Heide Gluesing-Luerssen (Germany) P. Fuhrmann (Israel) Koichi Hashimoto (Japan) I. Gohberg (Israel) Bernard Hanzon (The Netherlands) U. Helmke (Germany) Diederich Hinrichsen (Germany) J.W. Helton (USA) Aleksandar Kavcic (USA) A. Isidori (Italy) Matthias Kawski (USA) M.A. Kaashoek (The Netherlands) Belinda King (USA) H. Kimura (Japan) Wolfgang Kliemann (USA) A.J. Krener (USA) Margreet Kuijper (Australia) A.B. Kurzhansky (Russia) Naomi Leonard (USA) A. Lindquist (Sweden) Daniel Liberzon (USA) C.F. Martin (USA) Wei Lin (USA) G. Picci (Italy) Brian Marcus (USA) A.C.M. Ran (The Netherlands) Volker Mehrmann (Germany) A. Rantzer (Sweden) Raimund Ober (USA) J. Rosenthal (USA) Dieter Praetzel-Wolters (Germany) J.H. van Schuppen (The Netherlands) Eric Rogers (United Kingdom) Y. Yamamoto (Japan) Pierre Rouchon (France) Hans Schumacher (The Netherlands) HONORARY MEMBERS Mark Shayman (USA) C.A. Desoer (USA) Rodolphe Sepulchre (Belgium) R.W. Newcomb (USA) Anton Stoorvogel (The Netherlands) A.H. Zemanian (USA). Maria Elena Valcher (Italy) MTNS – 2002 3 WELCOME TO MTNS 2002

On behalf of the Organizing Committee of MTNS 2002 I welcome you to the 15th edition of MTNS, the International Symposium on the Mathematical Theory of Networks and Systems.

The symposium is organized every two years and traditionally covers areas involv- ing a wide range of research directions in mathematical systems, networks and control theory. Mathematical methods which play a role in the areas mentioned above stem from a broad range of fields of pure and applied mathematics, including ordinary and partial differential equations, real and complex analysis, numerical analysis, probability theory and stochastic analysis, operator theory, linear and commutative algebra as well as algebraic and differential geometry. There are a wide range of applications ranging from problems in biology, communications and mathematical finance to problems in chemical engineering, aerospace engineering and robotics. One of the special features of MTNS 2002 will be 5 Mini-symposia, each consisting of a whole series of sessions. The Mini-sypomposia reflect areas where systems and control theory play a significant role.

MTNS 2002 features a total of 20 plenary and semi-plenary talks by some of the leading researchers in the area of systems and control theory. There will also be 90 sessions lasting 2 hours each. Members of the International Program Committee and Members from the Steering Committee were actively involved in the reviewing process and in the organization of sessions. I would like to express my sincere thank to all the members of these committees. I would also like to thank the Staff of the Department of Mathematics and the Center for Continuing Education at Notre Dame for their great help in the organization. A special thanks goes to David Gilliam, who worked countless hours, day and night, to make the submission of abstracts and papers and the completion of the conference proceedings a reality.

MTNS 2002 received generous support from the National Science Foundation, The Institute for Mathematics and its Applications (IMA) in Minnesota and from various departments and colleges at the University of Notre Dame. This support made it possible that many young researchers received travel support to attend MTNS 2002.

We hope you will find the 15th edition of MTNS interesting and rewarding.

Joachim Rosenthal MTNS 2002 Symposium Chair MTNS – 2002 4

MTNS HISTORY

The fourteen previous MTNS meetings were held in:

1973 : College Park, Maryland, USA, 1987 : Phoenix, Arizona, USA,

1975 : Montreal,Canada, 1989 : Amsterdam, Netherlands,

1977 : Lubbock, Texas, USA, 1991 : Kobe, Japan,

1979 : Delft, Netherlands, 1993 : Regensburg, Germany,

1981 : Santa Monica, California, USA, 1996 : St. Louis, Missouri, USA,

1983 : Beer Sheva, Israel, 1998 : Padova, Italy,

1985 : Stockholm, Sweden, 2000 : Perpignan, France.

Financial Support was received for MTNS 2002 from: National Science Foundation (NSF) Institute for Mathematics and its Applications (IMA) Center for Applied Mathematics, Notre Dame Provost Office of University of Notre Dame College of Science, University of Notre Dame Graduate School at University of Notre Dame MTNS – 2002 5 Special Events at MTNS 2002

Minisymposium on Biological Systems: • Organizers: Mark Alber and Raimund Ober. Sessions: TUA2: Patterns in Biology TUM2: Immunology 1: Introduction and Microscopy TUP2: Immunology 2: Microscopy and Biophysics WA1: Immunology 3: Structure and Kinetics WM1: Immunology 4: Diffusion and Modelling WP1: Immunology 5: Cellular Aspects THA1: Complex Networks and Biological Applications 1 THM1: Complex Networks and Biological Applications 2 THP1: Complex Networks and Biological Applications 3 FA1: Genetic Networks

Minisymposium on Communication Systems: • Organizers: David Forney and Brian Marcus. Sessions: MA1: Minicourse on “Capacity of Multidimensional Codes” MM1: Capacity of Multi-Dimensional Codes Part II MP1: The Interaction of Control, Information and Communication TUA1: Design and Analysis of Block Codes, Part I TUM1: Design and Analysis of Block Codes, Part II TUP1: Convolutional Codes WA2: Computer Networks WM2: Control and Communications WP2: Cryptography

Minisymposium on Control and Computation: • Organizers: Paul van Dooren, Uwe Helmke and Volker Mehrmann Sessions: WA4: Model Reduction WM4: Control and Computation WP4: Large-Scale Computations in Control THA4: Fully Nonlinear, Three-Dimensional, Surface Water Waves in Arbitrary Depth THM4: Robust Control and Linear Matrix Inequalities THP4: Computational Methods for Structured Matrices and Applications FA4: Stability and Numerics FM4: Nonlinear Surface Water Waves: Theory, Computation and Experiment MTNS – 2002 6

Minisymposium on Financial Systems: • Organizers: Hans Schumacher and Michael Taksar. Sessions: THA3: Systems and Control Theory in Finance and Insurance 1 THM3: Systems and Control Theory in Finance and Insurance 2

Minisymposium on Multidimensional Systems: • Organizers: Krzysztof Galkowski, Eric Rogers and Victor Vinnikov. Sessions: TUA3: Minicourse A: Multidimensional Systems TUM3: Multidimensional Systems 1 TUP3: Multidimensional Systems 2 WA3: Minicourse B: Multidimensional Systems WM3: Multidimensional Systems 3 WP3: Multidimensional Systems 4 THA5: Multidimensional Systems 5

Workshop on Open Problems in Systems Theory: • Organizer: Vincent Blondel Time: Monday and Tuesday, August 12, 13 from 20:00–22:00 in Room 102 De Bartolo Hall.

Panel Discussion on “Future Directions of Research and Teaching on Mathe- • matical Control and Systems Theory”: Organizer: Biswa Datta Time: Wednesday Evening August 14, from 20:00–22:00 in Room 102 De Bartolo Hall. MTNS – 2002 7

Plenary Speakers:

Anthony Bloch (University of Michigan), William Helton (University of California), Bruce Hajek (University of Illinois), Gilbert Strang (MIT), Eduardo Sontag (Rutgers University).

Special Topic Invited Speaker:

Albert-Laszlo Barabasi (University of Notre Dame), Roger Brockett (Harvard University), Raffaello D’Andrea (Cornell University), Matthias Heinkenschloss (Rice University), Knut Hueper (University of Wuerzburg), Karl Kunisch (Graz University), Hans-Andrea Loeliger (ETH), Robert J. McEliece (Cal Tech), Wolfgang Runggaldier, (University of Padova), Arjan van der Schaft (University of Twente), Olof Staffans (Abo Akademi University), Allen Tannenbaum (Georgia Tech), Sjoerd Verduyn Lunel (University of Leiden), Jan Willems (University of Groningen), Jeffrey Wood (University of Southampton). MTNS – 2002 8 General Information

Registration Desk: The Registration desk is Recreational Activities: The registration located in McKenna Hall, also called Center for package will include information on possibilities Continuing Education (CCE). It will be open to use the recreational facilities at Notre Dame Sunday afternoon August 11, and staff will be like e.g. the swimming pools and the golf course. available throughout the conference week during business hours.

Social Events:

Shuttle Bus: There will be continuous loops from the two conference hotels (Comfort Suites Sunday 18:00-21:00: There is a Welcoming • and Inn at Saint Mary’s) to the conference center Party in the Center for Continuing Educa- (CCE) housed in McKenna Hall during the times: tion. Sunday, August 11 noon - 9:30 pm Thursday 19:00-22:00: Banquet Dinner Monday, August 12 7 am - 10:30 pm • Tuesday, August 13 7 am - 10:30 pm Friday 18:00-20:00: Farewell Party. Wednesday, August 14 7 am - 10:30 pm • Thursday, August 15 7 am - 10:30 pm Friday, August 16 7 am - 9 pm Notre Dame Tourism: The University offers a regular schedule of campus tours. Call the Eck Visitors’ Center at (574) 631-5726 for more infor- Computer Access: Every participant receives mation. his own login name and his own password. It will Next to De Bartolo Hall is the Snite Museum of allow her/him to access the Notre Dame com- Art with a good collection of Fine Art. Admis- puter system from a machine in De Bartolo Hall. sion is free. MTNS – 2002 9 M 20:00-22:00 Dinner M 16:30-18:30 Coffee Break M 14:00-16:00 Lunch Break M 11:00-13:00 Coffee Break M 9:30-10:30 M 8:30-9:30 M 8:00-8:30 Welcoming Remarks Roger Bruce Hajek Room 101 Brockett Workshop Open Communication Information of Control, The Codes Multidimensional Capacity Codes Multidimensional Capacity Andrea Room 102 Interaction Problems 2 1 Loeliger

of of and and Nonlinear and Nonlinear ControlApplications Control Control Room 126 Monday, August

Systems Systems 2 1 Application Control Stochastic System 1 Stochastic Room 129

and

its Inverse Indefinite Spaces Operator for Problems in State Equations Matrix and Operator Sjoerd Verduyn Lunel 12, 2002 Room 136/138 Space with Theory Metric I Methods and Systems Control of Linear Output Positive Room 208 Feedback Systems Systems Algebraic Optimal Optimization and Systems Mechanical Quantum Control of Room 209 Control Theory and Hybrid Control Adaptive Room 210 Control Systems MTNS – 2002 10

Tu 20:00-22:00 Dinner Tu 16:30-18:30 Coffee Break Tu 14:00-16:00 Lunch Break Tu 10:30-12:30 Coffee Break Tu 9:00-10:00 Tu 8:00-9:00

Arjan Gilbert Strang Room 101 Schaft van der Workshop Open Codes Convolutional Block Analysis Design Block Analysis Design Room 102 McEliece Robert Problems Codes Codes and and

of of 2 1 Biophysics Microscopy Immunology 2: Microscopy Introduction Immunology 1: Patterns Room 126 in Tuesday, August

Biology and and Systems Multidimensional Systems Multidimensional Systems Multidimensional Minicourse Jeffrey Room 129

2 1 Wood and Nonlinear Stability, Part Input-to-State 13, 2002 Control Room 136/138 Systems

3 I Completion Problems Interpolation Developments Recent Linear Room 208 Systems

and in Systems Dimensional Infinite Parameter Distributed Control of Engineering 1 Quantum Room 209

Systems and Infinity Control Robust Identification Filtering and and Identification, Estimation, Robust Room 210 Estimation Detection and

H- MTNS – 2002 11

22:00 W 20:00- Dinner 18:30 W 16:30- Coffee Break 16:00 W 14:00- Lunch Break 12:30 W 10:30- Coffee Break 10:00 W 9:00- W 8:00-9:00

Jan Willems Room 101 William Helton Discussion Panel Aspects 5: Cellular Immunology and Modelling 4: Diffusion Immunology and Kinetics 3: Structure Immunology Room 102 Barabasi Laszlo

Cryptography Communications Control and Networks Computer Room 126 Wednesday, August

Systems 4 Multidimensional Systems 3 Multidimensional Systems Multidimensional Minicourse Room 129

Control Computations in Large-Scale Computation Control and Model Reduction Room 136/138 Knut Hueper 14, 2002

Applications Together with Sums of Squares Polynomials as Expressing System 2 Stochastic Problems Numerical Systems and Time-Varying Room 208 Engineering 2 Quantum Systems Theory Geometry in Differential Algebraic and Control 4 Systems and Nonlinear Room 209 Diagnosis Synthesis and Analysis, System Control Hybrid Systems and Hybrid Discrete Event Room 210 MTNS – 2002 12

Dinner Banquet Th 16:30-18:30 Coffee Break Th 14:00-16:00 Lunch Break Th 10:30-12:30 Coffee Break Th 9:00-10:00 Th 8:00-9:00

Room 101 Eduardo Sontag Networks 3 Biological Networks 2 Biological Networks 1 Biological Room 102 Control, Part I Stabilization and Systems: Parameter Distributed Part II Systems: Parameter Distributed Part I Systems: Parameter Distributed Olof Room 126 Staffans Theory Theory Thursday, August Applications Theory Stochastic Insurance, in Control Systems Insurance, in Control Systems Runggaldier Finance Finance Wolfgang Room 129 and Theory Theory and and Part Part and and its 2 1 and for Computational Linear Matrix Robust Arbitrary Water Dimensional, Fully Nonlinear, Three- Structured Heinkenschloss Applications Room 136/138 Waves Control Matthias 15, 2002 Depth Surface Inequalities in Matrices Methods and to Dynamic The 5 Multidimensional Behavioral Room 208 Systems Approach Systems Systems Mechanical Dynamics of Control Systems Mechanical Dynamics of Control Servoing Robust Globally Stable

Room 209 Visual and and 2 1 Control Adaptive Approaches New Algebra Control Equations Operator Matrix and Room 210 and 2 to MTNS – 2002 13

Farewell Party F 16:30-18:30 Coffee Break F 14:00-16:00 Lunch Break F 10:30-12:30 Coffee Break F 9:00-10:00 F 8:00-9:00

Anthony Room 101 D'Andrea Raffaello Bloch Circuits Networks Theory Mathematical Genetic Networks Allen Room 102 Tannenbaum of

and I Computation, Applications Systems: Parameter Distributed Control, Part II Stabilization and Systems: Parameter Distributed II Computation, Applications Systems: Parameter Distributed Karl Room 126 Friday, August Kunisch

and and Part Part Methods Theoretic Operator Estimation Control Stochastic Room 129

and 16, 2002 and Experiment Theory, Water Nonlinear Surface Numerics Stability Room 136/138 Waves: Computation and Theory and Systems, Approach A Behavioral Coding Room 208 Control to of Vehicle Coordinated Room 209 Networks Control Stability, Part Input-to-State Applications Control Nonlinear

Room 210 and II MTNS – 2002 14 Schedule of Events

Sunday August 11, 2002 Morning:

18:00—21:00 Welcoming Reception Room: 102, Session: MA1 Center for Continuing Education Chair: Shmuel Friedland, Brian Marcus Title: Capacity of Multi-Dimensional Codes Part I 11:00-13:00 Minicourse on “Capacity of Monday August 12, 2002 Multidimensional Codes”, Shmuel Friedland 8:00-8:30 Room: 101 Welcoming Remarks Room: 126, Session: MA2 Panos J Antsaklis, Director, Center for Applied Mathematics Chair: Michael D. Lemmon Steven Buechler, Title: Control Applications Chair, Department of Mathematics 11:00-11:20 Cancer Treatment Using Mul- Jeffrey C Kantor, tiple Chemotherapeutic Agents Subject Vice President to Drug Resistance, John Westman, Bruce Fabijonas, Daniel Kern, Floyd Hanson 11:25-11:45 Selection of Decentralized Con- trol Configurations Based on Distur- 8:30-9:30 Room: 101 Plenary Talk bance Rejection for Plants with Real In- Bruce Hajek, tegrators, Henning Schmidt A Basket of System Theoretic Problems 11:50-12:10 Synergetic Control of the Un- in Communications stable Two-Mass System, Alexander Kolesnikov 12:15-12:35 Synergetic Control for Elec- 9:30-10:30 Room: 101 Invited Talk tromechanical Systems, Andrey Popov, Roger Brockett, Anatoly Kolesnikov, Gennady Veselov, Alexan- Optimal System Identification for NMR der Kolesnikov, Roger Dougal Applications 12:40-13:00 Modeling of Out-of-Plane Hy- groinstability of Multi-Ply Paperboard, Gianantonio Bortolin, Per-Olof Gutman 9:30-10:30 Room: 102 Invited Talk Hans-Andrea Loeliger, Factor Graphs, Least Squares and Room: 136, Session: MA4 Kalman Filtering Chair: Bill Helton, Andre Ran, Leiba Rodman Title: Matrix and Operator Equations I 9:30-10:30 Room: 136 Invited Talk 11:00-11:30 Noncanonical Almost Periodic Sjoerd Verduyn Lunel, Factorization and Toeplitz Operators Control and Stabilization of Systems with Almost Periodic Symbols, Leiba Rod- with Time Delays man, I. M. Spitkovsky, H. J. Woerdeman MTNS – 2002 15

11:30-12:00 Symmetric Nonsquare Factoriza- Room: 210, Session: MA7 tion of Selfadjoint Rational Matrix Func- Chair: Peter Bauer tions and Algebraic Riccati Inequalities, Title: Adaptive Control A. C. M. Ran, Mark A. Petersen 12:00-12:30 Extremal Problems of Interpola- 11:00-11:20 Gap Metric Robustness of Adap- tion Theory, L. A. Sakhnovich tive Controllers, Mark French 12:30-13:00 Convex Invertible Cones, 11:20-11:40 Adaptive Predictive Control Nevalinna-Pick Interpolation and the Set with Controllers of Restricted Structure, of Lyapunov Solutions, Izchak Lewkowicz, Michael Grimble, Peter Martin Nir Cohen 11:40-12:00 Output Adaptive Model Refer- ence Control of Linear Continuous State- Delay Plant, Boris Mirkin, Per-Olof Gutman 12:00-12:20 A Comparison Between Ro- Room: 208, Session: MA5 bust Adaptive Controllers w.r.t a Non– Chair: Lorenzo Farina, Maria Elena Valcher singular Transient Cost, Ahmad Sanei, Title: Positive Systems Mark French 12:20-12:40 A Manifold Structure on the set 11:00-11:30 Positive Systems in the State of Functional Observers, Jochen Trumpf, Space Approach: Main Issues and Re- Uwe Helmke cent Results, Lorenzo Farina 14:00-15:00 On the Capacity of 2-D Con- 11:30-12:00 Positive Systems in the Behav- strained Codes and Consequences for ioral Approach: Main Issues and Recent Full-Surface Data Channels, William Results, Maria Elena Valcher Weeks 12:00-12:30 Feedback Stabilisation with Posi- tive Control of Dissipative Compartmen- tal Systems, Georges Bastin, A. Provost 12:30-13:00 Feedback Control for a Chemo- Middle: stat with two Organisms, Patrick De Leen- Room: 102, Session: MM1 heer, Hal Smith Chair: Shmuel Friedland, Brian Marcus Title: Capacity of Multi-Dimensional Codes Part II Room: 209, Session: MA6 14:00-15:00 On the Capacity of 2-D Con- Chair: Augusto Ferrante, Michele Pavon strained Codes and Consequences for Title: Control of Quantum Mechanical Full-Surface Data Channels, William Systems Weeks 11:00-11:30 Sufficient Conditions for Control- 15:00-16:00 Counting Independent Sets in lability of Finite Level Quantum Systems The Grid, And Similar Questions, Neil via Structure Theory of Semisimple Lie Calkin Algebras, Claudio Altafini 11:30-12:00 Geometric Control of Quantum Room: 126, Session: MM2 Mechanical Systems in a Noisy Environ- Chair: Wei Lin ment, Domenico D’Alessandro Title: Nonlinear Systems and Control 1 12:00-12:30 Control of Quantum Systems Using Model-based Feedback Strategies, 14:00-14:20 Estimating Generalized Gradi- Augusto Ferrante, Michele Pavon, Giorgio ents of Value Function in Optimal Con- Raccanelli trol Problems for Differential-Difference 12:30-13:00 Quantum Control of Dissipative Inclusions, Leonid Minchenko, Aleksey Volo- Systems, Sonia G. Schirmer, A. I. Solomon sevich MTNS – 2002 16

14:20-14:40 Interconnected Systems of Fliess 15:30-16:00 State Space Method, Explicit So- Operators, W. Steven Gray, Yaqin Li lutions of Scattering Problems, and Non- 14:40-15:00 Controllability Analysis of A linear Integrable Equations, Alexander L. Two Degree of Freedom Nonlinear At- Sakhnovich titude Control System, Jinglai Shen, Amit K. Sanyal, N. Harris McClamroch Room: 208, Session: MM5 15:00-15:20 Sliding Mode Idle Speed Ignition Control Strategies for Automotive En- Chair: Xiaochang Wang gines, Manjit Singh Srai, H. Sindano, N. E. Title: Output Feedback Control of Linear Gough, A. C. Cole Systems 15:20-15:40 Truncation and Approximation 14:00-14:30 Counterexamples to Pole Place- Errors in the Max-Plus Algorithm for H- ment by Real Static Output Feedback, infinity Control, William McEneaney Alex Eremenko, A. Gabrielov 15:40-16:00 Solution of Second Order Lin- 14:30-15:00 Numerical Homotopy Algo- earization, Rajagopalan Devanathan rithms for Satellite Trajectory Control by Pole Placement, Jan Verschelde, Yusong Wang Room: 129, Session: MM3 15:00-15:30 Numerical Schubert Calculus by the Pieri Homotopy Algorithm, Tien-Yien Chair: Giorgio Picci, Augusto Ferrante Li, Xiaoshen Wang, Mengnien Wu Title: Stochastic Systems 1 15:30-16:00 On Minimal Order Decentralized 14:00-14:30 Canonical Correlations Between Output Feedback Pole Assignment Prob- Input and Output Processes of Linear lems, Xiaochang Wang Stochastic Models, Katrien De Cock, Bart De Moor Room: 209, Session: MM6 14:30-15:00 A Regularized Cepstrum and Covariance Matching Method for Chair: Yutaka Yamamoto ARMA(n,m) Design, Per Enquist Title: Optimization and Optimal Control 15:00-15:30 On Some Interpolation Prob- 14:00-14:20 A Jacobi-like Method for the lems, Gyorgy Michaletzky, A. Gombani Indefinite Generalized Hermitian Eigen- 15:30-16:00 Non-regular Processes and Sin- value Problem, Christian Mehl gular Kalman Filtering, Augusto Ferrante, 14:20-14:40 Disturbed Discrete Time Linear- Stefano Pinzoni, Giorgio Picci Quadratic Open-Loop Nash games, Ger- hard Jank, Dirk Kremer 14:40-15:00 Linear Matrix Inequalities for Room: 136, Session: MM4 Global Optimiztion of Rational Func- tions and H2 Optimal Model Reduction, Chair: I. Gohberg, M.A. Kaashoek Dorina Jibetean, Bernard Hanzon Title: State Space Methods for Problems 15:00-15:20 Newton’s Method for Optimiza- in Operator Theory tion in Jordan Algebras, Sandra Ricardo, 14:00-14:30 State Space Methods, Repro- Uwe Helmke, Shintaro Yoshizawa ducing Kernel Spaces and Applications, 15:20-15:40 Non-symmetric Riccati Theory Harry Dym and Linear Quadratic Nash Games., Dirk 14:30-15:00 A Beurling–Lax Type Theorem Kremer, Radu Stefan in the Unit Ball, Daniel Alpay, Aad Dijksma, 15:40-16:00 Some New Results on Linear Jim Rovnyak Quadratic Regulator Design for Lossless 15:00-15:30 A Naimark Dilation Perspective Systems, Maria Gabriella Xibilia, Luigi For- on Positive Real Interpolation, A. Frazho tuna, Giovanni Muscato MTNS – 2002 17

Afternoon: Room: 129, Session: MP3 Room: 102, Session: MP1 Chair: Amarjit Budhiraja Title: Stochastic Control and its Applica- Chair: Sandro Zampieri tions Title: The Interaction of Control, Infor- mation and Communication 16:30-17:00 Nonlinear Filtering in Correlated Noise: a Wiener Chaos Approach, Sergey 16:30-16:50 Minimum Data Rates for Stabil- Lototsky ising Linear Systems with Unknown Pa- 17:00-17:30 Stationary Solutions and For- rameters, Girish Nair, Robin J. Evans, Bj¨orn ward Equations for Controlled and Sin- Wittenmark gular Martingale Problems, Richard H. 16:50-17:10 A Graphical Model Approach to Stockbridge Distributed Control, Sekhar Tatikonda 17:30-18:00 An Investment Model with Liq- 17:10-17:30 Quantized Stabilization of Single- uidity Risk, Hui Wang input Nonlinear Affine Systems, Jialing 18:00-18:30 Mean-Variance Portfolio Selec- Liu, Nicola Elia tion under Markov Regime: Discrete- 17:30-17:50 Distributed Robust Controller time Models and Continuos-time Limits, for Complex Networks, Wing Shing Wong George Yin, X. Y. Zhou 17:50-18:10 Stabilizing Quantized Feedback with Minimal Information Flow: the Scalar Case, Fabio Fagnani, Sandro Zampieri Room: 136, Session: MP4 18:10-18:30 Systems of Dynamics and their Chair: Harry Dym, Heinz Langer Cohomological Invariants, Reuben Rabi, Title: Spaces with Indefinite Metrics and Sanjoy Mitter Inverse 16 :30-17:00 Regular and Singular Point-like Room: 126, Session: MP2 Perturbations of some Differential Oper- Chair: Matthias Kawski ators in Pontryagin Spaces, Aad Dijksma, Title: Nonlinear Systems and Control 2 Yuri Shondin 17:00-17:30 Applications of Spaces with In- 16:30-16:50 Skorokhod-Neumann Boundary definite Metrics, Babak Hassibi Conditions in Robust Queueing Service 17:30-18:00 Sturm-Liouville Inverse Spectral Models, Martin Day Problems with Boundary Conditions De- 16:50-17:10 Optimization Methods for Tar- pending on the Spectral Parameter, Cor- get Problems of Control, Alexander B. nelis van der Mee, Vjacheslav Pivovarchik Kurzhanski, Pravin Varaiya 18:00-18:30 Variational Principles for Block 17:10-17:30 On Optimal Quadratic Lya- Operator Matrices, Christiane Tretter, punov Functions for Polynomial Sys- Heinz Langer, Matthias Langer tems, Graziano Chesi, Alberto Tesi, Antonio Vicino 17:30-17:50 The Maximum Principle for an Room: 209, Session: MP5 Optimal Solution to a Differential In- Chair: Paul Fuhrmann clusion with State Constraints, Aurelian Title: Algebraic Systems Theory Cernea 17:50-18:10 Synergetic Synthesis of Nonlin- 16:30-16:50 Further Results on Intercon- ear Interconnected Control for Turbo- nection and Elimination for Delay- generators, Anatoly Kolesnikov, Andrew Kuz- Differential Systems, Heide Gluesing- menko Luerssen 18:10-18:30 Stabilities and Controllabilities 16:50-17:10 Reduction of Affine Systems of Switched Systems (with Applications on Polytopes, Jan H. van Schuppen, Luc to the Quantum Systems), Leonid Gurvits C.G.J.M. Habets MTNS – 2002 18

17:10-17:30 State Feedback Stabilization 9:00-10:00 Room: 101 Invited Talk with Guaranteed Transient Bounds, Arjan van der Schaft, Fabian Wirth, Diederich Hinrichsen, Elmar Mathematical Theory of Network Mod- Plischke els of Physical Systems 17:30-17:50 Reduction of the Number of Pa- rameters for all Stabilizing Controllers, Kazuyoshi Mori 9:00-10:00 Room: 102 Invited Talk 17:50-18:10 Structural Properties of LTI Sin- Robert J. McEliece, gular Systems by Output Feedback, Runyi Belief Propagation on Partially Ordered Yu, Dianhui Wang Sets 18:10-18:30 On Fliess Models over a Commu- tative Ring, Vakhtang Lomadze 9:00-10:00 Room: 129 Invited Talk Jeff Wood, Room: 210, Session: MP6 Modules and Behaviors: Re-examining Oberst’s Duality Chair: Daniel Liberzon Title: Hybrid Systems and Control 16:30-17:00 Nonlinear and Hybrid Control Morning: via RRTs, Michael Branicky, Michael M. Room: 102, Session: TUA1 Curtiss 17:00-17:30 Reachability Analysis of Hybrid Chair: Heide Gluesing-Luerssen Systems with Linear Dynamics, Mireille Title: Design and Analysis of Block Broucke Codes, Part I 17:30-18:00 Towards the Control of Linear 10:30-11:30 Iterative Decoding and Design of Systems with Minimum Bit-Rate, Joao Codes on Graphs, Pascal O. Vontobel Hespanha, Antonio Ortega, Lavanya Vasude- 11:30-12:00 Codes for Networks, Ralf Koetter van 12:00-12:30 Unitary Constellation Design 18:00-18:30 Control of Hybrid Systems along with Application to Space-time Coding, Limit Cycles, Milos Zefran, Guobiao Song, Guangyue Han, Joachim Rosenthal Francesco Bullo Room: 126, Session: TUA2 Chair: Wijesuriya P. Dayawansa 20:00–22:00 Room: 102 Title: Patterns in Biology 10:30-11:00 Visual Systems, Bijoy Gosh, A. Pol- Workshop on Open Problems in Sys- pitiya tems Theory 11:00-11:30 The Dynamics of Avian Kinesis, Chairs: Vincent Blondel, Roger Brockett Lawrence Schovanec, Alan Barhorst, Sankar Chatterjee 11:30-12:00 Spiral Waves in the Heart, Clyde Martin, P. Marcus 12:00-12:30 Large Amplitude Travelling Tuesday August 13, 2002 Waves in Coupled Oscillater Networks, Wijesura P. Dayawansa, Clyde Martin 8:00-9:00 Room: 101 Plenary Talk

Gilbert Strang, Filtering and Signal Processing MTNS – 2002 19

Room: 129, Session: TUA3 12:10-12:30 Inclusion of Frequency Domain Behaviors, Stephen Prajna, Pablo A. Parrilo Chair: Victor Vinnikov, Joseph A. Ball Title: Minicourse A: Multidimensional Systems 10:30-11:30 Overdetermined Multidimen- sional Systems and Applications, Victor Room: 209, Session: TUA6 Vinnikov 11:30-12:30 Overdetermined Multidimen- sional Systems and Applications, Joseph Chair: Viswanath Ramakrishna A. Ball Title: Quantum Engineering I 10:30-11:10 A Numerical Approach to the Design of Strongly Modulating Pulses to Room: 136, Session: TUA4 Implement Precise Effective Hamiltoni- Chair: Lars Gruene, Fabian Wirth ans for Quantum Information Process- Title: Input-to-State Stability, Part I ing, Timothy Havel, Nicolas Boulant, David G. Cory, Evan M. Fortunato, Marco A. Pravia, 10:30-11:00 Attractors, Input-to-State- Grum Teklemariam Stability, and Control Sets, Fritz Colonius, 11:10-11:50 System Theoretic Aspects of W. Kliemann NMR Spectroscopy , Raimund J. Ober 11:00-11:30 Output-Input Stability of Non- 11:50-12:10 Local and Global Control of Pop- linear Systems and Input/Output Oper- ulation Transfer in Quantum Systems, ators, Daniel Liberzon, Eduardo Sontag Vladimir Malinovsky 11:30-12:00 A Parameter-Robust Observer 12:10-12:30 Hartree-Fock Models in Elec- as an Application of ISS Techniques, tronic Structure Computations, Gabriel Madalena Chaves Turinici 12:00-12:30 Quantitative Aspects of the Input-to-state Stability Property, Lars Gruene

Room: 208, Session: TUA5 Room: 210, Session: TUA7 Chair: Anders Rantzer Title: Linear Systems Chair: Stephen Campbell, Ramine Nikoukhah Title: Robust Estimation, Identification, 10:30-10:50 A New Property of La- and Detection guerre Functions, Luigi Fortuna, Riccardo Caponetto, Mattia Frasca 10:30-11:00 A Survey of Input-Output Meth- 10:50-11:10 Communication-Limited Stabil- ods in Robust Estimation, Babak Hassibi isability of Jump Markov Linear Sys- 11:00-11:30 Robust Least-Squares Filtering tems, Girish Nair, Subhrakanti Dey, Robin with a Relative Entropy Constraint, Evans Bernard Levy, Ramine Nikoukhah 11:10-11:30 Equivalence of Finite Pole 11:30-12:00 Bounding the Solution Set of Assignability of LTI Singular Systems Uncertain Linear Equations: a Convex by Output Feedback, Runyi Yu, Dianhui Relaxation Approach, Giuseppe Calafiore, Wang Laurent El Ghaoui 11:30-11:50 On Kalman Models over a Com- 12:00-12:30 The Design of Auxiliary Sig- mutative Ring, Vakhtang Lomadze nals for Robust Active Failure Detection 11:50-12:10 On Rosenbrock Models over a in Uncertain Systems, Stephen Campbell, Commutative Ring, Vakhtang Lomadze Ramine Nikoukhah MTNS – 2002 20

Middle: 15:30-16:00 Algebraic Algorithm for 2D Sta- bility Test Based on a Lyapunov Equa- Room: 102, Session: TUM1 tion, Minoru Yamada, Li Xu, Osami Saito Chair: Daniel Costello Title: Design and Analysis of Block Codes, Part II Room: 208, Session: TUM4 14:00-14:30 On a Few Classes of Optimal and Near-optimal Polynomial Codes, Nuh Ay- Chair: Joseph A. Ball, Hugo Woerdeman din Title: Recent Developments on Interpola- 14:30-15:00 Building Low-Density Parity- tion and Completion Problems Check Codes with Affine Permuta- 14:00-14:20 Feedback Control for Multidi- tion Matrices, Michael O’Sullivan, Marcus mensional Systems and Interpolation Greferath, Roxana Smarandache Problems for Multivariable Functions, 15:00-15:30 On Plotkin and Elias Bounds for Joseph A. Ball, Tanit Malakorn Codes over Frobenius Rings under the 14:20-14:40 On the Caratheodory-Fejer In- Homogeneous Weight, Marcus Greferath terpolation Problem for Generalized 15:30-16:00 Four and Six-Dimensional Signal Schur Functions, Vladimir Bolotnikov Constellations from Algebraic Lattices, 14:40-15:00 Abstract Interpolation in Scat- Carmelo Interlando, Michele Elia tering Setting, Alexander Kheifets 15:00-15:20 A Convex Optimization Ap- proach to Generalized Moment Prob- Room: 126, Session: TUM2 lems, Anders Lindquist, C. I. Byrnes 15:20-15:40 Extremal Properties of Outer Chair: Raimund Ober Factors, Scott McCullough Title: Immunology 1: Introduction and 15:40-16:00 On the Realization of Inverse Microscopy Stieltjes Functions, E. R. Tsekanovskii, 14:00-14:40 Introduction to Workshop and Sergey Belyi, Seppo Hassi, Henk de Snoo Overview, Raimund Ober 14:40-15:20 T Cell Receptor MHC Interac- tions: An Overview, E. Sally Ward 15:20-16:00 Image Formation and Deconvo- Room: 209, Session: TUM5 lution for 3 Dimensional Microscopy of Cell Samples, Jose Angel Conchello Chair: Damir Arov Title: Control of Distributed Parameter Systems Room: 129, Session: TUM3 14:00-14:30 Optimal Control and Riccati Equations for a Degenerate Parabolic Chair: Krzystof Galkowski, Eric Rogers, Vic- System, Jean-Marie Buchot, Jean-Pierre tor Vinnikov Raymond Title: Multidimensional Systems 1 14:30-15:00 Nonlinear Predictive Control of 14:00-15:00 2D Linear Control Systems - Flexible Manipulator Systems, Alaa Mo- From Theory to Experiment to Theory, hamedy, Andrzej Ordys, Michael Grimble Eric Rogers, Tarek Al-Towlem, James Rad- 15:00-15:30 Furtivity and Masking Problems cliffe, Paul Lewin, Krzysztof Galkowski, David in Acoustics, Francesco Zirilli Owens 15:30-16:00 Approximation of Optimal Con- 15:00-15:30 Stability Analysis of 2D Dynam- trols for Semi-Linear Parabolic PDE by ics in Roessers Model, Tatsushi Ooba, Ya- Solving Hamilton-Jacobi-Bellman Equa- suyuki Funahashi tions, Sophie Gombao MTNS – 2002 21

Room: 210, Session: TUM6 16:30-17:10 Microscopic Investigation of Synapse Formation, Michael Dustin Chair: Anders Lindquist 17:10-17:50 Studying Protein-Protein Inter- Title: Filtering and Identification actions: Biosensor Technology, Peter 14:00-14:20 System Identification of Nonlin- Schuck ear Dynamic Systems with Multiple In- 17:50-18:30 Protein Dynamics Near Mem- puts and Single Output Using Discrete- brane Surfaces: New Aspects of Local Time Volterra Type Equations, Thomas Coupled Reaction and Transport, Nancy Treichl, Stefan Hofmann, Dierk Schr¨oder L. Thompson 14:25-14:45 Data Driven Local Coordinates, Thomas Ribarits, Manfred Deistler, Bernard Hanzon Room: 129, Session: TUP3 14:50-15:10 Using Rank Order Filters to De- Chair: Krzystof Galkowski, Eric Rogers, Vic- compose the Electromyogram, Dawnlee tor Vinnikov Roberson, Cheryl Schrader Title: Multidimensional Systems 2 15:15-15:35 Conditioning Analysis of a Con- tinuous Time Subspace-Based Model 16:30-17:00 State Representation of nD Be- Identification Algorithm, Juan Carlos haviors, Isabel Br´as,Paula Rocha Martinez-Garcia, G.H. Sal`azar-Silva,R. Gar- 17:00-17:30 The Bang-Bang Principle for the rido Goursat-Darboux Problem, Dariusz Id- 15:40-16:00 On Model and State Estimation czak under Mixed Uncertainty, Irina Digailova, 17:30-18:00 Elimination of Anticipation of Alexander B.Kurzhanski Singular 2D Roesser Model, Tadeusz Kac- zorek 18:00-18:30 Difference Equations and n-D Discrete Systems, Jiri Gregor Afternoon: Room: 102, Session: TUP1 Room: 136, Session: TUP4 Chair: Heide Gluesing-Luerssen Title: Convolutional Codes Chair: Matthias Kawski Title: Nonlinear Systems and Control 3 16:30-17:00 Construction and Decoding of Strongly MDS Convolutional Codes, Rox- 16:30-16:50 Disturbance Attenuation for a ana Smarandache, Heide Gluesing-Luerssen, Class of Nonlinear Systems by Output Joachim Rosenthal Feedback, Wei Lin, Xianqing Huang, Chun- 17:00-17:30 On Observers and Behaviors, jiang Qian Paul A. Fuhrmann 16:50-17:10 A Linear Controller for a Mul- 17:30-18:00 On the Convergence of Non- tifrequency Model of a Pulse-Width- systematic Turbo Codes, Daniel Costello Modulated Cuk Converter, Yusuf Fuad, Jr., Adrish Banerjee, Francesca Vatta, Bartolo J.W. van der Woude, W.L. de Koning Scanavino 17:10-17:30 Synergetic Synthesis of Nonlin- 18:00-18:30 Some Small Cyclic Convolu- ear Kinematics Regulators for Mobile tional Codes, Heide Gluesing-Luerssen, Wi- Robots, Boris Topchiev land Schmale, Melissa Striha 17:30-17:50 On the Convergence of a Feed- back Control Strategy for Multilevel Quantum Systems, Paolo Vettori Room: 126, Session: TUP2 18:10-18:30 Global Output Feedback Control Chair: Raimund Ober with Disturbance Attenuation for a Class Title: Immunology 2: Microscopy and of Nonlinear Systems, Xianqing Huang, Wei Biophysics Lin MTNS – 2002 22

Room: 209, Session: TUP5 20:00–22:00 Room: 102 Workshop on Open Problems in Sys- Chair: Ruth Curtain tems Theory Title: Infinite Dimensional Systems Chairs: Anders Rantzer, Eduardo Sontag and 16:30-17:00 Observability Analysis of a Non- Jan C. Willems linear Tubular Reactor, Cedric Delattre, Denis Dochain, Joseph Winkin 17:00-17:30 A Hilbert Space Approach to Self-Similar Systems, Mamadou Mboup Wednesday August 14, 2002 17:30-18:00 Boundary Observability in the Quasi-Static Thermoelastic Contact 8:00-9:00 Room: 101 Plenary Talk Problem, Michael Polis, Irina Sivergina J. William Helton, 18:00-18:30 Modeling Distributed Parame- Manipulating Matrix Inequalities Auto- ter Systems with Discrete Element Net- matically works, Fabien Soulier, Patrick Lagonotte

9:00-10:00 Room: 101 Invited Talk Jan C. Willems, Dissipative Distributed Systems

Room: 210, Session: TUP6 9:00-10:00 Room: 102 Invited Talk Albert-Laszlo Barabasi, Chair: Avraham Feintuch The Architecture of Complexity: Emer- Title: Robust and H-Infinity Control and gence of Scaling in Complex Networks Estimation 16:30-16:50 Simultaneous Robust Regulation 9:00-10:00 Room: 136 Invited Talk and Robust Stabilization with Degree Constraint, Ryozo Nagamune Knut Hueper, 16:50-17:10 Closed-Loop Structure of Dis- The Dynamics of Matrix Eigenvalue Al- creteTime H-infinity Controller, Wa- gorithms ree Kongprawechnon, Shun Ushida, Hidenori Kimura 17:10-17:30 On a Recursive State- Morning: space Method for Discrete-time H2- Approximation, Ralf Peeters, Martine Olivi, Room: 102, Session: WA1 Bernard Hanzon Chair: Raimund Ober 17:30-17:50 PID Robust Control via Genetic Title: Immunology 3: Structure and Ki- Algorithms and Integral Criteria Mini- netics mization, Catalin Nicolae Calistru, Oana Ge- man 10:30-11:10 Geometrical Methods in Struc- 17:50-18:10 MIMO Systems Properties tural Molecular Biology, Timothy F. Havel Preservation under SPR Substitutions, 11:10-11:50 Kinetic aspects of TcR-MHC and Juan Carlos Martinez-Garcia, G. Fern`andez- Antibody-Antigen Interactions, Jefferson Anaya Foote 18:10-18:30 State Feedback Mixed H2/H- 11:50-12:30 Biophysical Considerations of Infinity Problem for Linear Systems with T-Cell Receptor-Peptide/MHC Interac- Finite Jumps, Vasile Dragan, Adrian Stoica tions, Brian M. Baker MTNS – 2002 23

Room: 126, Session: WA2 10:50-11:10 Analysis of Smith-Type Methods for Lyapunov Equations and Balanced Chair: Martin Haenggi Model Reduction, Dan Sorensen Title: Computer Networks 11:10-11:30 Krylov Subspace Techniques for 10:30-10:50 Min-Plus System Theory Ap- Reduced Order Modeling of Nonlinear plied to Communication Networks, Dynamical System, Daniel Skoogh, Zhaojun Patrick Thiran, Jean-Yves Le Boudec Bai 10:55-11:15 Elements of Probabilistic Net- 11:30-11:50 Model Reduction of Second Or- work Calculus for Packet Scale Rate der Systems, Guarantee Nodes, Milan Vojnovic, Jean- Younes Chahlaoui, D. Lemonnier, K. Meerbergen, Yves Le Boudec A. Vandendorpe, P. Van Dooren 11:20-11:40 Statistical Performance Analysis 11:50-12:10 Model Reduction via Tangential of a Generalized Processor Sharing Sys- Interpolation, Antoine Vandendorpe, K. Gal- tem by Using Large Deviations, Min Xie, livan, P. Van Dooren Martin Haenggi 11:45-12:05 Resource Allocation and Con- gestion Control in Distributed Sensor Networks - a Network Calculus Ap- Room: 208, Session: WA5 proach, Jinsong Zhang, Kamal Premaratne, Chair: Daniel Alpay, Yuli Eidelman Peter Bauer Title: Time-Varying Systems and Numer- 12:10-12:30 Optimal Media Streaming in a ical Problems Rate-Distortion Sense For Guaranteed Service Networks, Olivier Verscheure, Pas- 10:30-11:00 Unbounded J-inner Sections, cal Frossard Patrick Dewilde, Daniel Alpay 11:00-11:30 Linear Time-Varying Darlington Synthesis, Avraham Feintuch Room: 129, Session: WA3 11:30-12:00 Reduction to System Methods for Inversion of Diagonal Plus Semisep- Chair: Eric Rogers arable Operator Matrices, Yuli Eidelman, Title: Minicourse B: Multidimensional Israel Gohberg Systems 10:30-11:10 Recent Results on Multidimen- sional Behaviors, Eva Zerz 11:10-11:50 Motivation and General Con- Room: 209, Session: WA6 cepts in Behavioral Systems, Jan C. Willems Chair: Erik Verriest 11:50-12:30 Similarities/Differences between Title: Nonlinear Systems and Control 4 the Behavioral Approach for Multidi- 10:30-11:00 Parameter Tuning of a Non Inte- mensional versus Delay-Differential Sys- ger Order PID Controller, Luigi Fortuna, tems, Heide Gluesing-Luerssen Riccardo Caponetto, Domenico Porto 11:00-11:30 Nonlinear Discrete-Time Ob- server Design with Linearizable Er- Room: 136, Session: WA4 ror Dynamics, MingQing Xiao, Nikolaos Kazantzis, Costas Kravaris, Arthur J Krener Chair: Paul Van Dooren 11:30-12:00 Analysis of Periodic Solutions of Title: Model Reduction Tapping-Mode AFM: An IQC Approach, 10:30-10:50 An Overview of Model Reduc- Murti Salapaka, Abu Sebastian tion Methods for Large-Scale Dynamical 12:00-12:30 Bifurcations of the Controlled Systems, Thanos Antoulas Escape Equation, Tobias Gayer MTNS – 2002 24

Room: 210, Session: WA7 Room: 126, Session: WM2

Chair: Rodolphe Sepulchre Chair: Aleksandar Kavcic Title: Discrete Event and Hybrid Systems Title: Control and Communications 10:30-10:50 Switched Systems that are Peri- 14:00-14:20 Feedback Capacity, Sekhar odically Stable may be Unstable, Jacques Tatikonda, Sanjoy Mitter Theys, Vincent Blondel, Alexander Vladimirov 14:20-14:40 Sum-Product Algorithm and 10:50:11:10 The Servo Problem for Piece- Feedback Capacity, Shaohua Yang, Aleksan- wise Linear Systems, Stefan Solyom, Anders dar Kavcic Rantzer 14:40-15:00 Kalman Filtering, Factor 11:10-11:30 Stability of Hybrid Control Graphs, and Electrical Networks, Pas- Systems Based on Time-State Control cal O. Vontobel, Dani Lippuner, Hans-Andrea Forms, Yoshikatsu Hoshi, Mitsuji Sampei, Loeliger Shigeki Nakaura, 15:00-15:20 Kalman Filtering Applied to 11:30-11:50 Discrete-Time Modeling and Timing Recovery in Tracking Mode, Panu Analysis of Pulse-Width-Modulated Chaichanavong, Brian Marcus Switched Power Converters, Willem L. 15:20-15:40 Lower Bounds for the Perfor- De Koning mance of Iterative Timing Recovery at 11:50-12:10 On the Control of the Reso- low SNR, Aravind Nayak, J. Barry, S. nant Converter: A Hybrid-Flatness Ap- McLaughlin proach, Hebert Sira-Ramirez, Ramon Silva- 15:40-16:00 Classical Capacity of Quantum Ortigoza Channels, Navin Khaneja 12:10-12:30 Controllability of Periodically Switched Linear Systems with Delay in Control, Guangming Xie, Long Wang, Yijing Room: 129, Session: WM3 Wang Chair: Krzystof Galkowski, Eric Rogers, Vic- tor Vinnikov Title: Multidimensional Systems 3 14:00-15:00 Conservative Multidimensional Systems: A Survey, Joseph A. Ball Middle: 15:00-15:30 On J-Conservative Scattering nD System Realizations, Dmitry Kalyuzhniy- Room: 102, Session: WM1 Verbovetzky 15:30-16:00 Factorization of M-D Polynomial Chair: Raimund Ober Matrices for Design of M-D Multirate Title: Immunology 4: Diffusion and Mod- Systems, Mikhail Tchobanou, Cynthia Wood- elling burn 14:00-14:40 Measuring Lateral Diffusion and Associations of MHC Molecules in Mem- branes of the ER and at the Cell Surface, Room: 136, Session: WM4 Michael Edidin Chair: Uwe Helmke 14:40-15:20 A Computational Model for T Title: Control and Computation Cell Receptor Signal Integration, Mark Alber, Arancha Casal, Cenk Sumen, Tim 14:00-14:30 Continuation of Eigendecomposi- Reddy, Mark Davis, Peter Lee tions, Luca Dieci 15:20-16:00 Immunological Synapse Forma- 14:30-5:00 Numerical Solution of Euclidean tion: A Crossroad of Physical Chemistry Balanced Norm Realizations via Gradi- and Cell Biology, Arup K. Chakraborty ent Flows, N. Del Buono, L. Lopez MTNS – 2002 25

15:00-15:30 Controllability of the QR Algo- 15:00-15:30 The Controlled Composition rithm on Hessenberg Flags, Uwe Helmke, Analysis of Hybrid Automata, Ying Shang, Jens Jordan M.D. Lemmon 15:30-16:00 The Continuos-Time Rayleigh 15:30-16:00 Monitoring and Diagnosis of Quotient Flow on the Grassmann Mani- Hybrid Systems Using Particle Filter- fold, Rodolphe Sepulchre, P.-A. Absil, R. Ma- ing Methods, Xenofon Koutsoukos, James hony Kurien, Feng Zhao

Room: 209, Session: WM5 Room: 208, Session: WM7

Chair: Anthony Bloch Chair: Giorgio Picci Title: Algebraic and Differential Geome- Title: Stochastic Systems 2 try in Systems Theory 14:00-14:30 State Space Realization of Ran- 14:00-14:20 Hamiltonian Structure of the dom Processes with Feedback, Giorgio Algebraic Riccati Equation and its In- Picci, Alessandro Chiuso finitesimal V-Stability, Nanaz Fathpour, 14:30-15:00 Approximate Realization of Hid- Edmond A. Jonckheere den Markov Chains, Lorenzo Finesso 14:20-14:40 Global Transformation of Non- 15:00-15:30 Random Sampling of a linear Dynamic Systems into Canonical Continuous-Time Stochastic Dynamical Forms, Anna Michtchenko, Aleksey Zhirabok System, Mario Micheli, Michael I. Jordan 14:40-15:00 A Lie-Group Approach for Non- 15:30-16:00 The Hilbert Space of an Ergodic linear Dynamic Systems Described by Sequence, Giorgio Picci Implicit Ordinary Differential Equations, Kurt Schlacher, Andreas Kugi, Kurt Zehetleit- ner Afternoon: 15:00-15:20 Quotients of Fully Nonlinear Control Systems, Paulo Tabuada, George J. Room: 102, Session: WP1 Pappas 15:20-15:40 The Wave Equation as a Port- Chair: Raimund Ober Hamiltonian System, and a Finite Di- Title: Immunology 5: Cellular Aspects mensional Approximation, Viswanath Ta- 16:30-17:30 Staining Antigen Specific CD4+ lasila, Goran Golo, Arjan van der Schaft T -Cells with Class II MHC Oligomers, 15:40-16:00 Pseudo Balancing for Discrete Lawrence Stern Nonlinear Systems, Erik Verriest 17:10-17:40 The Roles of Serial Engagement and Kinetic Proofreading in Peptide- Induced T-Cell Activation, Dan Coombs, Room: 210, Session: WM6 Carla Wofsy, Byron Goldstein Chair: Panos Antsaklis, Anthony Michel Title: Hybrid Control System Analysis, Room: 126, Session: WP2 Synthesis and Diagnosis Chair: Roxana Smarandache 14:00-14:30 Partial Stability of Dynamical Title: Cryptography Systems, Ye Sun, A.N. Michel, A.P. Molchanov 16:30-17:00 A High-Speed Processing for 14:30-15:00 An Approach to General RSA Cryptograms Using High-Radix Switched Linear Quadratic Optimal Con- Signed-Digit Numbers and a New Algo- trol Problems with State Jumps, Xuping rithm of Modulo Operation, Yoshinori Fu- Xu, Panos Antsaklis jisawa, Yasushi Fuwa MTNS – 2002 26

17:00-17:30 On the Rational Cubic Curve Room: 208, Session: WP5 Cryptosystems, Xiaochang Wang, Heather Chair: J. William Helton Henkel Title: Expressing Polynomials as Sums of 17:30-18:00 Public Key Cryptography Based Squares Together with Applications on Simple Modules over Simple Rings, Gerard Maze, Christopher Monico, Joan-Josep 16:30-17:10 How to Write a Polynomial as a Climent, Joachim Rosenthal Sum of Squares of Polynomials, and Why You’d Want to Do So, Bruce Reznick 17:10-17:30 Applications of Our Newfound Facility in Expressing Polynomials as Room: 129, Session: WP3 Sums of Squares., Pablo A. Parrilo 17:30-17:50 Reduced Representations of Pos- Chair: Krzystof Galkowski, Eric Rogers, Vic- itive Polynomials, Mihai Putinar tor Vinnikov 17:50-18:10 Recent Progress in Polynomial Title: Multidimensional Systems 4 Optimization, Ruchira Datta 16:30-17:00 Spatial Restoration with Re- 18:10-18:30 Bounding Linear PDEs via duced Boundary Error, Nirmal Bose, Jae- Semidefinite Optimization, Constantine hoon Koo Caramanis, Dimitris Bertsimas 17:00-17:30 On Succesive Packing Approach to Multidimensional (M-D) Interleaving, Room: 209, Session: WP6 Sankar Basu, Xi Min Zhang, Yun Q. Shi 17:30-18:00 Matrix Functions in Homomor- Chair: Viswanath Ramakrishna phic Signal Processing, Eduard Krajnik Title: Quantum Engineering II 18:00-18:30 Cellular Automata in Image 16:30-17:10 Optimal Control of Laser Cool- Processing, Adriana Popovici, Dan Emanuel ing: A Theory of Purity Increasing Popovici Transformations, David Tannor, Shlomo Sklarz 17:10-17:30 Controllability of Pairs of Cou- pled Quantum Dots, Viswanath Ramakr- Room: 136, Session: WP4 ishna Chair: Biswa Nath Datta, Floyd B. Hanson 17:30-17:50 Constructive Control of Quan- Title: Large-Scale Computations in Con- tum Systems, trol Sonia Schirmer, A.D. Greemtree 17:50-18:10 Use of Wei-Norman Formulae 16:30-16:50 Projection Methods for Reduced and Parameter Differentiation in Quan- Order Modeling with Guaranteed Stabil- tum Control,Claudio Altafini ity, Thanos Antoulas 18:10-18:30 Control of Quantum Mechani- 16:55-17:15 Computational Methods for cal Systems with Minimum Number of Portfolio and Consumption Policy Opti- Switches, Domenico D’ Alessandro mization in Log-Normal Diffusion, Log- Uniform Jump Environments, Floyd B. Hanson, J. J. Westman 17:20-17:40 Partial Eigenvalue Assignment in 20:00–22:00 Room: 102 Linear Systems: Existence, Uniqueness Panel Discussion on Future Directions and Numerical Solution, Biswa N. Datta, of Research and Teaching in Mathe- Daniil Sarkissian matical Control and Systems Theory, 17:45-18:05 Model Reduction via an Explic- itly Restarted Lanczos Algorithm, Vasil- Biswa Datta, Organizer. ios Papakos, Imad M. Jaimoukha MTNS – 2002 27

Thursday August 15, 2002 10:55-11:15 Explicit Formulae for J-Spectral Factors for Well-Posed Systems, Ruth 8:00-9:00 Room: 101 Plenary Talk Curtain, Amol J. Sasane 11:20-11:40 A Riccati Equation Approach Eduardo Sontag, to the Standard Infinite-Dimensional H- On Systems Molecular Biology and Con- Infinity Problem, Kalle M. Mikkola, Olof trol Theory Staffans 11:45-12:05 Sub-optimal Hankel Norm Ap- 9:00-10:00 Room: 126 Invited Talk proximation for the Wiener Class, Orest Iftime, Amol Sasane Olof Staffans, 12:10-12:30 LQG Balancing in Infinite Di- Passive and Conservative Infinite- mensions, Mark R. Opmeer, Ruth Curtain Dimensional Impedance and Scattering Systems (from a Personal Point of View) Room: 129, Session: THA3 9:00-10:00 Room: 129 Invited Talk Chair: J.M. (Hans) Schumacher Wolfgang J. Runggaldier, Title: Systems and Control Theory in Fi- On Stochastic Control in Finance nance and Insurance 1 10:30-11:30 Control and Financial Engineer- 9:00-10:00 Room: 136 Invited Talk ing, J. M. (Hans) Schumacher 11:30-12:00 Dynamic Risk Sensitive Asset Matthias Heinkenschloss, Management With Nonnegative Multi- Domain Decomposition Approaches for ple Factor Constraints, Arunabha Bagchi, the Optimization of Distributed Systems K. Suresh Kumar 12:00-12:30 A Filtered No-arbitrage Model for Term Structures from Noisy Data, Morning: Andrea Gombani, Stefan R. Jaschke, Wolfgang J. Runggaldier Room: 102, Session: THA1 Chair: Mark Alber Title: Complex Networks and Biological Room: 136, Session: THA4 Applications 1 Chair: David Nicholls 10:30-11:10 The Spread of Infections on So- Title: Fully Nonlinear, Three- cial Networks, Mark Newman Dimensional, Surface Water Waves in 11:10-11:50 Information Theory Aspects of Arbitrary Depth Signal Transduction and Gene Regula- 10:30-11:00 Experiments on Deep-Water tion, Andrea Levchenko Waves with Two-Dimensional Surface Patterns, Diane Henderson Room: 126, Session: THA2 11:00-11:30 Instability of Bounded Solutions of the 2-D Cubic Nonlinear Schrodinger Chair: Ruth Curtain, Olof Staffans Equation, John Carter Title: Distributed Parameter Systems: 11:30-12:00 Computing (quasi) Periodic Theory Part I Waves in Shallow Water, Bernard Decon- 10:30-10:50 Some Results on the Theory of inck Linear Time-Invariant Dissipative Sys- 12:30-13:00 Mathematical Models of Deep- tems with Hilbert and Pontryagin State Water Waves with two-Dimensional Sur- Spaces, Damir Arov face Patterns, Harvey Segur MTNS – 2002 28

Room: 208, Session: THA5 11:30-12:00 The Symmetric Linear Matrix Equation, Martine C. B. Reurings Chair: Krzystof Galkowski, Eric Rogers, Vic- 12:00-12:30 Investigating Duality on Stability tor Vinnikov Conditions, Mauricio de Oliveira Title: Multidimensional Systems 5 10:30-11:00 Robust Stability and Stabiliza- tion of n-D Systems, Jiang-Qian Ying, Li Xu, Masayuki Kawamata Middle: 11:00-11:30 Succesive stabilization of a class Room: 102, Session: THM1 of 2D systems, Krzysztof Galkowski, Bartek Sulikowski, Eric Rogers, David H. Owens Chair: Mark Alber 11:30-12:00 Optimal Control for a Class of Title: Complex Networks and Biological Differential Linear Repetitive Processes, Applications 2 Eric Rogers, S. Dymkou, M. Dymkov, K. 14:00-14:40 Synchronization of Oscillators in Galkowski, D. H Owens Small World Systems, Lou Pecora 12:00-12:30 Relation between Eigenvalues 14:40-15:20 Intracellular signaling is depen- and Singular Values in the Problem of dent on the cytoskeleton. Evidence from Stability Maintenance of Ellipsoidal Es- proteomics., Gabor Forgas timates, Taalaibek A. Akunov, Anatoly V. 15:20-16:00 The Role of Scale-free Connec- Ushakov tivity Patterns in Spreading Phenomena, Alessandro Vespignani Room: 209, Session: THA6

Chair: Koichi Hashimoto Room: 126, Session: THM2 Title: Globally Stable Robust Visual Ser- voing Chair: Ruth Curtain, Olof Staffans Title: Distributed Parameter Systems: 10:30-11:00 Keeping Features in the Cam- Theory Part II era’s Field of View: a Visual Servoing Strategy, Graziano Chesi, K. Hashimoto, D. 14:00-14:30 Zeros of SISO Infinite- Prattichizzo, A. Vicino Dimensional Systems, Kirsten Morris, 11:00-11:30 Binocular Visual Servoing with a Richard Rebarber Limited Field of View, Noah Cowan 14:30-15:00 Stabilizability of Systems with 11:30-12:00 Visual Servoing with Dynamics: Signals in `2(Z), Birgit Jacob Control of an Unmanned Blimp, Jim Os- 15:00-15:30 Stability and Boundedness of trowski Continuous and Discrete-Time Systems, 12:00-12:30 Enlarging the Stable Region of Hans Zwart, B.Z. Guo Image Based Control by Path Planning, 15:30-16:00 Coprimeness Conditions for Youcef Mezouar Pseudorational Transfer Functions, Yu- taka Yamamoto

Room: 210, Session: THA7 Room: 129, Session: THM3 Chair: Bill Helton, Andre Ran, Leiba Rodman Title: Matrix and Operator Equations II Chair: J.M. (Hans) Schumacher Title: Systems and Control Theory in Fi- 10:30-11:00 Noncommutative Convexity of nance and Insurance 2 Functions and Sets, J. William Helton 11:00-11:30 Symmetry Groups, Semidefinite 14:00-15:00 Ruin Probabilities Minimization Programming,and Sums of Squares, Pablo and Dividend Distribution Optimization A. Parrilo in Diffusion Models, Michael Taksar MTNS – 2002 29

15:00-15:30 Continuous-Time Mean-Variance 14:20-14:40 Hamiltonian Attitude Dynamics Portfolio Selection with Markov- for a Spacecraft with a Point Mass Oscil- Modulated Market Parameters, Xun Yu lator, Craig Woolsey Zhou 14:40-15:00 Controllable Kinematic Reduc- 15:30-16:00 Stock Selection Based on Clus- tions for Mechanical Systems: Concepts, ter and Outlier Analysis, Steven Craighead, Computational Tools, and Examples, An- Bruce Klemesrud drew Lewis, Francesco Bullo, Kevin M. Lynch 15:00-15:20 Matching and Stabilization of Room: 136, Session: THM4 Linear Mechanical Systems, Dimitri Zenkov Chair: Paul Van Dooren 15:20-15:40 Matching and Stabilization of Title: Robust Control and Linear Matrix Constrained Systems, Guido Blankenstein Inequalities 15:40-16:00 Extremal Flows on Stiefel Man- 14:00-14:30 Linear Matrix Inequalities in Ro- ifolds, and Riemannian Potatoes, Peter bust Control: A Brief Survey, Venkatara- Crouch, Anthony M. Bloch manan Balakrishnan 14:30-15:00 Periodic Multirate Systems, nu- Gap and Robust Stabilization, Li Qiu, Li Room: 210, Session: THM7 Chai 15:00-15:30 Spectral Factorization and Sums Chair: Jan van Schuppen of Squares via Semidefinite Program- Title: Control and Algebra ming, Hugo Woerdeman 14:00-14:30 Control and Algebra - An Intro- 15:30-16:00 Robustness Analysis via Stabil- duction, Jan H. van Schuppen ity Radii, Spectral Value Sets and mu- 14:30-15:00 Towards an Algebraic Systems Functions, Michael Karow Theory of Hybrid Systems, George J. Pap- pas Room: 208, Session: THM5 15:00-15:30 The Category of a Affine Connec- tion Control Systems, Andrew Lewis Chair: Maria Elena Valcher 15:30-16:00 Coalgebra and Supervisory Con- Title: The Behavioral Approach to Dy- trol with Partial Observations, Jan namic Systems Komenda 14:00-14:30 Deterministic Kalman Filtering, Jan C. Willems 14:30-15:00 Over-Determined Systems, Eva Zerz Afternoon: 15:00-15:30 Regular Implementability nD Behaviors, Paula Rocha Room: 102, Session: THP1 15:30-16:00 Cones of Trajectories as Sub- sets of Linear Systems: the Autonomous Chair: Mark Alber Case, Andrea Morettin Title: Complex Networks and Biological Applications 3 Room: 209, Session: THM6 16:30-17:00 Connections Matter: A Boolean Model for the Segment Polarity Network Chair: Naomi Leonard of Drosophila Melanogaster, Reka Albert Title: Control and Dynamics of Mechani- 17:00-17:30 Modeling Mesenchymal Con- cal Systems I densation during Limb Chondrogenesis, 14:00-14:20 Composition of Dirac Structures Gilberto Tomas and Control of Port-Hamiltonian Sys- 17:30-18:00 Classification of scale-free net- tems, Arjan van der Schaft, J. Cervera works, Byungnam Kahng MTNS – 2002 30

18:00-18:30 Prediction of Protein Essentiality 17:00-17:30 Structured LDPC Codes, Amin Based on Genomic Data, Hawoong Jeoong, Shokrollahi Zoltan N. Oltvai and Albert-Laszlo Barabasi 17:30-18:00 Efficient Matrix Computations in Wideband Communications, Patrick Room: 126, Session: THP2 Dewilde, Lang Tong, Alle-Jan van der Veen 18:00-18:30 Stable Factorization of Han- Chair: Kirsten Morris, Olof Staffans kel and Hankel-like Matrices, Vadim Ol- Title: Distributed Parameter Systems: shevsky, Michael Stewart Stabilization and Control, Part I 16:30-17:00 Reciprocals of Regular Linear Room: 209, Session: THP5 Systems: a Survey., Ruth Curtain 17:00-17:30 H-infinity Control of Acoustic Chair: Naomi Leonard Noise in a Duct with a Feedforward Con- Title: Control and Dynamics of Mechani- figuration, Kirsten Morris cal Systems II 17:30-18:00 Positivity and Dissipativity of 16:30-16:50 On the Ball and Beam Prob- Oscillating Diffusive Filters, Application lem: Regulation with Guaranteed Tran- to the Stability of Coupled Systems, G. sient Performance and Tracking Periodic Dauphin, Denis Matignon Orbits, Romeo Ortega, Fabio Gomez-Estern, 18:00-18:30 Can Positive Pseudo-Differential Javier Aracil, Francisco Gordillo Operators of Diffusive Type Help Stabi- 16:50-17:10 Reduction of Controlled La- lize Unstable Systems?, Denis Matignon grangian Systems with Symmetries, Dong Eui Chang Room: 129, Session: THP3 17:10-17:30 Constrained Mechanical Systems with Impacts, Patrick Hagerty Chair: Tyrone Duncan 17:30-17:50 Adjoints of Hamiltonian Systems Title: Stochastic Theory and Applications and Iterative Learning Control, Kenji Fu- 16:30-17:00 An Approach to Stochastic In- jimoto, Toshiharu Sugie tegration for Fractional Brownian Mo- 17:50-18:10 Controllability of Mechanical tion in a Hilbert Space, Tyrone Duncan, B. Systems with Constraints and Symme- Pasik-Duncan, J. Jakubowski try, Jorge Cortes, Sonia Mart´inez, Jim P. Os- 17:00-17:30 A Class of Tractable Partially trowski, Hong Zhang Observed Discrete Stochastic Games, 18:10-18:30 The Use of Information in Swarm William McEneaney Motions of Autonomous Vehicles, John 17:30-18:00 Hybrid Stock Models and Pa- Baillieul rameter Estimation, George Yin, Q. Zhang, K. Yin Room: 210, Session: THP6 18:00-18:30 Jump-Diffusion Stock Return Models in Finance: Stochastic Process Chair: Jan Willem Polderman Density with Uniform-Jump Amplitude, Title: New Approaches to Adaptive Con- Floyd B. Hanson, J. J. Westman trol 16:30-16:50 Cautious Hierarchical Switching Room: 136, Session: THP4 Control of Stochastic Linear Systems, Marco Campi, Jaoa Hespanha, M. Prandini Chair: Georg Heinig, Vadim Olshevski 16:50-17:10 Strong Robustness in Multi- Title: Computational Methods for Struc- Phase Adaptive Control: the Basic tured Matrices and Applications Scheme, Maria Cadic, Jan Willem Polderman 16:30-17:00 Split Algorithms for Toeplitz 17:10-17:30 Near Optimal LQR Performance and Toeplitz-plus-Hankel Matrices, Georg for Uncertain First Order Systems, Daniel Heinig Miller, Li Luo MTNS – 2002 31

17:30-17:50 Self-Tuning Control for Polyno- Morning: mial Systems: an Algorithmic Perspec- tive, Iven Mareels Room: 102, Session: FA1 17:50-18:10 Geometry of Adaptive Control, Part II: Optimization and Geodesics, Fe- Chair: Reinhard Laubenbacher lipe Pait, Diego Colon Title: Genetic Networks 18:10-18:30 Two Scale High Gain Adaptive 10:30-11:00 Biochemistry by Numbers: Mod- Control, Jan Willem Polderman, Iven Ma- eling, Signaling and Genetic Networks, reels Pedro Mendes, Alberto de la Fuente, Paul Brazhnik, Stefan Hoops 11:00-11:30 Designer Gene Networks, Mads Kaern, James J. Collins 19:00—22:00 Banquet Dinner 11:30-12:00 Function, Design, and Gene Cir- cuitry, Michael A. Savageau Center for Continuing Education 12:00-12:30 Comparative analysis of mathe- matical models of intracellular networks, Vassily Hatzimanikatis, Amit Mehra, Michael Beste

Friday August 16, 2002 Room: 126, Session: FA2 8:00-9:00 Room: 101 Plenary Talk Chair: Belinda King, Kirsten Morris Anthony Bloch, Title: Distributed Parameter Systems: Conservative and Dissipative Dynamics Stabilization and Control, Part II in Classical and Quantum Systems. 10:30-10:50 An Example of Output Regu- lation for Distributed Parameter Sys- tems with Infinite Dimensional Exosys- 9:00-10:00 Room: 101 Invited Talk tem, David Gilliam, Christopher I. Byrnes, Jeff B. Hood, Victor I. Shubov Raffaello D’Andrea, 10:50-11:10 Control of Systems with In- A State Space Approach to Control of finitely Many Unstable Modes and Spatially Interconnected Systems Strongly Stabilizing Controllers Achiev- ing a Desired Sensitivity, Suat G¨um¨u¸ssoy, Hitay Ozbay,¨ 9:00-10:00 Room: 102 Invited Talk 11:10-11:30 Receding Horizon Control and Reduced-Order Methods, Ito Kazufumi Allen Tannenbaum, 11:30-11:50 Some Problems of Control for Controlled Active Vision in Image Nonlinear Partial Differential Equations, Guided Surgery and Therapy. David Russell 11:50-12:10 Global Stabilization of Systems of Partial Differential Equations Us- 9:00-10:00 Room: 126 Invited Talk ing Finite Dimensional Controllers, Igor Mezic Karl Kunisch, 12:10-12:30 Ouput Regulation of Nonlinear From Viscoelastic Fluids to Constrained Systems with State Delay, Emilia Fridman Optimal Control MTNS – 2002 32

Room: 129, Session: FA3 Room: 208, Session: FA5

Chair: Wolfgang Kliemann Chair: Harry Trentelman Title: Stochastic Control and Estimation Title: A Behavioral Approach to Systems, 10:30-10:50 Algebraic Optimization Tech- Control and Coding Theory niques for the Estimation of Zero-Beta 10:30-10:50 A Behavioral Approach to List Pricing Models, Bernard Hanzon Decoding, Jan Willem Polderman, Margreta 10:50-11:10 Trajectory Planning Under a Kuijper Stochastic Uncertainty, Ulf J¨onsson,Clyde 10:55-11:15 Linear Hamiltonian systems, Martin, Yishao Zhou Paolo Rapisarda, H.L. Trentelman 11:10-11:30 An Addendum to the Problem 11:20-11:40 Approximate Time-Controllability of Stochastic Observability, Vasile Dragan, versus Time-Controllability, Amol Sasane, Teodor Morozan M.K. C¸amlibel 11:30-11:50 Combined Optimization of Port- 11:45:12:05 On a Class of Time-Varying Be- folio and Risk Exposure of an Insur- haviors, Madhu Belur, M.K. C¸amlibel, A.J. ance Company, Daniel Cajueiro, Takashi Sasane, J.C. Willems Yoneyama 12:10-12:30 Synthesis of Strictly Dissipative 11:50-12:10 On a Unitary Model for Two- Systems and the Strictly Suboptimal Time Parameter Stationary Processes, State Space H-infinity Control Problem, Dan Emanuel Popovici Harry. L. Trentelman

Room: 138, Session: FA4 Room: 209, Session: FA6 Chair: Patrick Dewilde Title: Stability and Numerics Chair: Naomi Leonard Title: Coordinated Control of Vehicle 10:30-10:50 Parameter Dependent Extremal Networks Norms for Linear Parameter Varying Systems, Fabian Wirth 10:30-10:50 Stability of Systems of Self- 10:50-11:10 On the Sensitivity of Algebraic Driven Particles Undergoing Phase Riccati Equations, Ji-guang Sun Transitions, A. Stephen Morse 11:10-11:30 A Numerically Reliable Method 10:50-11:10 Stability Properties of Intercon- for a Neglected but Unsolved Problem: nected Vehicles, Vijay Kumar, Herbert Tan- State Feedback Decoupling with Stabil- ner, George Pappas ity for (A, B, C, D) Quadruples, Delin 11:10-11:30 Formations with a Mission: Sta- Chu ble Coordination of Vehicle Group Ma- 11:30-11:50 Large Stability Property of Solu- neuvers, Naomi Leonard, Petter Ogren, Ed- tions of Large-Scale Discrete-Time Sys- ward Fiorelli tems, Tanya Lukyanova, Anatoliy Martynyuk 11:30-11:50 Coordinated Control Strategies 11:50-12:10 Pole Placement Under Output for Networked Vehicles: An Applica- Feedback: A Simplification of the Prob- tion to Autonomous Underwater Vehi- lem, Michael Schilmoeller, Joyce O’Halloran cles, Joao Sousa, Fernando Pereira 12:10-12:30 To the Problem of Construction 11:50-12:10 Group Shape Feedback Control, of Liapunov Functions for Continuous Raffaello D’Andrea Large Scale Systems, Vitaliy Slyn’ko, Ana- 12:10-12:30 Hamiltonian Structures for In- toliy Martynyuk teracting Satellites, P.S. Krishnaprasad MTNS – 2002 33

Room: 210, Session: FA7 Room: 126, Session: FM2

Chair: Mrdjan Jankovic Chair: Belinda King, Kirsten Morris Title: Nonlinear Control and Applica- Title: Distributed Parameter Systems: tions Applications and Computation, Part I 10:30-11:00 Application of Nonlinear 14:00-14:30 Performance Enhancement of Lyapunov-based Controllers and Ob- Controlled Diffusion Processes by Mov- servers to Gasoline Direct Injection En- ing Actuators, Michael Demetriou, Nikolaos gine Charge and Torque Control, Ilya Kazantzis Kolmanovsky 14:30-15:00 Equilibrium Profiles of Tubular 11:00-11:30 Multivariable Extremum Seeking Reactor Nonlinear Models, M. Laabissi, Feedback: Analysis and Design, Kartik B. M. E. Achhab,Joseph Winkin, D. Dochain Ariyur, Miroslav Krstic 15:00-15:30 Control of Electronic Material, 11:30-12:00 Stabilization of Sets Parametrized Katherine Kime by a Single Variable: Application to 15:30-16:00 Active Sound Field Attenuation Ship Maneuvering, Roger Skjetne, Andrew via Acoustic Arrays, H.T. Banks R. Teel, Petar V. Kokotovic 12:00-12:30 Nonlinear Control and Automo- Room: 129, Session: FM3 tive Engine Applications, Mrdjan Jankovic Chair: William Helton Title: Operator Theoretic Methods 14:00-14:20 A Nehari Theorem for Continuous-Time FIR Systems, Gjerrit Middle: Meinsma, Mirkin, Zhong 14:25-14:45 Optimal Approximation of Lin- Room: 102, Session: FM1 ear Operators: a Singular Value Decom- position Approach, Siep Weiland, Hardy Chair: Martin Haenggi Siahaan, Anton Stoorvogel Title: Mathematical Theory of Networks 14:50-15:10 Geometrical and Spectral Prop- and Circuits erties of the Time-Varying Riccati Dif- 14:00-14:20 On Switched Hamiltonian Sys- ference Equation, Nevio Carpanese tems, Arjan van der Schaft, Maurice Heemels, 15:15-15:35 A Generalization of the Karin Gerritsen Widrow’s Quantization Theorem, Alexan- 14:20-14:40 Parameter Influence on the Zeros dru Isar, Dorina Isar of Network Determinants, Sven Feldmann 15:40-16:00 Functions of System and Their 14:40-15:00 Canonical Realizations of Linear Perturbations, Alexey (Olexiy) Tikhonov Time-Varying Systems, Fred Neerhoff, P. van der Kloet Room: 138, Session: FM4 15:00-15:20 In Search of Sensitivity in Net- Chair: David Nicholls work Optimization, Mike Chen, Charuhas Title: Nonlinear Surface Water Waves: Pandit, Sean Meyn Theory, Computation and Experiment 15:20-15:40 Dynamic Eigenvalues for Scalar Linear Time-Varying Systems, Pieter Van 14:00-14:30 Numerical Simulation of Blow- der Kloet, F.L. Neerhoff up Solutions of the Vector Nonlinear 15:40-16:00 Interconnection Structures in Schr¨odingerEquation, Catherine Sulem Physical Systems: a Mathematical For- 14:30-15:00 Existence Theory for Traveling mulation, Goran Golo, Orest V. Iftime, Arjan Water Waves in Three Dimensions, van der Schaft Walter Craig MTNS – 2002 34

15:00-15:30 Numerical Simulation of Travel- Room: 126, Session: FP2 ing Water Waves, David Nicholls Chair: Belinda King, Ruth Curtain 15:30-16:00 Similarities between the Quasi- Title: Distributed Parameter Systems: Bubble and the Generalized Wave Con- Applications and Computation, Part II tinuity Equation Solutions to the Shal- low Water Equations, John H.Atkinson, 16:30-16:50 POD Based Control of Beam Vi- Joannes Westerink brations: Methodology and Experimen- tal Implementations, Brian Lewis, Gregory P. Hicks 16:50-17:10 A Comparison of Balancing Tech- niques for Reduced Order Controllers for Systems of PDEs, Belinda King, Katie A. E. Camp 17:10-17:30 Modeling and Control Issues As- sociated with Atomic Force Microscopy, Room: 210, Session: FM5 Ralph Smith 17:30-17:50 The Effect on Control Design of a Stabilized Finite Element Approxima- Chair: Lars Gruene, Fabian Wirth tion for Burgers’ Equation, Belinda King, Title: Input-to-State Stability, Part II 17:50-18:10 Functional Gain Computations 14:00-14:30 Input-to-state stability of pulse for a 1D Parabolic Equation Using Non- width modulated control systems, Andrew Uniform Meshes., John Burns, Belinda B. Teel, L. Moreau, D. Nesic King, Lizette Zietsman 14:30-15:00 ISS for Dynamic Inputs, Fabian 18:10-18:30 A Continuous Control Design Wirth Method, Jeff Borggaard 15:00-15:30 A Relaxation Theorem for Dif- ferential Inclusions with Applications to Stability Properties, Yuan Wang, Eduardo 18:00—20:00 Farewell Party Sontag, B. Ingalls Center for Continuing Education 15:30-16:00 Characterization of the Non- Uniform in Time ISS Property and Ap- plications, Iasson Karafyllis, J. Tsinias MTNS – 2002 35 Abstracts of Presented Talks

Monday August 12, 2002 use of stochastic models for the dynamics and mea- surement processes. In this talk we take a fresh look at 8:30-9:30 Room: 101 Plenary Talk problems in this area with a view toward finding com- Bruce Hajek (University of Illinois) putational procedures that will determine the inputs A Basket of System Theoretic Problems in which will optimize specific performance measures. In Communications particular, we explore performance measures related Researchers are currently faced with a rich set of prob- to conditional entropy, and in this way develop a for- lems in the design and analysis of wireless communica- malism for establishing the mathematical limits on tion systems and high-speed communication networks. what can be accomplished with better input design. The purpose of this talk is to survey several significant open problems which system theory could play a role in solving. The problems include: (a) Providing de- lay constraints in a large communication network in a distributed way, using delay calculus based on the max-plus algebra, (b) Finding the capacity of cellu- 9:30-10:30 Room: 102 Invited Talk lar networks (will survey the recent use of the theory of large random matrices and related free probabil- Hans-Andrea Loeliger (ETH) ity theory), (c) Predicting efficiency and fairness, and Factor Graphs, Least Squares and Kalman Fil- designing allocation mechanisms, for the Internet, as tering thousands of autonomous systems interact through Factor graphs are graphical models with origins in self-interested pricing and congestion, and (d) Dis- coding theory and with close relations to Willems-type tributing limited routing information in a large amor- behavioral system theory. The sum(mary)-product al- phous network, such as a peer-to-peer network or ad gorithm, which operates by message passing in the hoc sensor array, to facilitate position location. factor graph, subsumes a great variety of algorithms in coding, signal processing, and artificial intelligence. In this talk, we apply Forney-style factor graphs to lin- ear models with quadratic cost functions and we show 9:30-10:30 Room: 101 Invited Talk how general versions of Kalman filtering and recursive least squares algorithms arise as instances of the sum- Roger Brockett (Harvard University) product algorithm. (An ”isomorphism” of such factor Optimal System Identification for NMR Appli- graphs with electrical networks is presented in a sep- cations arate talk.) In a wide variety of settings, the measurement of nu- clear magnetic resonance (NMR) effects has proven to be a remarkably effective for investigating unknown structures on both large and small scales. Over the years a large body of technique has been developed for improving the sensitivity and resolution of NMR mea- surements and many recent advances in biochemistry 9:30-10:30 Room: 136 Invited Talk and medicine are dependent on the sophisticated sig- nal processing techniques now used routinely. From Sjoerd Verduyn Lunel (University of Leiden) a system theoretic perspective, problems in this area Control and Stabilization of Systems with can be thought of as identification problems involv- Time Delays ing bilinear systems. They are distinguished from lin- For dynamical systems governed by feedback laws, ear system identification problems by the fact that time delays arise naturally in the feedback loop to the quality of the identification is strongly dependent represent effects due to communication, transmission, upon the form of the excitatory input applied to the transportation or tern effects. The introduction of system. Many ingenious techniques, such as the “two time delays in a system of differential equations re- dimensional” Fourier transform procedure have been sults in an infinite dimensional state space. In this developed based on particular types of input patterns. paper we give an overview of the basic theory and dis- Because of the low signal to noise ratios inherent in cuss recent developments concerning the control and NMR, the optimization of such methods requires the stabilization of systems with delays. MTNS – 2002 36

Morning: subsystems is addressed. As a special case we con- sider systems that contain one or more integrators Room: 102, Session: MA1 and which give rise to a number of zero elements in Chair: Shmuel Friedland, Brian Marcus the transfer-matrix of the overall system. Title: Capacity of Multi-Dimensional Codes Part I 11:00-13:00 Minicourse on “Capacity of Multi- 11:50-12:10 Synergetic Control of the Unstable dimensional Codes”, Two-Mass System, Shmuel Friedland (University of Illinois at Chicago) Alexander Kolesnikov (Taganrod State University, A multi-dimensional code (mdc) is a filling of points Russia) with integer coordinates (lattice) in the d dimensional The paper is devoted to a problem of synergetic reg- space with a given alphabet according to some rule. ulators synthesis for nonlinear unstable two-mass sys- Example: put in any point of the lattice either 0 or 1, tem “inverted pendulum”. The different variants of with the condition that any 1 must be surrounded by synergetic regulators synthesis for this system are pre- zeros. In this talk we review the available techniques sented. Synthesized regulators are constructed on ba- to compute the capacity (density) of mdc. sis of nonlinear mathematical models. The synthe- sized regulators allow not only to solve the pendulum stabilization task at the top unstable position but also form the new modes of dynamic behavior such as auto- oscillations of the pendulum and the cart around the Room: 126, Session: MA2 set position Chair: Michael D. Lemmon Title: Control Applications 12:15-12:35 Synergetic Control for Electrome- 11:00-11:20 Cancer Treatment Using Multiple chanical Systems, Chemotherapeutic Agents Subject to Drug Re- Andrey Popov, Anatoly Kolesnikov, Gennady Veselov, sistance, Alexander Kolesnikov (Taganrod State University, John Westman (UCLA), Russia); Bruce Fabijonas (Southern Methodist University), Roger Dougal (University of South Carolina) Daniel Kern (University of Minnesota), The paper is devoted to a problem of synergetic reg- Floyd Hanson (University of Illinois at Chicago) ulators synthesis for nonlinear electromechanical sys- A compartment model for the evolution of cancer sub- tems (EMS). The different variants of synergetic reg- ject to multiple chemotherapeutic agents is presented. ulators synthesis for such EMS as DC electric drives, The formulation accounts for the heterogeneous na- asynchronous electric drives and synchronous electric ture of cancer and drug resistance, leading to a sys- drives are presented. Synthesized vector regulators tem that can be used as a tool for optimizing treat- are constructed on basis of nonlinear mathematical ments. Chemotherapeutic or cytotoxic agents have models and ensure to perform the standard techno- limited effectiveness due to of toxicity and the devel- logical tasks: angle and angular speed stabilization, opment of drug resistance. Toxicity limits the amount torque stabilization etc. Also examples of EMS using of the agent that can be used and acts as an upper in mechanical oscillator mode and the task of EMS’s bound on the dosage. Drug resistance causes the clin- power-saving control are considered. ical value of the cytotoxic agent to become nominal after a small number of treatments. Therefore, sev- eral non-cross resistant cytotoxic agents are used to 12:40-13:00 Modeling of Out-of-Plane Hygroin- extend the number of treatments that can be admin- stability of Multi-Ply Paperboard, istered. Gianantonio Bortolin, Per-Olof Gutman (KTH); Bengt Nilsson This paper describes a semi-physical model for the 11:25-11:45 Selection of Decentralized Control dimensional stability properties (i.e. curl) of the car- Configurations Based on Disturbance Rejec- ton board produced at AssiDomn Frvi, Sweden. The tion for Plants with Real Integrators, main equations are based on classical lamination the- Henning Schmidt, Elling Jacobson ory of composite materials, and each constituent ply (KTH) is considered as a macroscopic homogeneous, elastic The paper considers the effect of interactions on the medium. The model shows a general agreement be- disturbance rejection properties of decentralized con- tween predicted and measured curvatures. trol systems. In particular, the problem of selecting the control structure yielding the best disturbance re- jection under independent tuning of the individual MTNS – 2002 37

Room: 136, Session: MA4 inary axis, it is shown that the three following, seem- Chair: Bill Helton, Andre Ran, Leiba Rodman ingly independent problems, are in fact equivalent. Title: Matrix and Operator Equations I Characterizing the set of real symmetric solutions of the algebraic Lyapunov inclusion associated with A. 11:00-11:30 Noncanonical Almost Periodic Fac- The image of all Nevalinna-Pick interpolations asso- torization and Toeplitz Operators with Almost ciated with the spectrum of A. The structure of the Periodic Symbols, Convex Invertible Cone generated by A. Leiba Rodman, I. M. Spitkovsky, H. J. Woerdeman (College of William and Mary) Noncanonical factorizations of almost periodic matrix valued functions of several real variables are studied. In particular, results are proved on behavior of factor- ization indices under small perturbations, connections between left and right indices, and relations between factorization and Fredholmness properties of the as- sociated Toeplitz operators. Room: 208, Session: MA5

11:30-12:00 Symmetric Nonsquare Factorization Chair: Lorenzo Farina, Maria Elena Valcher of Selfadjoint Rational Matrix Functions and Title: Positive Systems Algebraic Riccati Inequalities, 11:00-11:30 Positive Systems in the State Space A. C. M. Ran Mark A. Petersen Approach: Main Issues and Recent Results, (Vrije Universiteit, Amsterdam) Lorenzo Farina (Universit`adi Roma “La Sapienza”) In this talk we shall present a parametrization of all A positive system is a system in which the state vari- symmetric, possibly nonsquare minimal factorizations ables are always positive in value. In this introductory of a positive semidefinite rational matrix function. It tutorial paper, basic results on positive systems are turns out that a pole-pair of such a nonsquare factor reviewed and recent developments and open problems is the same as a pole pair for a specific square factor. are addressed. The location of the zeros In this talk we present a parametrization of all symmetric, possibly nonsquare minimal factorizations of a positive semidefinite ratio- 11:30-12:00 Positive Systems in the Behavioral nal matrix function. It turns out that a pole-pair of Approach: Main Issues and Recent Results, such a factor is the same as a pole pair for a specific Maria Elena Valcher (University of Padova) square factor. The location of the zeros is determined The aim of this paper is to provide an overview of by a solution to a certain algebraic Riccati inequal- the most significant results about positive systems, ity. We also consider the case where the function has obtained within the behavioral approach, and, specif- constant signature. A connection with Bezoutians is ically the main results about positive autonomous be- discussed. haviors, and to present some new results concerned with the positive realization problem for controllable behaviors. 12:00-12:30 Extremal Problems of Interpolation Theory, L. A. Sakhnovich 12:00-12:30 Feedback Stabilisation with Positive We consider interpolation problems the solutions of Control of Dissipative Compartmental Sys- which satisfy in addition an extremal condition.We tems, show that under some assumptions the extremal in- Georges Bastin, A. Provost terpolation problem has a unique solution.We com- (Universite’ Catholique de Louvain) pare our approach with optimal and superoptimal ap- In many process control applications, the system un- proaches. The results of our talk were obtained jointly der consideration is compartmental and positive. For with J.W.Helton. such systems, the design of state feedback controllers makes sense only if the control function is guaranteed to provide a non-negative value at any time instant. 12:30-13:00 Convex Invertible Cones, Nevalinna- The purpose of this paper is to present a positive Pick Interpolation and the Set of Lyapunov So- control law for the feedback stabilisation of positive lutions, compartmental systems which are dissipative but can Izchak Lewkowicz, Nir Cohen nevertheless be globally unstable. The approach is il- (Ben-Gurion University of the Negev) lustrated with an application to an industrial grinding For a real matrix A whose spectrum avoids the imag- circuit. MTNS – 2002 38

12:30-13:00 Feedback Control for a Chemostat the nonlinear problem can be numerically solved, we with two Organisms, can construct explicitly the control functions and then Patrick De Leenheer, Hal Smith implement them in open-loop to achieve the desired (Arizona State University) transfer. It is shown that a chemostat with two organisms can be made coexistent by means of feedback control of the dilution rate. 12:30-13:00 Quantum Control of Dissipative Sys- tems, Sonia G. Schirmer, A. I. Solomon (The Open University, Milton Keynes) We study the effect of dissipation on one’s ability to Room: 209, Session: MA6 control a quantum system. In particular, we show Chair: Augusto Ferrante, Michele Pavon that dissipation, although often considered undesir- Title: Control of Quantum Mechanical Systems able, opens new possibilities for quantum control by removing the constraint of unitary evolution, which 11:00-11:30 Sufficient Conditions for Controlla- restricts the set of reachable sets and imposes bounds bility of Finite Level Quantum Systems via on the optimization of observables. Structure Theory of Semisimple Lie Algebras, Claudio Altafini (ISAS, Italy) The controllability of the unitary propagator of a fi- nite level quantum system is studied in this paper by analyzing the structure of the semisimple Lie algebra su(N).

11:30-12:00 Geometric Control of Quantum Me- Room: 210, Session: MA7 chanical Systems in a Noisy Environment, Domenico D’Alessandro (Iowa State University) Chair: Peter Bauer We study the control problem for a general (finite di- Title: Adaptive Control mensional) quantum system interacting with a bath. 11:00-11:20 Gap Metric Robustness of Adaptive In the Hamiltonian formulation, the overall (System Controllers, plus Environment) system varies on a Hilbert space Mark French (University of Southampton) which is the tensor product of the system and envi- We consider the construction of adaptive controllers ronment Hilbert spaces. We use sensitivity functions for minimum phase linear systems which achieve non- as a tool to compare control laws and trajectories in zero robustness margins in the sense of the (linear) L2 this formulation. We treat the example of a spin 1/2 gap metric. The gap perturbations may be more con- particle in a spin bath in detail. For this system, us- strained for larger disturbances and for larger para- ing optimal control theory and first order sensitivity metric uncertainty. Working within the framework of as the cost, we explicitly derive the minimum deco- the nonlinear gap metric due to Georgiou and Smith, herence controls. universal adaptive controllers are first given achiev- ing this goal for first order plants, and the results are then generalised to relative degree one, minimum 12:00-12:30 Control of Quantum Systems Using phase plants. Model-based Feedback Strategies, Augusto Ferrante, Michele Pavon (Universit`adi Padova); 11:20-11:40 Adaptive Predictive Control with Giorgio Raccanelli Controllers of Restricted Structure, Coherent control of quantum-mechanical systems is a Michael Grimble, Peter Martin growing field due to recent advances in laser technol- (University of Strathclyde) ogy and to the miniaturization of electronic devices. A novel adaptive predictive optimal controller of low We present in this paper a new approach to this prob- order, involving a multi-step cost index and future set- lem. We consider feedback control strategies for the point knowledge, is considered. A non-linear system, Schroedinger equation of a multilevel quantum sys- represented by a set of multiple linear discrete-time tem that eventually produce the decrease of the dis- state-space models plus one identified model is to be tance between the state (or the propagator) and the controlled. Simulation results applied to a ship roll ex- target. Once the functional form of these controls has ample demonstrate that the controller is more robust been obtained, plugging them into the evolution equa- than a standard self-tuning scheme, but is capable of tion we get a nonlinear initial value problem. When good performance. MTNS – 2002 39

11:40-12:00 Output Adaptive Model Reference z(0) = f(x(0)) it will continue to produce an exact Control of Linear Continuous State-Delay estimate z(t) = f(x(t)) for all future times t. It is Plant, well known that a tracking observer for f exists if Boris Mirkin, Per-Olof Gutman and only if the kernel of f is a conditioned invariant (Technion - Israel Institute of Technology) subspace with respect to the observed system. The This paper develops a new approach for the output set M of tracking observer parameters of fixed size, model reference adaptive control of linear continuous- i.e. tracking observers of fixed order together with the time plants with state delays. The main idea is to functions they are tracking, is shown to be a smooth include into the control law a feedforward component manifold. Furthermore, the set of conditioned invari- which compensates for the delayed states, in addition ant subspaces of fixed codimension together with their to output feedback. The feedforward is formed by friends, i.e. the output injections making the subspace special adaptively adjusted prefilters as a function of invariant, is shown to carry a vector bundle structure the delayed state of the reference model. The output over M. Potential applications are in the area of pole feedback component is designed as for a plant with- placement by dynamic output feedback as well as the out delay, but applied to the time-delay plant. Such a analysis of the convergence behaviour of numerical al- controller structure containing adaptive output error gorithms for observer design. feedback and adaptive prefilters from the delayed ref- erence model makes it possible to solve the problem of adaptive exact asymptotic output tracking under parametric uncertainties. The stability is analyzed us- ing the Lyapunov-Krasovskii functional method. Sim- ulation results show the effectiveness of the proposed scheme. Middle:

12:00-12:20 A Comparison Between Robust Room: 102, Session: MM1 Adaptive Controllers w.r.t a Non–singular Transient Cost, Chair: Shmuel Friedland, Brian Marcus Ahmad Sanei, Mark French Title: Capacity of Multi-Dimensional Codes (University of Southampton) Part II A nonsingular performance comparison between two standard robust adaptive control designs based on the 14:00-15:00 On the Capacity of 2-D Constrained dead-zone and projection modifications is given. A Codes and Consequences for Full-Surface Data worst case transient cost functional penalising the L- Channels, infinity norm of the state, control and control deriva- William Weeks (University of Missouri–Rolla) tive has been chosen as the criterion of comparison. If The capacity of two-dimensional constrained binary a bound on the L-infinity norm of the disturbance is codes is bounded by considering strips in the plane of known, it is shown that the dead-zone controller out- width n and constructing matrix recursions. Curve- performs the projection controller when the a-priori fitting and bounding techniques allowe precise esti- information on the unknown system parameter is suf- mates of the capacity to be made. Finally, by assum- ficiently conservative. For clarity of the presentation, ing that capacity-achieving encoders and decoders ex- the result is restricted to scalar systems and the gen- ist, density improvements for full-surface channels are eralisations are only briefly discussed. calculated, and those calculations verified by numeri- cal simulation.

12:20-12:40 A Manifold Structure on the set of Functional Observers, 15:00-16:00 Counting Independent Sets in the Jochen Trumpf, Uwe Helmke Grid, And Similar Questions, (Universit¨atW¨urzburg) Neil Calkin (Clemson) Assume, a linear function f of the state x of a linear We shall consider various enumeration questions aris- control system in state space form is given. An ob- ing from statistical mechanics: for example, counting server for this function is another linear control sys- the number of independent sets in an n k grid, count- ing non-attacking configurations of Kings× on an n k tem taking the input and the output of the observed × system as its input and generating from it an esti- chessboard, etc. We shall show how to prove rigorous mate for f. The observer is called a tracking observer (albeit somewhat weak) upper and lower bounds on if its state z satisfies the tracking property: when- the entropy of such systems. We will also introduce a ever the observer is started with an exact estimate formulation which leads to many open questions about the eigenstructure of recursively defined matrices. MTNS – 2002 40

Room: 126, Session: MM2 , A. C. Cole (U of Central England in Birmingham) Chair: Wei Lin A potentially improved class of systems involves slid- Title: Nonlinear Systems and Control 1 ing mode control, in which the actuating signal is changed discontinuously, depending on the position of 14:00-14:20 Estimating Generalized Gradients of the system trajectory in state space, forces the state Value Function in Optimal Control Problems trajectory to track a defined hyperplane towards the for Differential-Difference Inclusions, desired steady-state. This affords the opportunity of Leonid Minchenko, Aleksey Volosevich simplifying the complex and uncertain processes that (Belarus State University) occur within SI engines and to develop new methods This paper deals with the optimal control problems for of controlling their performance. the differential- difference inclusions under endpoint constraints. The result obtained employs Clarke’s normal cones and the estimates of generalized gradi- 15:20-15:40 Truncation and Approximation Er- ents of the value function. The method for estimating rors in the Max-Plus Algorithm for H-infinity generalized gradients is developed. These estimates Control, are used for obtaining necessary optimality conditions. William McEneaney (University of California, San Diego) We consider the H-infinity problem for a nonlinear 14:20-14:40 Interconnected Systems of Fliess system. The semi-group is linear over the max-plus Operators, algebra. The value function is an eigenfunction cor- W. Steven Gray, Yaqin Li responding to the max-plus multiplicative identity. (Old Dominion University) One may truncate an expansion to obtain a finite- Given two analytic nonlinear input-output systems dimensional problem. We obtain some error estimates represented as Fliess operators, four system intercon- for the size of the errors introduced by this basis trun- nections are considered in a unified setting: the par- cation, and consider errors introduced by the approx- allel connection, product connection, cascade connec- imation of the elements of the matrix. tion and feedback connection. In each case, the cor- responding generating series is produced, when one exists, and conditions for convergence of the corre- 15:40-16:00 Solution of Second Order Lineariza- sponding Fliess operator are given. In the process, an tion, existing notion of a composition product for formal Rajagopalan Devanathan power series is generalized to the multivariable set- (Nanyang Technological University, Singapore) ting, and its set of known properties is expanded. In For a nonlinear system with a control input, a gen- addition, the notion of a feedback product for formal eralized form of the homological equation can be for- power series is introduced and characterized. mulated to reduce the system to its normal form. In this paper, it has been shown that for a linearly con- trollable system with an appropriate choice of state 14:40-15:00 Controllability Analysis of A Two feedback, the generalized homological equation can be Degree of Freedom Nonlinear Attitude Con- solved to give an explicit solution, of a reduced order, trol System, to the problem of second order linearization. Jinglai Shen, Amit K. Sanyal, N. Harris McClamroch (The University of Michigan) We study a physically simple two degree of freedom at- titude control system that has a single control input. The physical assumptions are described. Equations Room: 129, Session: MM3 of motion are derived and expressed in a nonlinear control form. We demonstrate that the system is in- Chair: Giorgio Picci, Augusto Ferrante herently nonlinear. Conditions for small time local Title: Stochastic Systems 1 controllability of the state and of the configuration 14:00-14:30 Canonical Correlations Between In- are presented. put and Output Processes of Linear Stochastic Models, Katrien De Cock, Bart De Moor 15:00-15:20 Sliding Mode Idle Speed Ignition (Universit´eCatholique de Louvain) Control Strategies for Automotive Engines, We obtain expressions for the principal angles between Manjit Singh Srai the row spaces of input and output data block Hankel (U of Central England in Birmingham), matrices of a linear stochastic model in terms of the H. Sindano (Sagem-Johnson, Birmingham), model parameters. The canonical correlations of the N. E. Gough (University of Wolverhampton) corresponding processes are equal to the limiting val- MTNS – 2002 41 ues of the cosines of the principal angles. From these “flipped” from z = to their reciprocals at z = 0. parametric expressions, the relations between the dif- It is well known that∞ for finite zeros the zero-flipping ferent sets of canonical correlations can be easily de- process takes place by multiplication of the underly- duced. ing spectral factor by a suitable rational all-pass ma- trix function. For infinite zeros, zero-flipping is imple- mented by a dual version of the Silverman structure 14:30-15:00 A Regularized Cepstrum and Co- algorithm. Using this interpretation, we derive a new variance Matching Method for ARMA(n,m) algorithm for filtering of non-regular processes, based Design, on a reduced-order Riccati equation. Per Enquist(LADSEB/CNR Padova, Italy) An ARMA(n,m) model is uniquely determined by its n first covariances and m first cepstrum coefficients. However, there does not always exist a model match- ing an estimated set of these parameters. We propose Room: 136, Session: MM4 a method determining an asymptotically stable mini- Chair: I. Gohberg, M.A. Kaashoek mum phase model that match the covariances exactly Title: State Space Methods for Problems in Op- and the cepstrum parameters approximatively. A con- erator Theory vex barrier term is used for regularization and the model is determined by a convex optimization prob- 14:00-14:30 State Space Methods, Reproducing lem. Kernel Spaces and Applications, Harry Dym (The Weizmann Institute) There is a natural connection between finite dimen- 15:00-15:30 On Some Interpolation Problems, sional reproducing kernel Hilbert and Krein spaces Gy¨orgy Michaletzky and state space theory which underlies many prob- (E¨otv¨osLor`andUniversity, Budapest), lems that have been investigated in the last several A. Gombani (LADSEB-CNR Padova, Italy) years. The purpose of this talk is to explain this con- Some aspect of the partial realization problem formu- nection and illustrate it by means of examples drawn lated for Schur-functions considered on the right half largely from problems of interpolation, extension and plane is investigated. We show among others that in factorization, as time permits. The talk will be ex- the linear fractional parameterization of all solutions pository. the condition on the McMillan-degree of the solution is transformed into an interpolatiom condition. This analysis can be considered to be partially complemen- 14:30-15:00 A Beurling–Lax Type Theorem in tary to the results of A. Lindquist, C. Byrnes et al. the Unit Ball, on Carathodory functions. Daniel Alpay, Aad Dijksma, Jim Rovnyak (Ben-Gurion University of the Negev) Schur multipliers on the unit ball are operator-valued 15:30-16:00 Non-regular Processes and Singular functions for which the N-variable Schwarz-Pick ker- Kalman Filtering, nel is nonnegative. In this paper, the coefficient spaces Augusto Ferrante, Stefano Pinzoni, Giorgio Picci are assumed to be Pontryagin spaces having the same (Universit`adi Padova) negative index. The associated reproducing kernel Contrary to the continuous-time case, a discrete-time Hilbert spaces are characterized in terms of general- process y can be represented by minimal linear mod- ized difference-quotient transformations. els, which may either have a non-singular or a singular D matrix. In fact, models with D=0 have been com- monly used in the statistical literature. On the other hand, for models with a singular D matrix the Ric- 15:00-15:30 A Naimark Dilation Perspective on cati difference equation of Kalman filtering involves Positive Real Interpolation, in general the pseudo-inversion of a singular matrix. A. Frazho (Purdue University) This “cheap filtering” problem has been studied for No Abstract several decades in connection with the so-called “in- 15:30-16:00 State Space Method, Explicit Solu- variant directions” of the Riccati equation. For a sin- tions of Scattering Problems, and Nonlinear gular D, a reduction in the order of the Riccati equa- Integrable Equations, tion is in general possible. In this paper, we provide Alexander L. Sakhnovich an explanation of this phenomenon from the classical We review recent results on the canonical differen- point of view of “zero flippin” among minimal spec- tial systems with the pseudo-exponential potentials tral factors. Changing D’s occurs whenever zeros are (and related nonlinear integrable equations) that in- volve applications of state space techniques. Such sys- MTNS – 2002 42 tems are characterized by the requirement that the 14:30-15:00 Numerical Homotopy Algorithms associated Weyl function is a bi-proper positive-real for Satellite Trajectory Control by Pole Place- stable rational matrix function. Explicit formulas for ment, the fundamental solution, the spectral and scattering Jan Verschelde, Yusong Wang functions and bound states are derived. (University of Illinois at Chicago) The aim of this paper is to illustrate the application of numerical homotopy algorithms to control the trajec- tory of a satellite. We design output feedback laws via pole placement. The output feedback laws are com- puted by numerical homotopy algorithms, which are based on enumerative geometry. We combine the use Room: 208, Session: MM5 of MATLAB with PHCpack. Chair: Xiaochang Wang Title: Output Feedback Control of Linear Sys- 15:00-15:30 Numerical Schubert Calculus by the tems Pieri Homotopy Algorithm, 14:00-14:30 Counterexamples to Pole Placement Tien-Yien Li, Xiaoshen Wang, Mengnien Wu by Real Static Output Feedback, (Michigan State University) Alex Eremenko, A. Gabrielov Some control theory and enumerative geometry prob- (Purdue University) lems can be formulated as follows. Given linear sub- We consider linear systems with m inputs, p outputs spaces L1, ..., Ln (of certain vector space, satisfying and McMillan degree n controlled by static output certain conditions), find all subspaces of certain di- feedback with a gain matrix K. The pole placement mension which intersect the given subspaces non triv- map for such system is given by ially. We derived a modified Pieri homotopy algorithm for numerically finding all the solutions. In the pro- cess, all polynomial systems to be solved are square

MatR(m p) Poly (mp), systems. × → R K ψK = det(λI A BKC), 7→ − − 15:30-16:00 On Minimal Order Decentralized Output Feedback Pole Assignment Problems, where A MatR(n n),B MatR(n m) and ∈ × ∈ × Xiaochang Wang (Texas Tech University) C MatR(p n) are real matrices, and Poly (mp) is ∈ × R We give a new sufficient condition for the arbitrary the set of monic polynomials of λ of degree mp with pole assignability of a system by decentralized dy- real coefficients. namic compensators. Using such condition we are We say that for a given triple of integers (m, n, p) able to derive a new bound on the degrees of de- the pole placement map is generically surjective if centralized dynamic compensators so that the generic there is a dense open set of systems (A, B, C) ∈ systems have the arbitrary pole assignability. MatR(n n) MatR(n m) MatR(p n) such that the image× of× the map× (??)× contains an× open dense subset of PolyR(mn). It was known that mp n is necessary and mpn is sufficient for generic surjectivity.≥ It was also known that the real pole placement map is not generically Room: 209, Session: MM6 surjective for (m, p) = (2, 2) and (2, 4). We extend this result: Chair: Yutaka Yamamoto Theorem. For every pair (m, p) of even integers, Title: Optimization and Optimal Control there exists an open set U of systems S = (A, B, C) 14:00-14:20 A Jacobi-like Method for the Indef- with m inputs, p outputs and McMillan degree mp, inite Generalized Hermitian Eigenvalue Prob- such that for every S U the real pole placement lem, ∈ map (1) omits an open set in PolyR(mp). Christian Mehl (Technical University of Berlin) The open set U in this theorem is a neighborhood of Indefinite generalized Hermitian eigenvalue problems a system S0 defined by a rational curve in the Grass- arise in many applications, for example in the linear mannian G(m, m + p) which osculates the normal ra- quadratic optimal control problem. In this talk, we m+p 1 tional curve in the projective space RP − . The propose a Jacobi-like algorithm for the solution of the omitted subset in PolyR(mp) is a neighborhood of a indefinite generalized Hermitian eigenvalue problem set of polynomials whose zeros belong to a circle or- and discuss its convergence properties. Moreover, nu- thogonal to the real line. merical examples will be presented. MTNS – 2002 43

14:20-14:40 Disturbed Discrete Time Linear- 15:40-16:00 Some New Results on Linear Quadratic Open-Loop Nash games, Quadratic Regulator Design for Lossless Sys- Gerhard Jank, Dirk Kremer tems, (RWTH Aachen) Maria Gabriella Xibilia (University of Messina), We examine disturbed linear-quadratic games, where Luigi Fortuna, Giovanni Muscato each player chooses his strategy according to a mod- (University of Catania) ified Nash equilibrium model under open-loop infor- In the paper new results concerning LQR regulator mation structure. We give conditions for the existence design for lossless systems are proved. In particular and uniqueness of such an equilibrium. We also show it is given a necessary and sufficient condition which how these conditions are related to certain Riccati dif- allows to obtain the optimal gains of an LQR problem ference equations and a boundary value problem. for a lossless system without solving the CARE equa- tion, when a particular quadratic index is considered.

14:40-15:00 Linear Matrix Inequalities for Global Optimiztion of Rational Functions and H2 Optimal Model Reduction, Afternoon: Dorina Jibetean (CWI, Amsterdam), Bernard Hanzon Room: 102, Session: MP1 In this paper we study unconstrained global optimiza- Chair: Sandro Zampieri tion of rational functions. We give first few theoretical Title: The Interaction of Control, Information results and study then a relaxation of the initial prob- and Communication lem. The relaxation is solved using LMI techniques. 16:30-16:50 Minimum Data Rates for Stabilising Therefore, in general our procedure will produce a Linear Systems with Unknown Parameters, lower bound of the infimum of the original problem. Girish Nair, Robin J. Evans, Bj¨ornWittenmark The algorithm is then applied to the H2 optimal model (University of Melbourne) reduction problem. A fundamental question in the field of joint commu- nication and control is the following: if a plant and controller communicate at a limited data rate and no 15:00-15:20 Newton’s Method for Optimization restrictions but causality are placed on the coding in Jordan Algebras, and control policy, what is the smallest rate above Sandra Ricardo which stabilisation is possible? In this talk, the case (University of Tras-os-Montes e Alto Douro,Portugal), of discrete-time LTI systems with unknown parame- Uwe Helmke, Shintaro Yoshizawa ters is discussed. Necessary and sufficient data rate (University of W¨urzburg) bounds for stabilisability are presented. We consider a convex optimization problem on lin- early constrained cones in a Euclidean Jordan algebra. The cost function consists of a quadratic cost term 16:50-17:10 A Graphical Model Approach to plus a penalty function. A damped Newton algorithm Distributed Control, is proposed for minimization. Quadratic convergence Sekhar Tatikonda (UC-Berkeley) to the global minimum is shown using an explicit We develop a graphical model framework for visualiz- step-size selection. ing structural properties of distributed systems. We show that this graphical model can be used to deter- mine which agents influence other agents. Further- 15:20-15:40 Non-symmetric Riccati Theory and more it can be used to determine which elements in Linear Quadratic Nash Games., an agent’s information pattern are relevant to its de- Dirk Kremer (RWTH Aachen), cision. These structural observations suggest approx- Radu Stefan (University Polytechnica Bucharest) imation techniques. The existence of a stabilizing solution to the non- symmetric algebraic Riccati equation is shown to be equivalent to the invertibility of a certain Toeplitz op- 17:10-17:30 Quantized Stabilization of Single- erator, whose symbol is the transfer matrix function input Nonlinear Affine Systems, of an exponentially dichotomic system. We also show Jialing Liu, Nicola Elia that this theory has an important application to non- (Iowa State University) cooperative differential games. In this paper we show that, for a single-input nonlin- ear affine continuous-time system, a (robustly) stabi- lizing quantizer can be constructed based on a (ro- bust) control Lyapunov function. We characterize MTNS – 2002 44 the coarsest quantizer under certain conditions. The 18:10-18:30 Systems of Dynamics and their Co- quantized control scheme provides understanding to homological Invariants, the problem of how much interaction between the con- Reuben Rabi, Sanjoy Mitter troller and the system dynamics is needed for stabi- (MIT), lization. This talk introduces the idea that an open nonlin- ear dynamical system of the affine kind can, when composed of an interacting assembly of subsystems, 17:30-17:50 Distributed Robust Controller for be viewed “as a space” and demonstrates that this Complex Networks, idea has clear and concrete implications for our un- Wing Shing Wong (Chinese University of Hong Kong) derstanding of the structure of such systems, which Recently, control problems defined over a distributed, has traditionally been a difficult area resisting mean- complex networking environment have received much ingful progress. We produce invariants of an assembly attention. Examples of these problems include conges- of interconnected systems which reflect the manner in tion control over the Internet and the power control which the (sub-)systems are interconnected. Our in- problem of a wireless communication system. Con- variants are cohomology groups which are sensitive to trollers for these systems should possess certain char- dynamics around closed loops of the underlying graph acteristics. For example, they should depend only on a of the system of dynamics. small amount of observation data in order to minimize communication overhead. They have to relatively in- sensitive to the system parameters and maybe even to the system structure, since there are great uncertain- ties about the system model. Moreover, the decision rules are distributed to a large number of independent players, who may or may not be cooperating. In this talk, we report on some robust convergence results of a common class of distributed, low data rate feedback controllers. Room: 126, Session: MP2

Chair: Matthias Kawski 17:50-18:10 Stabilizing Quantized Feedback with Title: Nonlinear Systems and Control 2 Minimal Information Flow: the Scalar Case, 16:30-16:50 Skorokhod-Neumann Boundary Fabio Fagnani (Politecnico di Torino), Conditions in Robust Queueing Service Mod- Sandro Zampieri (Universit´adi Padova) els, In the performance analysis of a quantized stabiliza- Martin Day (Virginia Tech) tion strategy, two parameters play a central role: the We introduce the idea of boundary extremals in con- number of quantization levels used by the feedback trol problems for simple fluid models of queueing sys- and the convergence time of the closed loop system. tems with Skorokohd dynamics, using an L2-gain ro- In this paper we propose a definition of optimality for bust control formulation. Connections with viscosity- a quantized stabilization strategy based on how these sense Neumann boundary conditions are mentioned. parameters grow with the contraction, which is the ra- A simple example is presented in which these ex- tio between the measure of the starting region and the tremals play a fundamental role in the construction target region of the state space. Then, we analyze the of the optimal strategy. performance and prove the optimality of three differ- ent stabilizing quantized feedbacks strategy for scalar linear systems. A state feedback with finitely many 16:50-17:10 Optimization Methods for Target quantization levels yields only the so called practical Problems of Control, stabilization, namely the convergence of any initial Alexander B. Kurzhanski, Pravin Varaiya state belonging to a bigger bounded region into an- (University of California at Berkeley) other smaller target region of the state space. The The present paper indicates an array of reachabil- ratio between the measure of the starting region and the target region is called contraction of the closed ity problems relevant for nonstandard target prob- loop system. lems of control. The problems are solved through dy- namic programming techniques for systems with non- integral costs. This leads to new types of general- ized Hamilton-Jacobi-Bellman type equations in the general case and in the linear case allows treatment through duality methods of convex analysis and min- max theory. MTNS – 2002 45

17:10-17:30 On Optimal Quadratic Lyapunov ear switched systems, including those appearing in the Functions for Polynomial Systems, quantum computations context . A number of new re- Graziano Chesi (University of Siena), sults and connections is presented, most of them with Alberto Tesi (University of Florence), proofs . Antonio Vicino (University of Siena) The problem of estimating the domain of attraction of equilibria of polynomial systems is considered. Specif- ically, the computation of the quadratic Lyapunov function which maximizes the volume of the estimate is addressed. In order to solve this double non-convex optimization problem, a semi-convex approach based Room: 129, Session: MP3 on LMIs is proposed. Moreover, for the case of odd polynomial systems, a relaxed criterion for obtaining Chair: Amarjit Budhiraja an effective starting candidate is presented. Title: Stochastic Control and its Applications 16:30-17:00 Nonlinear Filtering in Correlated Noise: a Wiener Chaos Approach, 17:30-17:50 The Maximum Principle for an Op- Sergey Lototsky (USC) timal Solution to a Differential Inclusion with The solution to the Zakai filterling equation for the State Constraints, time homogeneous diffusion filtering problem with Aurelian Cernea (University of Bucharest) correlated noise is approximated using the Galerkin We consider an optimal control problem given by approximation followed by the Wiener Chaos decom- a differential inclusion, whose trajectories are con- position. The approximation error is estimated and a strained to the closure of an open set. Using recent corresponding numerical algorithm is described. The results on existence and relaxation of solutions of the algorithm has a very simple real-time part and can be constrained differential inclusion necessary optimality used with both continuous and discrete time observa- conditions are obtained in the form of a maximum tions. principle.

17:00-17:30 Stationary Solutions and Forward 17:50-18:10 Synergetic Synthesis of Nonlinear Equations for Controlled and Singular Martin- Interconnected Control for Turbogenerators, gale Problems, Anatoly Kolesnikov, Andrew Kuzmenko Richard H. Stockbridge (Taganrog State University, Russia) (University of Wisconsin Milwaukee) The analytical synthesis problem of coordinating reg- Stationary distributions of Markov processes can typ- ulator for power system, consisting of two turbogener- ically be characterized as probability measures µ that ators, working at big power network, is considered in annihilate the generator A of the process; that is, for this article. We used nonlinear, non-conservative and each such µ, there exists a stationary solution of the interconnected model of such power system. There martingale problem for A with marginal distribution are four control channels in this case. To solve this µ. This result is extended to martingale problems that hard task we used main approach of synergetic con- include absolutely continuous and singular ( with re- trol theory. This approach was developed by pro- spect to time) components and controls. Analogous fessor A.A. Kolesnikov. Briefly synthesis procedure results for the forward equation follow as a corollary. is described. The synthesized regulator provides sta- bilization of the generators excitation currents (syn- chronous EMFs), coordinated turbines rotation fre- 17:30-18:00 An Investment Model with Liquid- quency, asymptotic stability of the closed-loop sys- ity Risk, tem in the whole and compensates the external low- Hui Wang frequency harmonic disturbances that act to genera- We develop a continuous-time investment model to tors. Example of synthesis is considered. incorporate the presence of Liquidity risk. It is math- ematically equivalent to an optimal stopping problem with multiple entries and forced exits. We obtain the 18:10-18:30 Stabilities and Controllabilities of closed-form solution, under the assumptions of geo- Switched Systems (with Applications to the metric Brownian motion output price and Poisson liq- Quantum Systems), uidity shocks. It is shown that the liquidity risk is a Leonid Gurvits (Los Alamos National Laboratory) true sense risk if and only if the entry cost is large We study various stabilities and controllabilities of lin- enough compared to the running cost. MTNS – 2002 46

18:00-18:30 Mean-Variance Portfolio Selection problem is to describe these. That is the topic of under Markov Regime: Discrete-time Models the lecture. The lecture is based on joint papers and Continuos-time Limits, with Heinz Langer (University of Technology, Vienna) George Yin (Wayne State University), and Yuri Shondin (Pedagogical Institute, Nizhny Nov- X. Y. Zhou (Chinese University of Hong Kong) gorod). In this paper, we propose a discrete-time model for mean-variance portfolio selection. One of the distinct features is that the system under consideration is a 17:00-17:30 Applications of Spaces with Indefi- Markov modulated system. We show that under suit- nite Metrics, able conditions and scaling, the process of interest Babak Hassibi (California Institute of Technology) goes to a switching diffusion limit. Related issues on No Abstract optimal strategies and efficient frontier will also be mentioned. 17:30-18:00 Sturm-Liouville Inverse Spectral Problems with Boundary Conditions Depend- ing on the Spectral Parameter, Room: 136, Session: MP4 Cornelis van der Mee, Vjacheslav Pivovarchik (University of Cagliari, Italy) Chair: Harry Dym, Heinz Langer The potential in a Schro”dinger equation on a finite Title: Spaces with Indefinite Metrics and In- interval with energy dependent boundary conditions verse is derived from its eigenvalues using the theory of 16 :30-17:00 Regular and Singular Point-like Hermite-Biehler class entire functions. Perturbations of some Differential Operators in Pontryagin Spaces, Aad Dijksma (University of Groningen) 18:00-18:30 Variational Principles for Block Op- Yuri Shondin (Nizhny Novgorod) erator Matrices, The minimal realization of the Bessel differential ex- Christiane Tretter (Universitaet Bremen), pression Heinz Langer, Matthias Langer (University of Technology, Vienna) 2 2 d α 1/4 In this talk variational principles for block operator + − , 0 = α ( 1, 1), −dx2 x2 6 ∈ − matrices are presented which are based on function- als associated with the quadratic numerical range and in the Hilbert space L2((0, ); dx) is symmetric and ∞ which allow to characterize, e.g., eigenvalues in gaps has defect numbers (1, 1). The Q-function for this of the essential spectrum. symmetric operator and one of its canonical selfad- joint extensions is

π( z)α Q(z) = − . −2 sin (πα)

The minimal realization of the Laguerre expression Room: 209, Session: MP5 d2 d x (1 + α x) ), 0 = α ( 1, 1), Chair: Paul Fuhrmann − dx2 − − dx 6 ∈ − Title: Algebraic Systems Theory 2 α x 16:30-16:50 Further Results on Interconnection in the Hilbert space L ((0, ); x e− dx) is symmet- ric and has defect numbers∞ (1, 1). The Q-function for and Elimination for Delay-Differential Sys- this symmetric operator and one of its canonical self- tems, adjoint extensions is Heide Gluesing-Luerssen (University of Oldenburg, Germany) π Γ( z) We will consider a problem concerning the intercon- Q(z) = − . −sin (πα) Γ( z α) nection of systems from the behavioral point of view. − − The systems under consideration are linear, smooth, Both functions are classical Nevanlinna functions for time-invariant, and have commensurate delays. We 0 = α ( 1, 1), but for all other real noninteger val- will give a characterization for the achievability of ues6 of ∈α −they are generalized Nevanlinna functions, a given subsystem via regular interconnection from and hence are Q-functions for some symmetric oper- the overall system. This also leads to a criterion for ator in a Pontryagin space with defect numbers (1, 1) achievability via regular interconnection followed by and one of its canonical selfadjoint extensions. The an elimination of the (then) latent variables. MTNS – 2002 47

16:50-17:10 Reduction of Affine Systems on Fliess models come as natural generalizations of clas- Polytopes, sical polynomial models. As known, they are defined Jan H. van Schuppen, Luc C.G.J.M. Habets in terms of finitely generated modules and provide an (Eindhoven University of Technology) adequate description of arbitrary linear systems (not In an affine system on a polytope, state trajectories necessarily observable or controllable). In this note we are terminated when they reach a facet of the poly- extend Fliess’ approach to linear systems defined over tope and attempt to exit. The realization problem for an arbitrary commutative ring. We believe that the these systems is based on their behaviors, i.e. the set exposition will be of interest even for the field case. of input-output trajectories on time-intervals of either finite or infinite length. The state set can be reduced if and only if a subspace of the classical unobservable subspace, characterized by the normal vectors of the Room: 210, Session: MP6 exit facets, is nontrivial. Chair: Daniel Liberzon Title: Hybrid Systems and Control 17:10-17:30 State Feedback Stabilization with 16:30-17:00 Nonlinear and Hybrid Control via Guaranteed Transient Bounds, RRTs, Fabian Wirth, Diederich Hinrichsen, Elmar Plischke Michael Branicky, Michael M. Curtiss (Universitaet Bremen) (Case Western Reserve University) We analyze under which conditions linear time- In this paper, we review rapidly-exploring random invariant state space systems can be stabilized by trees (RRTs) for motion planning, experiment with static linear state feedback such that prescribed tran- them on standard control problems, and extend them sient bounds hold pointwise in time. Necessary and to the case of hybrid systems. sufficient conditions for stabilizability in this sense are derived for the quadratic stability definitions. While the general case is open, we obtain necessary and suf- 17:00-17:30 Reachability Analysis of Hybrid Sys- ficient criteria for generating a strict contraction semi- tems with Linear Dynamics, group w.r.t. the spectral norm. Mireille Broucke (University of Toronto) We present results on the symbolic reachability anal- ysis for hybrid systems with linear autonomous dy- 17:30-17:50 Reduction of the Number of Param- namics in each location. eters for all Stabilizing Controllers, Kazuyoshi Mori (The University of Aizu) This paper presents a parameterization method of sta- 17:30-18:00 Towards the Control of Linear Sys- bilizing controllers that needs smaller number of pa- tems with Minimum Bit-Rate, rameters than previous. The result in this paper will Joao Hespanha, Antonio Ortega, Lavanya Vasudevan not assume the existence of the coprime factorizability (UC Santa Barbara) and not employ the Youla parameterization. We address the problem of determining the min- imum bit-rate needed to stabilize a linear time- invariant process. For the noise free case, we de- 17:50-18:10 Structural Properties of LTI Singu- termine a bit-rate below which stabilization is not lar Systems by Output Feedback, possible and above which asymptotic stabilization Runyi Yu (Eastern Mediterranean University), can be achieved. Inspired by differential pulse code Dianhui Wang (La Trobe University, Australia) modulation techniques, we propose practical encod- Generic results regarding the structural properties ing/decoding schemes that guarantee boundedness of of LTI singular system are presented in this paper. the state for the case of a noisy linear time-invariant These include the structure of the system poles (both process. finite and infinite), assignment of the finite poles, elimination of the impulse mode, and controllability and observability of the closed-loop system. These 18:00-18:30 Control of Hybrid Systems along properties are characterized by some new concepts de- Limit Cycles, fined in this paper. Milos Zefran, Guobiao Song, Francesco Bullo (University of Illinois) In this work we use semidefinite programming and the 18:10-18:30 On Fliess Models over a Commuta- framework of piecewise linear systems to design con- tive Ring, trollers that stabilize a hybrid system to a limit cycle. Vakhtang Lomadze (University of Southampton) Dynamics of a hybrid system along a periodic orbit is not continuous so the usual arguments that rely MTNS – 2002 48 on the smoothness of the system do not hold. This plicated and without apparent structure, and easily problem is motivated primarily by robotic walking de- contain algebraic constraints arising from the inter- vices, but it is relevant also for flight control systems connection of the sub-systems. As such, they may and electrical power networks. not be well-suited for analysis,simulation and control. The aim of this presentation is to show how a particu- lar type of network modeling, namely port-based mod- eling, where the sub-systems are interacting with each other through power exchange represented by pairs of Tuesday August 13, 2002 conjugated variables, immediately leads to generalized Hamiltonian equations of motion. In fact, the inter- 8:00-9:00 Room: 101 Plenary Talk connection structure of the complex system, together with power-preserving elements like transformers and Gilbert Strang (MIT), workless constraints, defines a geometric object; called Filtering and Signal Processing a Dirac structure. The equations of motion are Hamil- We discuss two filters that are frequently used to tonian with respect to this Dirac structure and the smooth data. One is the (nonlinear) median filter, Hamiltonian defined by the total energy of the sys- that chooses the median of the sample values in the tem, together with the energy-dissipation structure. sliding window. This deals effectively with ”outliers” The resulting class of geometrically defined systems that are beyond the correct sample range, and will are called port-Hamiltonian systems. This framework never be chosen as the median. A straightforward im- offers many possibilities for analysis, simulation and plementation of the filter is expensive, particularly in control, which we will briefly address. In particular, two dimensions (for images). The second filter is lin- we will indicate how a given port-Hamiltonian plant ear, and known as ”Savitzky-Golay”. It is frequently system can be controlled by interconnecting it with used in spectroscopy, to locate positions and peaks a controller port-Hamiltonian system, and how this and widths of spectral lines. This filter is based on a extends classical passivity theory. Finally, we show least-squares fit of the samples in the sliding window how the framework can be extended to boundary con- to a polynomial of relatively low degree. The filter co- trolled distributed-parameter systems, by considering efficients are unlike the equiripple filter that is optimal infinite-dimensional Dirac structures based on Stokes’ in the maximum norm, and the ”maxflat” filters that theorem. are central in wavelet constructions. We will discuss the analysis and the implementation of both filters.

9:00-10:00 Room: 101 Invited Talk Arjan van der Schaft (University of Twente), Mathematical Theory of Network Models of Physical Systems 9:00-10:00 Room: 102 Invited Talk Prevailing trend in the modeling and simulation of complex (lumped-parameter) physical systems is modular modeling, where the complex physical sys- Robert J. McEliece (Cal Tech), tem is represented as the network interconnection of Belief Propagation on Partially Ordered Sets ideal elements. This has many advantages in terms The field of error-correcting codes has been revolu- of flexibility, re-usability of model parts, and sup- tionized by the advent of suboptimal iterative de- port for automated modeling. The viewpoint of net- coding methods, for example turbo-decoding and it- work modeling necessarily leads to the consideration erative decoding of low-density parity-check (LDPC) of open dynamical systems, that is, dynamical systems codes. Several years ago it was recognized that all with external variables, which can be interconnected scuh decoding algorithms are special cases of Pearl’s to other open systems. This viewpoint is fundamen- ”belief propagation” algorithm applied to graphs with tal to systems and control theory, where the analysis cycles, although BP isn’t supposed to work when cy- of complex systems is based on the properties of the cles are present! In this talk I will review the state of sub-systems and the way they are interconnected to our understanding of loopy BP, and will, motivated by each other, and the behavior of the complex system the recent work of Yedidia, Freeman, and Weiss, in- is sought to be influenced and regulated by the ad- troduce a promising generalization of BP in which the dition of additional feedback loops and control com- messages are passed not along the edges of a graph, ponents. On the other hand, the equations of motion but rather along the edges of the Hasse diagram of a obtained from direct network modeling are often com- finite partially ordered set. MTNS – 2002 49

9:00-10:00 Room: 129 Invited Talk pacity and lower the bit error probability at the same Jeff Wood (University of Southampton), time. In this paper, we design some specific constella- Modules and Behaviors: Re-examining tions according to two design criteria and their trans- Oberst’s Duality formations. Some of our codes feature good perfor- The module associated to a system, and the duality mance and efficient decoding algorithms. between system modules and system behaviors, will be introduced from a natural point of view. We will examine how injectivity of the signal space is related to the elimination problem, and the cogenerator prop- Room: 126, Session: TUA2 erty to identifiability. We will also introduce ‘abstract Chair: Wijesuriya P. Dayawansa behaviours’, which are independent of any embedding Title: Patterns in Biology in a trajectory space. However, most of the talk will 10:30-11:00 Visual Systems, be of a tutorial nature and will be illustrated by the Bijoy Gosh, A. Polpitiya case of systems defined by linear ordinary/ partial (Washington University St.Louis) constant coefficient differential equations. No Abstract

11:00-11:30 The Dynamics of Avian Kinesis, Morning: Lawrence Schovanec, Alan Barhorst, Sankar Chatter- jee Room: 102, Session: TUA1 (Texas Tech University) Chair: Heide Gluesing-Luerssen Unlike most vertebrates, birds can raise the upper jaw Title: Design and Analysis of Block Codes, with respect to the brain case, a feature known as Part I cranial kinesis. This property makes it possible for 10:30-11:30 Iterative Decoding and Design of birds to create a wide gape and fast movement of Codes on Graphs, the bills. This paper provides a method for assess- Pascal O. Vontobel (ETH) ing how neural controls and physiological variations The concept of factor graphs and the sum-product al- affect the gape, speed of movement, and the magni- gorithm is one possible way to understand the differ- tude of forces generated by the jaw. This is achieved ence between codes suitable for iterative decoding and by incorporating joint torques derived from muscul- ”traditionally good” codes. After stating some desir- tendon dynamics into the equations of motion for the able properties of factor graphs, we will review differ- system modeled as a system of articulating segmental ent approaches given in the literature leading to ran- links, closed loops and distributed elastic components. dom and algebraic constructions of turbo, low-density The method accounts for both the rigid body motions parity-check, and other codes. and the elastic deformations of the articulating com- ponents. This is accomplished by using a nonholo- nomic hybrid parameter projection method in which 11:30-12:00 Codes for Networks, the dynamics are projected onto the constraint-free Ralf Koetter (University of Illinois) manifold of the generalized speed space and thus re- We investigate connections between coding for net- quire no algebraic constraint conditions. Numerical works s suggested by Ahlswede et al. and linear sys- simulations and analysis of the system provide insight tem theory. In particular we focus on structure the- into how the biomechanical are related to functional orems for linear systems on arbitrary graphs. Such issues of the avian skull as well as evolution of the a linear system may be described by a trellis on the system. underlying graph. Employing various duality results from the theory of trellises and generalized state re- alizations, we give an algorithm that for a given net- 11:30-12:00 Spiral Waves in the Heart, working problem achieves a locally optimal solution. Clyde Martin, P. Marcus (Texas Tech University) A major problem that is being considered by several 12:00-12:30 Unitary Constellation Design with groups around the world is to model the electrical and Application to Space-time Coding, muscular actions of the mammalian heart. These ef- Guangyue Han, Joachim Rosenthal forts vary greatly in complexity. At the most com- (University of Notre Dame) plicated level the model is developed at the cellular Unitary modulation scheme in Rayleigh flat fading level and the detailed physiology is taken into account. channel with multiple antennas requires us to design These models are very difficult to simulate, often re- certain space-time codes to increase the channel ca- quiring hours or days of super computer time for a MTNS – 2002 50 few milliseconds to seconds of heart time. However between kernel bundles for determinantal representa- the result is very accurate simulation. At the simplest tions of an algebraic curve), and there is a Laplace level the heart is modeled as a single partial differen- transform (continuous time) or Z-transform (discrete tial equation. These models are easy to simulate and time) along the discriminant curve which implements several minutes of heart action can be realized in at the transformation from time domian to frequency do- most a few minutes of computer time. However, it main. In addition to analogues of the usual problems is very difficult to relate the results to physiological from system theory, we mention two applications be- assumptions. We have worked extensively with these yond the usual scope of system theory: wave-particle models and are able to develop some understanding of duality in quantum mechanics, and a mathematical the relation of the models to the cardiac fibrillation. model for DNA chains. We are now in the process of developing models that are intermediate to these two classes of models.

12:00-12:30 Large Amplitude Travelling Waves in Coupled Oscillater Networks, Wijesura P. Dayawansa, Clyde Martin (Texas Tech University) Room: 136, Session: TUA4 Theory of small amplitude oscillations in coupled os- Chair: Lars Gruene, Fabian Wirth cillator networks is well established. However, when Title: Input-to-State Stability, Part I the value of bifurcation parameters become large, the analysis of such networks become rather complex. A 10:30-11:00 Attractors, Input-to-State-Stability, network consisting of cubic oscillatos is considered and Control Sets, here, and it is shown that it is possible to increase bi- Fritz Colonius (University of Augsburg), furcation parameters so as to maintain the simplicity W. Kliemann (Iowa State University) of resulting oscillation patterns, leading to a mecha- The relations between attractors, ISS and controllabil- nism to generate complex spiking trains. ity properties are discussed. In particular, it is shown, that loss of the attractor property under perturbations is connected with a qualitative change in the controlla- bility properties due to a merger with a variant control set. Room: 129, Session: TUA3 Chair: Victor Vinnikov, Joseph A. Ball 11:00-11:30 Output-Input Stability of Nonlinear Title: Minicourse A: Multidimensional Systems Systems and Input/Output Operators, 10:30-12:30 Overdetermined Multidimensional Daniel Liberzon (University of Illinois) Systems and Applications, The recently proposed definition of output-input sta- Victor Vinnikov (Ben Gurion University of the bility requires the state and the input of the system Negev) to be bounded in terms of the output and derivatives Joseph A. Ball of the output. The present work extends this con- (Virginia Tech) cept to the setting of input/output operators. We We consider 2D ISO linear systems where the evolu- show that output-input stability of a system implies tion of the whole state is specified in two indepen- output-input stability of the associated input/output dent directions. The requirement that the value of operator, and that under suitable reachability and ob- the state at a given point be independent of the path servability assumptions, a converse result holds. from the origin chosen to arrive at the given point leads to nontrivial consistency conditons: the tran- sienct evolutions should commute and the input (and 11:30-12:00 A Parameter-Robust Observer as an output) signal should satisfy a nontrivial compatibil- Application of ISS Techniques, ity PDE/DE. Many of the standard structural prop- Madalena Chaves (Rutgers University) erties and operations for traditional linear systems Systems that model chemical networks are often de- (controllbility, observability, minimality, pairing with fined through a set of parameters (the reaction rate the adjoint system, feedback coupling, equivalence be- constants) whose values may be determined with a tween conservative systems and Lax-Phillips scatter- small margin of error. These parameters will typi- ing theory) carry over for this setting. There is also cally also appear in the construction of observers. A a frequency-domain theory for this class of systems definition of parameter-robustness is proposed and an (the transfer function is a bundle map between flat explicit observer for zero-deficiency chemical networks vector bundles on a Riemann surface, or equivalnetly is presented, which is robust in this sense. MTNS – 2002 51

12:00-12:30 Quantitative Aspects of the Input- 11:30-11:50 On Kalman Models over a Commu- to-state Stability property, tative Ring, Lars Gruene (J.W. Goethe-Universitaet Frankfurt) Vakhtang Lomadze (University of Southampton) In this talk we consider quantitative aspects of the There is a good notion of rational functions with coef- input-to-state stability (ISS) property. Our consider- ficients in a commutative ring. Using this notion, we ations lead to a new variant of ISS, called input-to- easily obtain a neat generalization of Chapter 10 of state dynamical stability (ISDS). The main feature of the classical book by Kalman et al. to linear systems ISDS is that it admits a quantitative Lyapunov func- over an arbitrary commutative ring. The generaliza- tion characterization. We clarify the relation to the tions certainly exist already. However, we believe that original ISS formulation and present several applica- the approach we present is more natural and straight- tions. forward.

11:50-12:10 On Rosenbrock Models over a Com- mutative Ring, Room: 208, Session: TUA5 Vakhtang Lomadze (University of Southampton) Chair: Anders Rantzer Rosenbrock’s notion of system equivalence is general Title: Linear Systems in nature; it is a kind of equivalence which in algebra 10:30-10:50 A New Property of Laguerre Func- is often termed as stable. We have shown recently tions, that Fuhrmann’s notion of system equivalence can be Luigi Fortuna, Riccardo Caponetto, Mattia Frasca viewed as a homotopy equivalence, and as such is also (University of Catania, Italy) general in nature. This note deals with a generaliza- Laguerre filters constitute an orthonormal basis for tion of the theory of system equivalences from the field the Hilbert space, for this they are used in system case to the commutative ring case. identification and reduced-order modelling. In this paper a new property of Laguerre filters is introduced: it is shown that the system having as transfer function 12:10-12:30 Inclusion of Frequency Domain Be- the sum of the first n+1 functions has all the singular haviors, values equals each other. On the basis of this property Stephen Prajna (Cal Tech), a generalization of Laguerre filters is proposed. Pablo A. Parrilo (ETH) This paper addresses inclusion of behaviors and its verification. It is shown that verifying inclusion of fre- 10:50-11:10 Communication-Limited Stabilis- quency domain behaviors defined by polynomial fre- ability of Jump Markov Linear Systems, quency domain equalities and inequalities amounts to Girish Nair, Subhrakanti Dey, Robin Evans proving emptiness of some basic semialgebraic sets. A (University of Melbourne) semidefinite programming relaxation method for solv- This paper investigates the control of fully observed, ing this problem is outlined. Some applications are scalar jump Markov linear systems in which feedback given to illustrate the use of the concepts. is transmitted at finite data rates over noiseless digi- tal channels. In particular, the objective is to find the infimum data rate, over all causal coding and control laws, at which asymptotic closed-loop stability in m- th absolute output moment is achievable. Room: 209, Session: TUA6

11:10-11:30 Equivalence of Finite Pole Chair: Viswanath Ramakrishna Assignability of LTI Singular Systems by Out- Title: Quantum Engineering I put Feedback, 10:30-11:10 A Numerical Approach to the De- Runyi Yu (Eastern Mediterranean University), sign of Strongly Modulating Pulses to Imple- Dianhui Wang (La Trobe University, Australia) ment Precise Effective Hamiltonians for Quan- This paper shows that the assignability of finite poles tum Information Processing, of a strongly controllable and observable singular sys- Timothy Havel, Nicolas Boulant, David G. Cory, tem (E, A, B, C) is equivalent to the pole assignabil- Evan M. Fortunato, Marco A. Pravia, Grum Tekle- ity of a non-singular system (As, Bs, Cs) of order mariam rank(E). Consequently, all the existing results on (MIT) pole assignment of non-singular systems can be ex- Unitary ”logic gates” to are used to process quantum tended to singular systems as far as the finite poles information: the state of a 2-D quantum system, or are concerned. ”qubit”. In many approaches to QIP, the gates are MTNS – 2002 52 implemented by modulating the system’s internal dy- 11:00-11:30 Robust Least-Squares Filtering with namics with piecewise constant external control fields, a Relative Entropy Constraint, or ”pulses”. This talk describes a numerical approach Bernard Levy (University of California at Davis), to designing pulse sequences which yield unitary gates Ramine Nikoukhah (INRIA) with high precision. The results of NMR experiments We consider a robust Wiener filtering problem for demonstrating these pulses will also be presented. wide-sense stationary processes in the presence of modelling errors. It requires solving a minimax prob- lem consisting of finding the best filter for the least- 11:10-11:50 System Theoretic Aspects of NMR favorable statistical model within a neighborhood of Spectroscopy, the nominal model, which is specified by the relative Raimund J. Ober (University of Texas at Dallas) entropy with respect to the nominal model. The stan- No Abstract dard noncausal Wiener filter is optimal, and in the causal case a characterization is provided for the best filter and corresponding least-favorable model. 11:50-12:10 Local and Global Control of Popula- tion Transfer in Quantum Systems, Vladimir S. Malinovsky (University of Michigan), 11:30-12:00 Bounding the Solution Set of Uncer- Stimulated Raman Adiabatic Passage (STIRAP) has tain Linear Equations: a Convex Relaxation proven to be an efficient and robust technique of pop- Approach, ulation transfer in three-level and multilevel systems. Giuseppe Calafiore (Politecnico di Torino), On the other hand, the studies of optimal pulse shap- Laurent El Ghaoui (UC Berkely) ing to control atomic or molecular dynamics have In this paper, we discuss semidefinite relaxation tech- reached a high degree of sophistication, specially since niques for computing minimal size ellipsoids that the advent of the use of Optimal Control Theory bound the solution set of a system of uncertain linear (OCT). In this contribution we show the connection equations (ULE). The proposed technique is based on between the counterintuitive pulse sequence in STI- the combination of a quadratic embedding of the un- RAP and the population locking condition of inter- certainty, and the S-procedure. The resulting bound- mediate levels. The relationship between several ver- ing condition is expressed as a Linear Matrix Inequal- sions of OCT called local optimization methods and ity (LMI) constraint on the ellipsoid parameters and STIRAP is also discussed. the additional scaling variables. This formulation leads to a convex optimization problem that can be efficiently solved by means of interior point barrier 12:10-12:30 Hartree-Fock Models in Electronic methods. Structure Computations, Gabriel Turinici (INRIA Rocquencourt/ CERMICS-ENPC, France) 12:00-12:30 The Design of Auxiliary Signals for This talk will present, in a pedagogical manner, an Robust Active Failure Detection in Uncertain algorithmic overview of the resolution of the Hartree- Systems, Fock problem. Although initially set as a minimi- Stephen Campbell (North Carolina State University), sation problem, most traditional implementation use Ramine Nikoukhah (INRIA) the Euler-Lagrange critical point equations that give By modeling the normal and the failed behaviors of rise to a nonlinear eigenvalue problem. It will be seen a process by two or more linear uncertain systems, next that the modern approaches try to come back to failure detectability in linear systems can be seen as a the minimisation framework in a search for efficiency. linear multi-model identification problem. This paper describes an active approach for multi-model identifi- cation using minimal auxiliary signals. Both additive and model uncertainty are included in this approach.

Room: 210, Session: TUA7 Chair: Stephen Campbell, Ramine Nikoukhah Title: Robust Estimation, Identification, and Detection 10:30-11:00 A Survey of Input-Output Methods in Robust Estimation, Babak Hassibi (Cal Tech) No Abstract MTNS – 2002 53

Middle: lations from lattices associated to rings of algebraic integers. The procedure provides a natural way to Room: 102, Session: TUM1 label the constellation points by elements of GF(q). Chair: Daniel Costello The labeling is proven to be linear which allows, at Title: Design and Analysis of Block Codes, the receiver, a fast way to map constellation points Part II into field elements. The constellations performance is determined by their minimum Euclidean distance and 14:00-14:30 On a Few Classes of Optimal and average energy, for rates from 1 up to 3. Near-optimal Polynomial Codes, Nuh Aydin (Ohio State University) We generalize a recent idea of constructing codes over a finite field F by evaluating certain collection of q Room: 126, Session: TUM2 polynomials over Fq at elements of an extension field. We show that many codes with the best-known pa- Chair: Raimund Ober rameters can be obtained with this construction. In Title: Immunology 1: Introduction and Mi- particular, a new linear code, a [40, 23, 10]-code, over croscopy F5 is discovered. Moreover, several families of optimal 14:00-14:40 Introduction to Workshop and and near-optimal codes are obtained by this method. Overview, We call a code near-optimal if its minimum distance Raimund Ober (University of Texas at Dallas) is within 1 unit of the known upper bound. A short introduction to minisymposium will be pre- sented by presented an overview of the topics that will be considered in the mini-symposium. 14:30-15:00 Building Low-Density Parity-Check Codes with Affine Permutation Matrices, Michael O’Sullivan, Marcus Greferath, Roxana 14:40-15:20 T Cell Receptor MHC Interactions: Smarandache An Overview, (San Diego State University) E. Sally Ward (University of Texas at Dallas) We will expand on a technique due to Hui et al for con- A tutorial exposition is given of the biology and in structing low density parity check codes using blocks particular immunology that is central for the under- of matrices, each constructed as a sum of permutation standing of the topics that will be presented in the matrices. The permutation matrices that we use are mini-symposium by other speakers. This presenta- defined by an affine congruence equation. tion should be of particular interest to attendees of the conference who are not familiar with these topics.

15:00-15:30 On Plotkin and Elias Bounds for Codes over Frobenius Rings under the Homo- 15:20-16:00 Image Formation and Deconvolution geneous Weight, for 3 Dimensional Microscopy of Cell Samples, Marcus Greferath (San Diego State Univ.) Jose Angel Conchello (Washington University Homogeneous weights were introduced by W. Heise St.Louis) and I. Constantinescu in 1995. They appear as a nat- No Abstract ural generalization of the Hamming weight on finite fields and the Lee weight on Z4 and have proven to be important in further significant papers. This arti- cle developes a Plotkin and an Elias bound for codes Room: 129, Session: TUM3 on finite Frobenius rings that are equipped with this weight. Surprisingly these bounds have the same form Chair: Krzystof Galkowski, Eric Rogers, Victor Vin- as in the case of the Hamming weight on a finite field. nikov Their proof however requires different techniques than Title: Multidimensional Systems 1 the classical proof: we make use of their representa- 14:00-15:00 2D Linear Control Systems - From tion by generating characters on the finite Frobenius Theory to Experiment to Theory, ring in question. Eric Rogers, Tarek Al-Towlem, James Radcliffe, Paul Lewin (University of Southampton), 15:30-16:00 Four and Six-Dimensional Signal Krzysztof Galkowski (University of Zielona Gora), Constellations from Algebraic Lattices, David Owens (University of Southampton) Carmelo Interlando (University of Notre Dame), The subject area of this paper is the application of sys- Michele Elia (Politecnico di Torino) tems theory developed for linear repetitive processes, We describe a procedure to construct signal constel- a distinct class of 2D linear systems, to linear itera- MTNS – 2002 54 tive learning control schemes. A unique feature is the tion problem for generalized Schur functions can be inclusion of experimental results obtained from the parametrized via a linear fractional transformation application of control laws designed using this theory A(z)E(z) + B(z) to an experimental rig in the form of a chain conveyor S(z) = , system. C(z)E(z) + B(z)

where A, B, C and D are polynomials depending ex- 15:00-15:30 Stability Analysis of 2D Dynamics plicitly on the data of the problem and E is the param- in Roessers Model, eter varying over the class of classical Schur functions Tatsushi Ooba, Yasuyuki Funahashi (i.e., analytic and contractive on the unit disk) such (Nagoya Institute of Technology) that CE + D does not vanish at interpolating points. State-space stability of linear shift-invariant discrete All other parameters E are called excluded and the two-dimensional (2-D) dynamics is considered. An ap- corresponding functions S may have a pole (simple or proach to the making of Lyapunov functions for 2-D multiple) in one or more of the interpolation points or dynamics is presented. It produces a quadratic form not satisfy one or more interpolation conditions. We involving finite cross-terms among local states. The present a classification of excluded parameters E and use of the quadratic form expands the scope of the related functions S in terms of the Pick matrix of the stability analysis based upon the notion of parallel interpolation problem. stability expands.

14:40-15:00 Abstract Interpolation in Scattering 15:30-16:00 Algebraic Algorithm for 2D Stabil- Setting, ity Test Based on a Lyapunov Equation, Alexander Kheifets Minoru Yamada (The College of William and Mary) (Gifu National College of Technology), Abstract Interpolation scheme is adapted to the case Li Xu (Akita Prefectural University), of nonanalytic (harmonic) interpolation. Unitary col- Osami Saito (Chiba University) ligation yields a scattering system, orthogonality as- Agathoklis proposed the 2D stability conditions based sumptions are given up (i.e. analyticity is not neces- on the Lyapunov approach. In this paper, some im- sarily holds). Wandering subspaces are replaced by provements have been proposed for the algorithm of an arbitrary scale (i.e. incoming and outgoing sub- Agathoklis so that 2D stability test can be realized by spaces are not distinguished). The modified version totally algebraic operations. covers, in particular, Nehari and Commutant Lifting problems directly.

15:00-15:20 A Convex Optimization Approach to Room: 208, Session: TUM4 Generalized Moment Problems, Chair: Joseph A. Ball, Hugo Woerdeman Anders Lindquist (KTH), Title: Recent Developments on Interpolation C. I. Byrnes (Washington University, St. Louis) and Completion Problems We present a universal solution to the generalized mo- 14:00-14:20 Feedback Control for Multidimen- ment problem, with a nonclassical complexity con- sional Systems and Interpolation Problems for straint, obtained by minimizing a strictly convex non- Multivariable Functions, linear functional.This optimization problem is derived Joseph A. Ball, Tanit Malakorn in two different ways, first geometrically by path inte- (Virginia Tech) gration of a one-form, and second via duality theory We examine the connections between and the state of mathematical programming. The solution is then of art on feedback stabilization and H-infinity con- applied to interpolation problems and to estimation trol, model matching problems, and multivariable of probability distributions.““ Nevanlinna-Pick interpolation problems for the case of multidimensional or nD linear systems. 15:20-15:40 Extremal Properties of Outer Fac- tors, 14:20-14:40 On the Caratheodory-Fejer Interpo- Scott McCullough (University of Florida) lation Problem for Generalized Schur Func- Each of the coefficients of the outer and -outer factor tions, of a polynomial p(t) which is non-negative∗ on for real Vladimir Bolotnikov t are larger, in absolute value, of the corresponding (The College of William and Mary) coefficient of any other factor even for matrix-valued The solutions of the Carath´eodory–Fej´erinterpola- factors. MTNS – 2002 55

15:40-16:00 On the Realization of Inverse Stielt- link while simultaneously suppressing structural vi- jes Functions, brations. The results are illustrated by a numerical E. R. Tsekanovskii (Niagara University), example. Sergey Belyi (Troy State University), Seppo Hassi (University of Vaasa, Finland), Henk de Snoo (University of Groningen) 15:00-15:30 Furtivity and Masking Problems in A survey of recent results in realization theory of Acoustics, Herglotz-Nevanlinna matrix-valued functions, inter- Francesco Zirilli (University “ La Sapienza” - Rome) polation problems and explicit system solutions ob- We consider the time dependent acoustic scattering tained by the author and D.Alpay, S.Belyi, S.Hassi, phenomena associated to an incoming acoustic field H.de Snoo is presented.We consider a new type of so- that hits a bounded simply connected obstacle with lutions of Nevanlinna-Pick interpolation problems, so Lipschitz boundary and known acoustic boundary called, explicit system solutions and find conditions on impedance. In this context we study furtivity and interpolation data of their existence, uniqueness and masking problems. These problems are formulated as restoration. optimal control problems for the wave equation and solved using the Pontryagin maximum principle via an appropriate numerical method.

Room: 209, Session: TUM5 15:30-16:00 Approximation of Optimal Controls Chair: Damir Arov for Semi-Linear Parabolic PDE by Solving Title: Control of Distributed Parameter Sys- Hamilton-Jacobi-Bellman Equations, tems Sophie Gombao (University Paul Sabatier) This paper presents a numerical approximation of op- 14:00-14:30 Optimal Control and Riccati Equa- timal controls by discretization of a Hamilton-Jacobi- tions for a Degenerate Parabolic System, Bellman (HJB) equation, coming from a family of op- Jean-Marie Buchot (Onera-Toulouse), timal control problems of parabolic PDE. The method Jean-Pierre Raymond (Universit´ePaul Sabatier) is based on model reduction, using POD (Proper Or- In this paper, we consider a stabilization problem for thogonal Decomposition), and on the approximation a fluid flow. For a perturbation of the velocity of an of the HJB equation of the reduced problem by finite incoming flow on a flat plate, the laminar-to-turbulent difference scheme. transition location varies. We want to stabilize it by a suction velocity trough the plate. We look for a suction velocity in a feedback form, determined by solving a LQR problem with an infinite time horizon. Room: 210, Session: TUM6 Chair: Anders Lindquist 14:30-15:00 Nonlinear Predictive Control of Title: Filtering and Identification Flexible Manipulator Systems, 14:00-14:20 System Identification of Nonlinear Alaa Mohamedy, Andrzej Ordys, Michael Grimble Dynamic Systems with Multiple Inputs and (University of Strathclyde) Single Output Using Discrete-Time Volterra The paper discusses predictive control algorithms in Type Equations, the context of applications to robotics and manufac- Thomas Treichl, Stefan Hofmann, Dierk Schr¨oder turing systems. Lagrangian mechanics and the as- (TU M¨unchen) sumed mode methods have been used to drive a pro- A theoretical framework about the identification of posed dynamic model of a single-link flexible manipu- nonlinear dynamic systems with multiple inputs and lator having a revolute joint. The model may be used single output is presented. Basic considerations about in general to investigate the motion of the manip- discrete-time Volterra type equations are performed. ulator in the horizontal plane rest-to-rest rotational It is shown how measuring equations can be obtained maneuver. Then, a nonlinear predictive controller from nonlinear and arbitrary coupled linear time- is designed on the basis of a Nonlinear Quadratic invariant transfer functions and nonlinear static func- Gaussian Predictive Optimal Control (NLQGPC) us- tions. With a simulation example the presented iden- ing the receding-horizon control approach. Based on tification method is evaluated. the (NLQGPC), the control law is derived by min- imizing a quadratic cost function that penalizes fu- ture tracking errors and control torques. Simulation 14:25-14:45 Data Driven Local Coordinates, results are presented for a single-link flexible manip- Thomas Ribarits, Manfred Deistler, Bernard Hanzon ulator to achieve a desired angular rotation of the (University of Technology Vienna) MTNS – 2002 56

In this paper we study a rather novel parametrization Afternoon: for state-space systems: data driven local coordinates (DDLC) as introduced in (McKelvey and Helmersson, 1999). We provide some insights into the geometry and topology of the DDLC construction and show a Room: 102, Session: TUP1 number of results for this parametrization which are also important for actual computations using DDLC. Chair: Heide Gluesing-Luerssen Title: Convolutional Codes 14:50-15:10 Using Rank Order Filters to Decom- pose the Electromyogram, 16:30-17:00 Construction and Decoding of Dawnlee Roberson, Cheryl Schrader Strongly MDS Convolutional Codes, (University of Texas at San Antonio) Roxana Smarandache (San Diego State University), This research applies nonlinear filters to model gen- Heide Gluesing-Luerssen (University of Oldenburg), erated muscle and nerve signals and compares the re- Joachim Rosenthal (University of Notre Dame) sults to determine correlation. The authors generate A new class of of rate 1/2 convolutional codes called three data test sets that range from simple (no noise) strongly MDS convolutional codes are introduced and to fairly complex (noisy, varied amplitude) and apply studied. These are codes having optimal column dis- a Rank Order filter family to the rectified generated tances. Properties of these codes are given and a con- electromyographic signals in an attempt to recover the crete construction is provided. This construction has original (nerve) signal. Promising results occur even the ability to correct δ errors in any sliding window in the most challenging case. of length 4δ + 2 whereas the best known MDS block code with parameters [n, n/2], n = 4δ + 2, can correct δ errors in any slotted window of length 4δ + 2. A de- 15:15-15:35 Conditioning Analysis of a Continu- coding algorithm for these codes is given in the end ous Time Subspace-Based Model Identification of the paper. Algorithm, Juan Carlos Martinez-Garcia (Instituto Mexicano del Petroleo), G.H. Sal`azar-Silva, 17:00-17:30 On Observers and Behaviors, R. Garrido Paul A. Fuhrmann (Ben-Gurion University) (CINVESTAV-IPN) In this talk, we explore the connection between the We present in this paper a study concerning the con- classic, input output based theory of observers for lin- ditioning analysis of a continuous-time deterministic ear functions of the state and the theory of behavior subspace-based model identification algorithm. We based observers as developed in the paper Valcher and show that the conditioning number of the associated Willems [1999]. extended observability matrices depends on an expo- nential way from both: the estimated order of the system and the dimension of the system output vec- 17:30-18:00 On the Convergence of Nonsystem- tor. atic Turbo Codes, Daniel Costello Jr., Adrish Banerjee (University of Notre Dame), 15:40-16:00 On Model and State Estimation un- Francesca Vatta (Universit`adi Trieste), der Mixed Uncertainty, Bartolo Scanavino (Politecnico di Torino) Irina Digailova, Alexander B.Kurzhanski In this paper, we study the convergence behavior of (Moscow State (Lomonosov) University) nonsystematic feedback convolutional encoders used The paper deals with joint estimation of the state as constituent encoders in a parallel concatenated space variables and the transition function of a sys- (turbo) coding scheme. We use mutual information tem with original linear structure through available based EXIT charts to visualize the decoding trajec- measurements corrupted by uncertain disturbances. tory of different nonsystematic encoders employing These are of the ”mixed’ type, combining unknown an iterative MAP decoding algorithm. It is observed but bounded, stochastic Gaussian and H-infinity type that encoders with low weight feedforward inverses disturbances. The results are given in either recur- perform better under low signal-to-noise ratio condi- rent pointwise form or through ”infomation sets” or tions. Catastrophic encoders, which have an infinite ”confidence domains”. weight feedforward inverse, can also be made to con- verge by doping the code with some systematic bits. We also present some BER performance curves for nonsystematic turbo codes and compare them to sys- tematic turbo codes. MTNS – 2002 57

18:00-18:30 Some Small Cyclic Convolutional 17:00-17:30 The Bang-Bang Principle for the Codes, Goursat-Darboux Problem, Heide Gluesing-Luerssen, Wiland Schmale, Melissa Dariusz Idczak (University of Lodz) Striha In the paper, the bang-bang principle for a control sys- (University of Oldenburg) tem connected with a system of linear nonautonomous In this note we will construct and investigate some partial differential equations of hyperbolic type (the small cyclic convolutional codes. Among other things so-called Goursat-Darboux problem or continuous we will present an infinite series of one-dimensional Fornasini-Marchesini problem) is proved. Some den- CCCs over the field with 4 elements having length 3 sity result for piecewise constant controls is also ob- and increasing constraint length (complexity). Our tained. computations show that the first codes in this series have very good free distances. 17:30-18:00 Elimination of Anticipation of Sin- gular 2D Roesser Model, Tadeusz Kaczorek (Technical University Warsaw) Necessary and sufficient conditions for the anticipa- Room: 126, Session: TUP2 tion of singular 2D Roesser model are established. Chair: Raimund Ober Three methods of elimination of the anticipation of Title: Immunology 2: Microscopy and Bio- the model are proposed. In the first method the elim- physics ination of the anticipation is achieved by connection 16:30-17:10 Microscopic Investigation of in series with the anticipatory model of a suitable Synapse Formation, number of delay elements. In the next two methods Michael Dustin (NYU School of Medicine) the elimination of anticipation is achieved by suitable No Abstract choice of feedback gain matrices. The methods are illustrated by numerical examples.

17:10-17:50 Studying Protein-Protein Interac- tions: Biosensor Technology, 18:00-18:30 Difference Equations and n-D Dis- Peter Schuck (National Institutes of Health) crete Systems, No Abstract Jiri Gregor (Czech Technical University) Discrete systems are often described by difference equations. With some attributes of initial state type, 17:50-18:30 Protein Dynamics near Membrane a difference equation defines a system, i.e. a set of Surfaces: New Aspects of Local Coupled Re- (input,output) pairs. For systems defined by linear action and Transport, difference equations (with constant or variable coeffi- Nancy L. Thompson (University of North Carolina at cients) on arbitrary subsets of the n-dimensional grid Chapel Hill) sufficient conditions of their stability are formulated. No Abstract These conditions also enable to obtain growth esti- mates for outputs of such systems. Examples illus- trate the results.

Room: 129, Session: TUP3 Chair: Krzystof Galkowski, Eric Rogers, Victor Vin- nikov Room: 136, Session: TUP4 Title: Multidimensional Systems 2 Chair: Matthias Kawski 16:30-17:00 State Representation of nD Behav- Title: Nonlinear Systems and Control 3 iors, 16:30-16:50 Disturbance Attenuation for a Class Isabel Br`as,Paula Rocha of Nonlinear Systems by Output Feedback, (University of Aveiro, Portugal) Wei Lin, Xianqing Huang In this paper we study the existence of state/driving- (Case Western Reserve University), variable (SDV) representations for discrete nD kernel Chunjiang Qian (University of Texas at San Antonio) behaviors. It turns out that, as in the 2D case, an This paper studies the problem of disturbance atten- nD kernel behavior is SDV-representable if and only uation with internal stability via output feedback for if it has an SDV-representable autonomous part. Ad- a family of nonlinear systems. Using a feedback domi- ditionally, we give a necessary condition for the state nation design which substantially differs from the sep- representability of nD autonomous behaviors. aration principle, we explicitly construct a dynamic MTNS – 2002 58 output compensator attenuating the disturbance’s ef- Room: 209, Session: TUP5 fect on the output to an arbitrary degree of accuracy Chair: Ruth Curtain in the L2 gain sense and achieving GAS in the absence Title: Infinite Dimensional Systems of disturbance. 16:30-17:00 Observability Analysis of a Nonlin- ear Tubular Reactor, 16:50-17:10 A Linear Controller for a Multi- Cedric Delattre, Denis Dochain (CESAME UCL), frequency Model of a Pulse-Width-Modulated Joseph Winkin (University of Namur (FUNDP)) Cuk Converter, An observability analysis is performed on a linear Yusuf Fuad, J.W. van der Woude, W.L. de Koning PDE with a spatial-dependent coefficient, that is (Delft University of Technology) the linearized model of a nonlinear tubular bioreac- This paper presents a linear controller for a mul- tor. It is shown that the associated linear infinite- tifrequency model of a pulse-width-modulated Cuk dimensional operator is a Riesz-spectral operator and converter. The controller is based on an equivalent that a finite number of dominant modes of the sys- discrete-time version of the linearized multifrequency tem are observable when the substrate concentration model and uses the continuous-time output over one is measured at the reactor output. This result is con- switching period. Using this combination we deter- firmed by a numerical simulation. mine the duty ratio at the beginning of every switch- ing period. We discuss how this controller, designed for an arbitrary number of harmonics, can be ap- 17:00-17:30 A Hilbert Space Approach to Self- plied in a realistic situation. Simulations are given Similar Systems, to demonstrate the influence of the controller on the Mamadou Mboup response of the system. (Universit´eRen´eDescartes - Paris V) This paper investigates the structural properties of linear self-similar systems, using an invariant subspace 17:10-17:30 Synergetic Synthesis of Nonlinear approach. The self-similar property is interpreted Kinematics Regulators for Mobile Robots, in terms of invariance of the corresponding transfer Boris Topchiev function space to a given transformation in a Hilbert (Taganrog State University, Russia) space, in a same way as the time invariance prop- We suggest new control strategy for mobile robots. erty for linear systems is related to the shift-invariance This strategy based on forming some desired behavior of the Hardy spaces. The transformation in question structure. Desired structure is introducing to control is exactly that defining the de Branges homogeneous laws by using methods of synergetic theory of control. spaces. The explicit form of the corresponding im- pulse response, which is shown to be described by a hyperbolic partial differential equation, is given. 17:30-17:50 On the Convergence of a Feedback Control Strategy for Multilevel Quantum Sys- tems, 17:30-18:00 Boundary Observability in the Paolo Vettori (University di Padova) Quasi-Static Thermoelastic Contact Problem, We present in this paper a class of feedback strate- Michael Polis, Irina Sivergina gies that solve the steering problem for finite dimen- (Oakland University) sional quantum systems. The control is designed to We study the observability properties of a system that let a suitable distance between the state and the tar- models the temperature evolution of a thermoelastic get decrease. Sufficient conditions are given to ensure rod that may come into contact with a rigid obsta- convergence of this process. cle. The system dynamics are described by a one- dimensional nonlocal PDE of parabolic type with a nonlinear and nonlocal boundary condition. For a 18:10-18:30 Global Output Feedback Control specified observation operator, we show the system to with Disturbance Attenuation for a Class of be observable and get the estimates for the boundary Nonlinear Systems, temperature and the current state of the system. Xianqing Huang, Wei Lin (Case Western Reserve University) No Abstract 18:00-18:30 Modeling Distributed Parameter Systems with Discrete Element Networks, Fabien Soulier, Patrick Lagonotte (LET - ENSMA) In this paper, we compare three general methods aimed to obtain reduced-order models of simple dis- MTNS – 2002 59 tributed parameter systems. We introduce the no- n, is orthogonal to that boundary, pointing inwards. tion of infinite-order impedance and then we use the Mittag-Leffler theorem or the continued fraction ex- pansion in order to represent models in the form of 17:30-17:50 PID Robust Control via Genetic Al- electrical networks. Eventually, the modeling of a gorithms and Integral Criteria Minimization, cooling fin is taken as an example. Catalin Nicolae Calistru, Oana Geman (Technical University of Ia¸si,Romania) No Abstract

Room: 210, Session: TUP6 Chair: Avraham Feintuch 17:50-18:10 MIMO Systems Properties Preser- Title: Robust and H-Infinity Control and Esti- vation under SPR Substitutions, mation Juan Carlos Martinez-Garcia (Instituto Mexicano del Petr´oleo, G. Fern`andez-Anaya 16:30-16:50 Simultaneous Robust Regulation (University Iberoamericana) and Robust Stabilization with Degree Con- This paper tackle the preservation of positive real straint, properties in Multi-Input Multi-Output transfer func- Ryozo Nagamune (KTH) tions, when performing substitutions of the Laplace This paper characterizes all controllers to a problem of variable s by strictly positive real functions of relative simultaneous robust regulation and robust stabiliza- degree equal to zero. We consider also the preserva- tion. The controller set will be represented in terms tion of stability properties of a class of unforced linear of a solution set to the boundary Nevanlinna-Pick in- time-invariant systems affected by a memoryless, pos- terpolation problem. It is shown that a certain degree sibly time-varying nonlinear, input which depends on restriction on the solution set leads to a reasonable the system output. degree bound of controllers. Using the freedom in the controller set, other specification may be satisfied without increasing the controller degree. 18:10-18:30 State Feedback Mixed H2/H- Infinity Problem for Linear Systems with Fi- nite Jumps, 16:50-17:10 Closed-Loop Structure of Discrete- Vasile Dragan, Adrian Stoica Time H-infinity Controller, (Romanian Academy) Waree Kongprawechnon (Thammasat University), The paper consider the mixed H2/ H-infinity problem Shun Ushida, Hidenori Kimura (University of Tokyo) for linear systems with jumps. There are two main This paper is concerned with the investigation of the reults proved in the paper. The first one provides an closed-loop structure of an discrete-time H-infinity evaluation of the norm induced by the inputs of ex- control system. It is shown that discrete-time H- ponentially stable systems with jumps. The second infinity controller is represented, like Linear Quadratic result gives the solution of the state feedback mixed Gaussian Control, as a pseudo state feedback, that H2/ H-infinity problem. is, a state feedback interconnected with an observer. However, in the discrete-time H-infinity control prob- lem the control structure is more complicated since we cannot choose the state feedback and the observer independently. Wednesday August 14, 2002

17:10-17:30 On a Recursive State-space Method 8:00-9:00 Room: 101 Plenary Talk for Discrete-time H2-Approximation, J. William Helton (University of California San Ralf Peeters (University of Maastricht), Diego) Martine Olivi (INRIA Sophia-Antipolis), Manipulating Matrix Inequalities Automati- Bernard Hanzon cally The discrete-time H2-approximation problem for Matrix inequalities have come to be extremely impor- MIMO systems is reduced into an optimization prob- tant in systems engineering in the past decade. This is lem over stable all-pass systems, for which a recursive because many systems problems convert directly into state-space method is considered. A particular atlas matrix inequalities. Matrix inequalities take the form of overlapping generic charts involving balanced real- of a list of requirements that polynomials or ratio- izations is used. It is shown that the gradient of the nal functions of matrices be positive semidefinite. Of H2-criterion at any local optimum of order n-1 em- course while some engineering problems present ra- bedded in the boundary of a chart of systems of order tional functions which are well behaved, many other MTNS – 2002 60 problems present rational functions which are badly 9:00-10:00 Room: 136 Invited Talk behaved. Thus taking the list of functions which a de- Knut Hueper (University of W¨urzburg, Germany), sign problem presents and converting these to a nice The Dynamics of Matrix Eigenvalue Algo- form, or at least checking if they already have or do rithms not have a nice form is a major enterprise. Since ma- We present a calculus approach to the convergence trix multiplication is not commutative, one sees much analysis of matrix eigenvalue algorithms including effort going into calculations (by hand) on noncom- those methods as Jacobi, RQI, shifted QR, or Ric- mutative rational functions. A major goal in sys- cati iterations. New algorithms are developed, being tems engineering is to convert “noncommutative in- somehow the true generalization of Rayleigh quotient equalities” to equivalent Linear Noncommutative In- iteration to full matrices, featuring cubic convergence equalities (that is, to LMI’s), if possible. The talk on all eigenvectors simultaneously. concerns efforts to process “noncommutative inequal- ities” using computer algebra. The most basic efforts, such as determining when noncommutative polynomi- als are positive, convex, convertible to noncommuta- tive LMI’s, transformable to convex inequalities, etc., Morning: force one to a rich area of undeveloped mathematics. Room: 102, Session: WA1 Chair: Raimund Ober Title: Immunology 3: Structure and Kinetics 10:30-11:10 Geometrical Methods in Structural Molecular Biology, 9:00-10:00 Room: 101 Invited Talk Timothy F. Havel (MIT) Structural molecular biology is concerned with the Jan C. Willems (University of Leuven, Belgium) role played by the three-dimensional structures of Dissipative Distributed Systems DNA, proteins and other molecules in biochemical This talk deals with systems described by constant processes. A number of important problems in this coefficient partial differential equations. We start by field can be formulated in purely geometric terms, us- introducing the behavior of such systems, controlla- ing lower and upper bounds on the interatomic dis- bility, and kernel and image representations. We then tances to describe the sets of configurations of inter- turn to dissipative systems. The main issue here is the est. The invariant theory of the Euclidean group, also construction of a storage function. This leads to an known as distance geometry, is the basis of practical application of Hilbert’s 17-th problem on the factor- algorithms for computer- aided geometric reasoning ization of polynomials in many variables. Throughout with such distance constraints. This talk will describe the talk, we use Maxwell’s equations as a running ex- the applications of these methods to protein homol- ample for motivating the problems and the concepts. ogy modeling and structure determination by nuclear The research on which this talk is based, is joint work magnetic resonance spectroscopy. with Harish Pillai from the IIT Bombay.

11:10-11:50 Kinetic aspects of TcR-MHC and Antibody-Antigen Interactions, Jefferson Foote (Fred Hutchinson Cancer Research 9:00-10:00 Room: 102 Invited Talk Center) Vertebrates develop immunity when they encounter a Albert-Laszlo Barabasi (University of Notre Dame), new microbe or foreign substance (antigen) because The Architecture of Complexity: Emergence of a set of cells called lymphocytes. The immune re- of Scaling in Complex Networks ponse is a Darwinian process in which small numbers Systems as diverse as the world wide web or the of antigen-recognizing lymphocytes are selected out of cell are described by networks with complex topol- a vast pool of lymphocytes that don’t recognize anti- ogy. Traditionally it has been assumed that these gen. Each lymphocyte has on its surface unique anti- networks are random. However, recent studies indi- gen receptors whose molecular structure determines cate that such complex networks are the result of self- fitness of the lymphocyte in such a selection. One organizing processes governed by simple but generic type of lymphocyte, a B cell, has a surface receptor laws, resulting in topologies strikingly different from called an antibody. When antigen is present, parame- those predicted by random networks. I will discuss the ters governing these cells’ fitness are the rate at which implications of these findings on the error and attack the antibody binds antigen (faster is better) and the tolerance of the Internet, the robustness of the cells, lifetime of the antibody-antigen complex once it has and other properties of complex evolving networks. formed (longer is better, up to a point). A second type MTNS – 2002 61 of lymphocyte, a T cell, has a surface antigen recep- vices. We review probabilistic performance bounds tor called simply a T cell receptor (TCR). A TCR is that apply to PSRG nodes, and improve a previous structurally similar to an antibody, but is specialized bound for loss probability. to react with the complex formed by embedding an antigen fragment in a cellular molecule called a Ma- jor Histocompatibility Complex molecule (MHC). At 11:20-11:40 Statistical Performance Analysis of a an early stage of development, T cells that either don’t Generalized Processor Sharing System by Us- bind MHC at all or bind it too tightly are eliminated. ing Large Deviations, Fitness here, in advance of any antigen contact, de- Min Xie, Martin Haenggi pends on a weak interaction with MHC that poten- (University of Notre Dame) tially could be improved if an antigen fragment were This paper reviews the statistical performance anal- present. Parameters subsequently governing fitness ysis of the tail of the steady-state queue length dis- during an immune response are again rate of forma- tribution in a Generalized Processor Sharing (GPS) tion and lifetime of the TCR:(antigen+MHC) com- system. In particular, we focus on the Large Devi- plex. Fast binding is probably an advantage, but a ations Principle (LDP) which can be used to esti- long lifetime is a distinct disadvantage. The optimal mate a wide range of traffic sources, including short- recognition signal occurs when a an antigen+MHC range dependent sources, like processes with Marko- molecule engages a TCR for some threshold time, dis- vian structure, and long-range dependent sources, like engages, engages with another TCR on the same cell, self-similar traffic. Generally, performance bounds, ei- disengages, and so on. ther upper or lower, are derived to characterize the asymptotic behavior of the queue length for each ses- sion which shares the GPS server with multiple ses- 11:50-12:30 Biophysical Considerations of T-Cell sions. Other analysis methods, such as service-curve Receptor-Peptide/MHC Interactions, based method, and bufferless fluid flow approxima- Brian M. Baker tions are introduced briefly. Finally, a comparison of No Abstract different methods and the range where they can be applied are given.

Room: 126, Session: WA2 11:45-12:05 Resource Allocation and Congestion Chair: Martin Haenggi Control in Distributed Sensor Networks - a Title: Computer Networks Network Calculus Approach, Jinsong Zhang, Kamal Premaratne 10:30-10:50 Min-Plus System Theory Applied to (University of Miami) Communication Networks, Peter Bauer (University of Notre Dame) Patrick Thiran, Jean-Yves Le Boudec The establishment of the overall objectives of a dis- (EPFL) tributed sensor network is a dynamic task so that it Network Calculus is a set of recent developments, may sufficiently well ’track’ its environment.To inte- which provide a deep insight into flow problems en- grately perform both resource allocation to each in- countered in networking. It can be viewed as the sys- put data flow and congestion control at each deci- tem theory that applies to computer networks. Con- sion node of such a network,a ’per-flow’ virtual queu- trary to traditional system theory, it relies on max- ing framework that decouples the input data flows to plus and min-plus algebra. In this paper, we show how each decision node is introduceded.Under this frame- a simple but important fixed-point theorem (residua- work,network calculus notions are utilized to model tion theorem in min-plus algebra) can be successfully the end-to-end flow and design a simple yet effective applied to a number of problems in networking, such feedback control law for each input data flow.The con- as window flow control, multimedia smoothing, and trol strategy is robust against the time-varying na- bounds on loss rates. ture of network delays and buffer depletion rate,and source-node rate cutoff/saturation. 10:55-11:15 Elements of Probabilistic Network Calculus for Packet Scale Rate Guarantee 12:10-12:30 Optimal Media Streaming in a Rate- Nodes, Distortion Sense For Guaranteed Service Net- Milan Vojnovic, Jean-Yves Le Boudec works, (EPFL) Olivier Verscheure, Pascal Frossard Packet Scale Rate Guarantee (PSRG) is a node model (IBM T.J. Watson Research Center), used by IETF for Expedited Forwarding, a priority Jean-Yves Le Boudec (EPFL) service defined in the context of Differentiated Ser- MTNS – 2002 62

We present an optimal low-complexity scheduling Room: 136, Session: WA4 strategy of continuous media in a rate-distortion sense for guaranteed service networks. First we consider a Chair: Paul Van Dooren stored and offline-compressed media stream. We give Title: Model Reduction an explicit representation of the optimal smoothing strategy, which minimizes the required playback de- 10:30-10:50 An Overview of Model Reduction lay and decoding buffer size. Then we provide the Methods for Large-Scale Dynamical Systems, theoretical bounds on the media rate such that (i) the Thanos Antoulas (Rice University) optimal smoothing solution meets some constraints No Abstract on the admissible playback delay and maximum de- coding buffer size, and (ii) the media size is max- imum. Finally we cast the rate-distortion problem 10:50-11:10 Analysis of Smith-Type Methods for as a piece-wise linear convex optimization algorithm, Lyapunov Equations and Balanced Model Re- which is solved efficiently using state-of-the-art linear duction, programming techniques. Dan Sorensen (Rice University) No Abstract

11:10-11:30 Krylov Subspace Techniques for Re- duced Order Modeling of Nonlinear Dynamical System, Room: 129, Session: WA3 Daniel Skoogh (Swedish Defence Research Agency), Zhaojun Bai (UC Davis) Chair: Eric Rogers We discuss reduced-order modeling of nonlinear dy- Title: Minicourse B: Multidimensional Systems namical systems by Krylov subspace methods. We 10:30-11:10 Recent Results on Multidimensional focus on a method which the nonlinear system is first Behaviors, approximated by a bilinear system through Carleman Eva Zerz (University of Kaiserslautern, Germany) bilinearization. A reduced-order bilinear system is In the behavioral framework, control is based on the constructed such that it matches certain number of interconnection of systems. We examine this approach multimoments corresponding to the first few kernels and give some related results for behaviors that obey of the Volterra-Wiener representation of the bilinear Oberst’s duality. There is a close relation to the ho- system. mological concept of extension modules. We will also address the question of causality, and we study first order representations of multidimensional behaviors. 11:30-11:50 Model Reduction of Second Order Systems, Younes Chahlaoui, D. Lemonnier, K. Meerbergen, A. 11:10-11:50 Motivation and General Concepts in Vandendorpe, P. Van Dooren Behavioral Systems, (Universit´eCatholique de Louvain) Jan C. Willems (University of Leuven, Belgium) In this paper we look at model reduction of second or- The aim of this introductory lecture is to outline tha der linear time-invariant system originating from me- basic philosophy that underlies the behavioral ap- chanical systems. We propose several methods pro- proach to modeling open dynamical systems. Special ducing reduced order models of second order as well, emphasis will be given to multi-dimensional systems. since it makes sense to impose the reduced-order sys- We will discuss some salient issues, such as kernel, tem to be of the same type than the original one. image and latent variable representations, the elimi- nation problem, and controllability. 11:50-12:10 Model Reduction via Tangential In- terpolation, 11:50-12:30 Similarities/Differences Between the Antoine Vandendorpe, K. Gallivan, P. Van Dooren Behavioral Approach for Multidimensional (Universit´eCatholique de Louvain) versus Delay-Differential Systems, In this paper, we study the problem of model reduc- Heide Gluesing-Luerssen, tion of a state space system via tangential interpola- (University of Oldenburg, Germany) tion. We show that the problem has a unique solution This presentation will focus on what the structural if a particular Loewner matrix constructed from the common ground between these the two different sys- data is nonsingular. Under these conditions the inter- tems classes and why do the theories kind of diverge polating system can also be obtained as a projection when it comes to the details. of the original system. This is a generalisation of the MTNS – 2002 63 well known Multipoint Pade technique for tangential the proposed controllers consists in the extension of interpolation. derivation and integration order from integer to non integer numbers. This approach provides a more flex- ible tuning strategy and therefore an easier achieving of control requirements with respects to classical con- Room: 208, Session: WA5 trollers. Chair: Daniel Alpay, Yuli Eidelman Title: Time-Varying Systems and Numerical 11:00-11:30 Nonlinear Discrete-Time Observer Problems Design with Linearizable Error Dynamics, 10:30-11:00 Unbounded J-inner Sections, MingQing Xiao (Southern Illinois University at Car- Patrick Dewilde (Delft University of Technology), bondale), Daniel Alpay (Ben-Gurion University of the Negev) Nikolaos Kazantzis (Worcester Polytechnic Institute), Rational J-inner-valued sections which are J-inner Costas Kravaris (University of Patras), w.r. to the unit circle (J is a signature matrix) play an Arthur J Krener (UC Davis) important role in interpolation theory and related the- A necessary and sufficient condition for the existence ories for optimal signal recovery and optimal control. of a discrete-time nonlinear observer with linearizable A now classical result is the extension of these theo- error dynamics is provided. The necessary and suffi- ries to the time-varying context. These theories work cient condition derived is associated with the solvabil- with bounded J-inner operators, here we consider the ity of a nonlinear functional equation. A corollary is unbounded case. It gives rise to new, interesting and Siegel’s theorem on the linearizability of a mapping. intriguing results. The method of observer design suggested by this theo- rem is constructive and can be applied approximately to any sufficiently smooth, linearly observable system 11:00-11:30 Linear Time-Varying Darlington yielding a local observer with approximately linear er- Synthesis, ror dynamics. Avraham Feintuch (Ben-Gurion University) No Abstract 11:30-12:00 Analysis of Periodic Solutions of Tapping-Mode AFM: An IQC Approach, 11:30-12:00 Reduction to System Methods for Murti Salapaka, Abu Sebastian Inversion of Diagonal Plus Semiseparable Op- (Iowa State University) erator Matrices, The feedback perspective with the cantilever viewed Yuli Eidelman, Israel Gohberg (Tel-Aviv University) as a linear system and the tip-sample interaction ap- We consider operator equations in a Hilbert space pearing as a nonlinear feedback is useful in analyzing with a bounded linear operator represented by a fi- AFM (Atomic Force Microscope) dynamics. Condi- nite operator matrix which is a sum of a diagonal tions for the existence and stability of periodic solu- and of a semiseparable operator matrices. We discuss tions for such a system when forced sinusoidally are three methods for solution of such equations which obtained. These results are applied to the case where are based on reduction of the considering problem the AFM is operated in the tapping-mode. The near to a finite system of linear difference equations with sinusoidal nature of periodic solutions is established boundary conditions. The obtained results yield new by obtaining bounds on the higher harmonics. The algorithms for solution of integral equations and for concept of Integral Quadratic Constraints (IQC) is inversion of matrices. widely used in the analysis.

12:00-12:30 Bifurcations of the Controlled Es- Room: 209, Session: WA6 cape Equation, Chair: Erik Verriest Tobias Gayer, (University of Augsburg) Title: Nonlinear Systems and Control 4 In this paper we present numerical methods for the analysis of nonlinear autonomous control systems and 10:30-11:00 Parameter Tuning of a Non Integer two conditions, local accessibility and an inner-pair Order PID Controller, condition, under which they can be applied. These Luigi Fortuna, Riccardo Caponetto (University of methods can be extended to work also for systems Catania) with time-periodic right hand side. In particular, the Domenico Porto (STMicroelectronics, Catania) escape equation with sinusoidal driving term and ad- In this paper a new type of PID controllers is intro- ditional control is analyzed. We will show that its duced and some properties are given. The novelty of MTNS – 2002 64 stability behavior undergoes interesting bifurcations invoke a very different behaviour of the feedback con- when the range of the control influence is varied. trol system. Subharmonic oscillations are shown to be normal unstable behaviour.

Room: 210, Session: WA7 11:50-12:10 On the Control of the Resonant Con- Chair: Rodolphe Sepulchre verter: A Hybrid-Flatness Approach, Title: Discrete Event and Hybrid Systems Hebert Sira-Ramirez, Ramon Silva-Ortigoza (CINVESTAV-IPN) 10:30-10:50 Switched Systems that are Periodi- In this article we show that the series resonant cally Stable may be Unstable, DC/DC converter, which is a hybrid system, is piece- Jacques Theys, Vincent Blondel wise differentially flat with a flat output which is in- (Catholic University of Louvain), variant with respect to the structural changes under- Alexander Vladimirov (Russian Academy of Science) gone by the system evolution. This fact considerably We prove the existence of pairs of matrices A0 and simplifies the design of a switching output feedback A1 with real entries that are such that all infinite pe- controller that can be essentially solved by linear tech- riodic products of the two matrices converge to zero, niques. Flatness clearly explains all practical issues but some infinite non-periodic product does not. associated with the normal operation of the converter.

10:50:11:10 The Servo Problem for Piecewise 12:10-12:30 Controllability of Periodically Linear Systems, Switched Linear Systems with Delay in Con- Stefan Solyom, Anders Rantzer trol, (Lund Institute of Technology) Guangming Xie, Long Wang, Yijing Wang The servo problem for a wide class of nonlinear sys- (Peking University, Beijing) tem is considered. A quantitative bound on system The controllability for switched linear systems with trajectories is derived. For piecewise linear systems time-delay in control are first formulated and investi- the bound is shown to be computable in terms of lin- gated. A sufficient and necessary condition for con- ear matrix inequalities. trollability of periodically switched linear systems is presented. Furthermore, it is proved that the control- lability can be realized in n + 1 periods at most. An 11:10-11:30 Stability of Hybrid Control Systems example illustrates the above results. Some further Based on Time-State Control Forms, results are also presented. Yoshikatsu Hoshi, Mitsuji Sampei, Shigeki Nakaura (Tokyo Institute of Technology) Time-State Control Form was proposed as a con- trol method for nonholonomic systems. This method needs input switching, so there is a drawback that Middle: the switching conditions may spoil the stability of the Room: 102, Session: WM1 system. From the above backgrounds, this paper con- siders the stability of Time-State Control Form from Chair: Raimund Ober the viewpoint of hybrid systems. We derive two kind Title: Immunology 4: Diffusion and Modelling of sufficient conditions which stabilize the systems by 14:00-14:40 Measuring Lateral Diffusion and As- using Lyapunov functions. sociations of MHC Molecules in Membranes of the ER and at the Cell Surface, Michael Edidin (Johns Hopkins University) 11:30-11:50 Discrete-Time Modeling and Analy- No Abstract sis of Pulse-Width-Modulated Switched Power Converters, Willem L. De Koning, Jacob W. Van Der Woude, 14:40-15:20 A Computational Model for T Cell Yusuf Fuad Receptor Signal Integration, (Delft University of Technology) Mark Alber (University of Notre Dame), The subject of this paper is a general theory for Arancha Casal, Cenk Sumen, Tim Reddy, Mark switched power converters, where the switch is pulse- Davis, Peter Lee width-modulated. The existence, uniqueness and (Stanford University) stabi- lity of stationary solutions is investigated. Fur- T cells play a central role in orchestrating the adap- thermore, feedback control situations are considered. tive immune response. Specificity of T cells is de- It is shown that a running and fixed modulator, may fined by each cell expressing a unique population of MTNS – 2002 65

T cell receptors (TCRs) which interact with specific by the forward recursion of the sum-product (Baum- peptide-MHC (pMHC) on other cells, such as antigen- Welch, BCJR) algorithm. This result drastically re- presenting cells (APC).XSCC During recognition, the duces the space over which the optimal feedback- T cell surface comes in direct contact with that of dependent source distribution needs to be sought. the APC. Within the first 30-60 seconds of contact, Further, the feedback capacity computation may then the T cell makes the critical decision to either sus- be formulated as an average-reward-per-stage stochas- tain the interaction, eventually leading to activation, tic control problem, for which numerical solutions or to disengage and move on. Over this brief period, of Bellman’s equation deliver the feedback-capacity- thousands of TCRs interact with thousands of differ- achieving source distribution. With the knowledge of ent pMHCs with a wide range of affinities. The T the capacity-achieving source distribution, the value cell must quickly survey and integrate this vast array of the capacity is easily estimated using accurate of signals into its critical decision. Learning how a Markov chain Monte Carlo methods. We demon- T cell perceives this vast, heterogeneous repertoire is strate the applicability of the method by computing critical to understanding not only T cell biology, but the feedback capacity of partial response channels and also certain disease evasion mechanisms. Mathemat- the feedback capacity of run-length-limited (RLL) se- ical models and computer simulations can help over- quences over binary symmetric channels (BSCs). come experimental limitations. If biologically-sound and well-informed, a computational model such as the one we propose here can result in valuable insights into 14:40-15:00 Kalman Filtering, Factor Graphs, the T cell recognition process. and Electrical Networks, Pascal O. Vontobel, Dani Lippuner, Hans-Andrea Loeliger 15:20-16:00 Immunological Synapse Formation: (ETH) A Crossroad of Physical Chemistry and Cell Factor graphs are graphical models with origins in Biology, coding theory. It is well known that Kalman filtering Arup K. Chakraborty is an instance of the generic sum(mary)-product algo- (University of California, Berkeley) rithm on the corresponding factor graph. In this pa- No Abstract per, a one-to-one correspondence between such factor graphs and a class of electrical networks is presented. The electrical network ”computes” correct Bayesian estimates even for factor graphs with cycles. Room: 126, Session: WM2 Chair: Aleksandar Kavcic Title: Control and Communications 15:00-15:20 Kalman Filtering Applied to Timing Recovery in Tracking Mode, 14:00-14:20 Feedback Capacity, Panu Chaichanavong (Stanford University) Sekhar Tatikonda, Sanjoy Mitter Brian Marcus (IBM Almaden) (MIT) This paper investigates the performance of the In this paper we provide a general framework for Kalman filter as a timing recovery system in track- characterizing the capacity of a large variety of chan- ing mode, subject to a wide range of operating con- nels with memory, side-information, intersymbol in- ditions. Our simulation compares the Kalman filter terference, and feedback. For Markov channels we and the phase-locked loop based on the number of di- show that one can formulate the capacity optimiza- vergences for various values of timing disturbance and tion problem as a dynamic program and compute the SNR. They are shown to work equally well in low SNR capacity using tools from approximate dynamic pro- and disturbance. However, the Kalman filter outper- gramming. Feedback codes are then discussed. forms the phase-locked loop when both SNR and dis- turbance are high. The simulation also suggests that the Kalman filter is more robust to variations in op- 14:20-14:40 Sum-Product Algorithm and Feed- erating conditions. back Capacity, Shaohua Yang, Aleksandar Kavcic (Harvard University) 15:20-15:40 Lower Bounds for the Performance In this paper, we explore the link between the sum- of Iterative Timing Recovery at low SNR, product algorithm and the feedback capacity of a Aravind Nayak, J. Barry, S. McLaughlin channel with memory. We show that the optimal (Georgia Institute of Technology) (i.e., capacity-achieving) feedback is captured by the We consider the problem of timing recovery at low causal posterior state probabilities. For finite-state signal-to-noise ratio. We first derive a lower bound for machine channels, the optimal feedback is captured MTNS – 2002 66 the timing estimation error variance in the presence computations may be reduced. For a 3-channel multi- of a random walk timing jitter, for the PR-IV chan- rate system, an algorithmic version of Suslin’s stabil- nel. Then, we look at the trained phase-locked loop, ity theorem may be useful for factoring the polyphase which gives a hueristic lower bound for the perfor- matrices. mance of iterative timing recovery schemes involving phase-locked loops.

15:40-16:00 Classical Capacity of Quantum Channels, Navin Khaneja (Harvard University) No Abstract Room: 136, Session: WM4 Chair: Uwe Helmke Title: Control and Computation 14:00-14:30 Continuation of Eigendecomposi- Room: 129, Session: WM3 tions, Chair: Krzystof Galkowski, Eric Rogers, Victor Vin- Luca Dieci (Georgia Tech) nikov In this talk we discuss the problem of smooth con- Title: Multidimensional Systems 3 tinuation of eigendecompositions. In particular, we 14:00-15:00 Conservative Multidimensional Sys- discuss the cases of the Schur decomposition and of tems: A Survey, bi-diagonal form. Connected to these, we discuss con- Joseph A. Ball (Virginia Tech) tinuation of Hessenberg forms. This is joint work with We survey a number of classes of ISO systems for A. Papini, Univ. of Florence. which an energy balance relation holds. Such systems lead to transfer functions with membership in various types of multivariable generalizations of the classical 14:30-5:00 Numerical Solution of Euclidean Bal- Schur class, including holomorphic functions of several anced Norm Realizations via Gradient Flows, variables, bundle maps between bundles over a Rie- N. Del Buono, L. Lopez mann surface and formal power series in noncommut- No Abstract ing indeterminates. Connections with Lax-Phillips scattering in this context will also be discussed. 15:00-15:30 Controllability of the QR Algorithm on Hessenberg Flags, 15:00-15:30 On J-Conservative Scattering nD Uwe Helmke, Jens Jordan System Realizations, (University of W¨urzburg) Dmitry Kalyuzhniy-Verbovetzky The shifted QR algorithm can be interpreted as a (The Weizmann Institute of Science) nonlinear discrete dynamical system on the flag man- We extend the notion of discrete conservative scat- ifold. In the complex case we describe the reachabil- tering nD system to the case where its state space ity sets as orbits of a group action and prove non- is endowed with the structure of a Krein space by controllability of the algorithm. In contrast, the algo- means of a canonical symmetry J. We call such sys- rithm restricted on the subset of Hessenberg flags is tems J-conservative. We prove that any operator- generically controllable. valued function holomorphic on some neighbourhood of the origin in the complex space of n variables and vanishing at the origin has a J-conservative scattering 15:30-16:00 The Continuos-Time Rayleigh Quo- nD system realization. tient Flow on the Grassmann Manifold, Rodolphe Sepulchre, P.-A. Absil (University of Liege), 15:30-16:00 Factorization of M-D Polynomial R. Mahony (Australian National University) Matrices for Design of M-D Multirate Systems, An extension of the Rayleigh quotient iteration (RQI) Mikhail Tchobanou (Moscow Technical University), to the Grassmann manifold has been recently pro- Cynthia Woodburn (Pittsburg State University) posed for computing a p-dimensional eigenspace of a The problem of the design of effective 2-D and 3-D symmetric matrix A. Here we analyze a continuous- multirate systems with prescribed properties is con- time flow analogous to this Grassmannian RQI. This sidered using tools from commutative algebra. Re- flow achieves deflation in finite time, i.e. it converges sults for factoring 2-channel polyphase matrices are in finite time to a subspace that includes an eigenvec- presented. After such a factorization, the number of tor of A. MTNS – 2002 67

Room: 209, Session: WM5 ularity assumptions, and characterize the structure of Chair: Anthony Bloch the quotient control bundle. We also introduce a no- Title: Algebraic and Differential Geometry in tion of projectability which turns out to be equivalent Systems Theory to controlled invariance. This allows to regard previ- ous work on symmetries, partial symmetries, and con- 14:00-14:20 Hamiltonian Structure of the Alge- trolled invariance as leading to special types of quo- braic Riccati Equation and its Infinitesimal V- tients. We also show the existence of quotients that Stability, are not induced by symmetries or controlled invari- Nanaz Fathpour, Edmond A. Jonckheere ance. (University of Southern California) We will investigate the stability behavior of quadratic maps in higher dimensions. We will establish the con- 15:20-15:40 The Wave Equation as a Port- nection between infinitesimal V-stability of solutions Hamiltonian System, and a Finite Dimensional to the Algebraic Riccati Equations, and the Hamil- Approximation, tonian eigenstructure of the solutions. Infinitesimal Viswanath Talasila, Goran Golo, Arjan van der V-stability of critical points of the Riccati map is cru- Schaft cially related to stability of the Riccati map. Grobner (University of Twente) Bases are used to implement the calculations. The problem of approximating a distributed param- eter system with free boundary conditions is solved for the 2-dimensional wave equation. To this end 14:20-14:40 Global Transformation of Nonlinear we first model the wave equation as a distributed- Dynamic Systems into Canonical Forms, parameter port Hamiltonian system. Then we em- Anna Michtchenko, Aleksey Zhirabok ploy the idea that it is natural to use different finite (Far Eastern State University) elements for the approximation of different geometric The problem of transformation of nonlinear systems variables (forms) describing a distributed-parameter into canonical forms is studied. Both continuous-time system, to spatially discretize the system and we show and discrete-time cases are considered. Sufficient and that we obtain a finite-dimensional port-Hamiltonian necessary conditions for such transformation are ob- system, which also preserves the conservation laws. tained. A special mathematical technique (so-called the algebra of functions) is used. 15:40-16:00 Pseudo Balancing for Discrete Non- linear Systems, 14:40-15:00 A Lie-Group Approach for Nonlin- Erik Verriest (Georgia Institute of Technology) ear Dynamic Systems Described by Implicit A rationale is presented to extend the notion of a bal- Ordinary Differential Equations, anced realization to nonlinear discrete time systems Kurt Schlacher (University of Linz), However, it is shown that even with a set of very rea- Andreas Kugi (University of Saarland), sonable assumptions, it is not possible to construct a Kurt Zehetleitner (University of Linz) globally balanced realization. The obstruction stems This contribution presents a Lie-group based ap- from certain integrability conditions which are gener- proach for the accessibility and the observability prob- ically not satisfied. One way around this is to relax lem of dynamic systems described by a set of im- the requirements of global balancing by restricting the plicit ODEs. It is shown that non-accessible or non- balancedness conditions to a discrete set of points in observable systems admit Lie groups acting on their the state space. solutions such that distinguished parts of the system remain unchanged. The presented methods use the fact that the dynamic system may be identified with a submanifold in a suitable jet-bundle. Room: 210, Session: WM6 Chair: Panos Antsaklis, Anthony Michel 15:00-15:20 Quotients of Fully Nonlinear Con- Title: Hybrid Control System Analysis, Synthe- trol Systems, sis and Diagnosis Paulo Tabuada, George J. Pappas 14:00-14:30 Partial Stability of Dynamical Sys- (University of Pennsylvania) tems, In this paper, we define and study quotients for fully Ye Sun, A.P. Molchanov, A.N. Michel nonlinear control systems. Our definition is inspired (University of Notre Dame) by categorical definitions of quotients as well as re- We develop new results for partial stability of general cent work on abstractions of affine control systems. dynamical systems with respect to invariant sets de- We show that quotients always exist under mild reg- fined on metric space, using stability preserving map- MTNS – 2002 68 pings. Our results are applicable to a much larger 15:30-16:00 Monitoring and Diagnosis of Hybrid class of systems than existing results, including to dy- Systems Using Particle Filtering Methods, namical systems that cannot be determined by the Xenofon Koutsoukos, James Kurien, Feng Zhao usual classical (differential) equations. Furthermore, (Palo Alto Research Center) in contrast to existing results which pertain primarily Embedded systems are composed of a large number of to the analysis of equilibria, the present results ap- components that interact with the physical world via ply to invariant sets (including equilibria as a special a set of sensors and actuators, have their own com- case). We apply our results in the analysis of a class putational capabilities, and communicate with each of discrete event systems (a computer load balancing other via a wired or wireless network. Diagnostic problem). systems for such applications must address new chal- lenges caused by the distribution of resources, the net- working environment, and the tight coupling between 14:30-15:00 An Approach to General Switched the computational and the physical worlds. Our ap- Linear Quadratic Optimal Control Problems proach is to move from centralized, discrete or con- with State Jumps, tinuous techniques toward a distributed, hybrid di- Xuping Xu (Penn State Erie), agnosis architecture. Monitoring and diagnosis of Panos Antsaklis (University of Notre Dame) any dynamical system depend crucially on the abil- Unlike conventional optimal control problems, op- ity to estimate the system state given the observa- timal control problems of switched systems require tions. Estimation for hybrid systems is particularly the solutions of not only the optimal continuous in- challenging because it requires keeping track of multi- puts but also the optimal switching sequences. In a ple models and the transitions between them. This previous paper by the authors, an approach for an paper presents a particle filtering based estimation important class of switched systems optimal control algorithm that addresses the challenge of the inter- problems, namely, general switched linear quadratic action between continuous and discrete dynamics in (GSLQ) problems where each subsystem is linear and hybrid systems. The hybrid estimation methodology the cost functionals are in general quadratic forms, has been demonstrated on a rocket propulsion system. was reported. In this paper, we extend the approach to GSLQ problems with state jumps at the switch- ing instants. For such problems, the cost functionals include not only the general quadratic cost terms for Room: 208, Session: WM7 the state and the input but also the costs for state Chair: Giorgio Picci jumps. The approach in this paper allows us to derive Title: Stochastic Systems 2 the derivatives of the optimal cost with respect to the switching instants based on the solution of the discon- 14:00-14:30 State Space Realization of Random tinuous Riccati equation parameterized by the switch- Processes with Feedback, ing instants along with its differentiations. With the Giorgio Picci, Alessandro Chiuso knowledge of the derivatives, nonlinear optimization (University of Padova) methods can be applied to locate the optimal switch- Construction of the state space of a stochastic process ing instants. An example is provided to illustrate the with exogenous inputs when feedback may be present. approach. 14:30-15:00 Approximate Realization of Hidden 15:00-15:30 The Controlled Composition Analy- Markov Chains, sis of Hybrid Automata, Lorenzo Finesso (LADSEB-CNR, Padova) Ying Shang, M.D. Lemmon Peter Spreij (University of Amsterdam) (University of Notre Dame) No Abstract This paper presents a sufficient condition for the exis- tence of global non-terminating solutions in controlled hybrid automata. The condition is based on a recur- 15:00-15:30 Random Sampling of a Continuous- sive algorithm that can always terminate after a finite Time Stochastic Dynamical System, number of iterations to a limit set of states called the Mario Micheli, Michael I. Jordan inner viability kernel. The inner viability kernel is (Brown University) easily computed in some cases. The inner viability We consider a dynamical system where the state equa- approach can ensure the composition analysis of con- tion is given by a linear stochastic differential equa- trolled hybrid automata. tion and noisy measurements occur at discrete times, in correspondence of the arrivals of a Poisson process. After formulating a Kalman Filter-based state esti- mation algorithm, we compute a complete statistical MTNS – 2002 69 description of the state estimation error process and Recent extensions of the model to deal with other sit- analyze its dependence on the continuous-time sys- uations will also be presented. tem’s dynamics and the sampling rate.

15:30-16:00 The Hilbert Space of an Ergodic Se- quence, Giorgio Picci, (University of Padova) Tom Taylor (Arizona State University) Given a semi infinite time series, it has been shown that under some natural assumptions of existence of Room: 126, Session: WP2 limits of certain time averages, an essentially unique stationary stochastic process exists which produces Chair: Roxana Smarandache the observed trajectory according to the classical Title: Cryptography “urn” scheme of probability theory. 16:30-17:00 A High-Speed Processing for RSA Cryptograms Using High-Radix Signed-Digit Numbers and a New Algorithm of Modulo Op- eration, Afternoon: Yoshinori Fujisawa, Yasushi Fuwa Room: 102, Session: WP1 (Nagano National College of Technology) In this work, we developed a high speed LSI for encod- Chair: Raimund Ober ing and decoding the RSA cryptogram and describe Title: Immunology 5: Cellular Aspects the processing method in this paper. In this paper, 16:30-17:30 Staining Antigen Specific CD4+ T we proposed two new processing algorithms for an -Cells with Class II MHC Oligomers, RSA public key cryptogram system. Then, we de- Lawrence Stern (MIT) veloped LSI to realize our proposed algorithm using No Abstract a sequential processing algorithm for calculating re- mainders. As a result, it was clear that the method using our proposed algorithm works faster than the 17:10-17:40 The Roles of Serial Engagement and previous method. Kinetic Proofreading in Peptide-Induced T- Cell Activation, Dan Coombs, Carla Wofsy, Byron Goldstein 17:00-17:30 On the Rational Cubic Curve Cryp- (Los Alamos National Laboratory) tosystems, The activation of a T-cell requires the formation Xiaochang Wang, Heather Henkel of a long-lived attachment to an antigen-presenting (Texas Tech University) cell (APC). APCs present peptide on their surfaces, In this paper, we study the group on the finite field held by a major-histocompatibility complex (MHC). of q elements plus infinity induced by rational cubic A given T-cell carries receptors (TCRs) specific for curves. We show that the group is isomorphic to either a particular MHC-peptide group, along with other a subgroup of order q+1 of the multiplicative group less specific adhesion and costimulatory molecules. of the finite field of q square elements, or the additive The stable region of close apposition (immunological group, or multiplicative group, of the original field. synapse) may facilitate signal transduction, by con- centrating the TCR and MHC-peptide together and allowing long-lived bond formation. A TCR will be- 17:30-18:00 Public Key Cryptography Based on come activated if it can form a sufficiently long-lived Simple Modules over Simple Rings, bond to allow multiple biochemical changes to occur Gerard Maze(EPFL), (the kinetic proofreading hypothesis). We have de- Christopher Monico (University of Notre Dame), veloped a mathematical model to examine the TCR- Joan-Josep Climent (University of Alicante), MHC-peptide interaction within the stable region and Joachim Rosenthal (University of Notre Dame) use it to study (1) the competing effects of serial en- We show how the action of a (semi)ring on a gagement (sequential activation of TCR by one MHC- (semi)module gives rise to a generalized Diffie- peptide) and kinetic proofreading, and (2) the possible Hellman protocol. This leads naturally to a crypto- role of TCR oligomerization in activation of T-cells. graphic protocol whose security relies on a finite ver- In conjunction with the model, recent experimental sion of a difficult control problem - steering the state data indicates that activated TCR must remain ac- of a dynamical system from an initial vector to some tive for a period after dissociation from MHC-peptide. final location. MTNS – 2002 70

Room: 129, Session: WP3 Adriana Popovici, Dan Emanuel Popovici Chair: Krzystof Galkowski, Eric Rogers, Victor Vin- (University of the West Timisoara, Romania) nikov Cellular automata can be successfully applied in im- Title: Multidimensional Systems 4 age processing. In this paper we discuss the applica- tion of two-dimensional cellular automata to the prob- 16:30-17:00 Spatial Restoration with Reduced lems of noise removal and border detection in digital Boundary Error, images. The proposed methods are compared with Nirmal Bose, Jaehoon Koo some classical or recent methods. A very important (The Pennsylvania State University) feature of the proposed methods is their intrinsic par- Since any symmetric BTHTHB matrix can be di- allelism, since they are implemented on well-known agonalized by the DCT matrix, a BTTB matrix parallel-working machines, as cellular automata are. encountered in the image reconstruction problem is sometimes approximated by a BTHTHB matrix for computational efficiency. In this paper, the error caused by the approximation of a BTTB matrix by a Room: 136, Session: WP4 BTHTHB matrix is analyzed. It is also shown that a simple modification of the observed image can achieve Chair: Biswa Nath Datta, Floyd B. Hanson both reduced boundary error and fast deblurring. Title: Large-Scale Computations in Control 16:30-16:50 Projection Methods for Reduced Order Modeling with Guaranteed Stability, 17:00-17:30 On Succesive Packing Approach to Thanos Antoulas (Rice University) Multidimensional (M-D) Interleaving, No Abstract Sankar Basu (IBM T. J. Watson Research Center), Xi Min Zhang, Yun Q. Shi (New Jersey Institute of Technology) 16:55-17:15 Computational Methods for Portfo- We propose an interleaving scheme for multidimen- lio and Consumption Policy Optimization in sional (M-D) interleaving. To achieved byusing a Log-Normal Diffusion, Log-Uniform Jump En- novel concept of basis interleaving array. A general vironments, method of obtaining a variety of basis interleaving ar- Floyd B. Hanson (University of Illinois at Chicago), rays is presented. Based on the basis interleaving ar- J. J. Westman (UCLA) ray, we then propose an interleaving technique, called Computational methods for a jump-diffusion portfo- successive packing, to generate the interleaved array lio and consumption policy optimization application of arbitrary size. It is shown that the proposed tech- using a log-uniform jump distribution with the usual k k nique can spread any error burst of m0 m1 within log-normal diffusion are considered. Jump-diffusion n n × m0 m1 array (1 k n 1) effectively so that the parameter estimations from a companion stochastic error× burst can be corrected≤ ≤ − with simple random error theory paper are applied for more model realism. correcting-code (provided the error correcting-code is The computational complexity of stochastic dynamic available). It is further shown that the technique is programming is reduced by the canonical Constant- optimal for combating a set of arbitrarily-shaped error Relative-Risk-Adverse utility model. bursts. Since this algorithm needs to be implemented only once for a given M-D array, the computational cost is is low. 17:20-17:40 Partial Eigenvalue Assignment in Linear Systems: Existence, Uniqueness and Numerical Solution, 17:30-18:00 Matrix Functions in Homomorphic Biswa N. Datta (Northern Illinois University), Signal Processing, Daniil R. Sarkissian (Mississippi State University) Eduard Krajnik (Czech Technical University) The problem of reassigning a part of the open-loop Yamada et al. suggested replacing the traditional cep- spectrum of a linear system by feedback control, leav- strum operator used in homomorphic signal process- ing the rest of the spectrum invariant, is called the ing by a finite-dimensional alternative called isomor- partial eigenvalue assignment problem. In this pa- phic operator. This paper sheds another light on the per, we derive new necessary and sufficient conditions isomorphic operator in terms of two matrix functions: for existence and uniqueness of solution of the par- exponential and logarithm. Closed form formulas for tial eigenvalue assignment problem and then present 1–D and 2–D cases are presented. a practical parametric algorithm to numerically solve it. The algorithm is feasible for large-scale solution and computationally viable. It also offers an oppor- 18:00-18:30 Cellular Automata in Image Pro- tunity to devise a robust solution to the problem by cessing, exploiting the arbitrary nature of the parameters. MTNS – 2002 71

17:45-18:05 Model Reduction via an Explicitly be written as a sum of squares of rational functions, Restarted Lanczos Algorithm, with some control over the form of the denominators. Vasilios Papakos, Imad M. Jaimoukha By using methods of operator theory (subnormal tu- (Imperial College,London) ples, hereditary calculus, moment problems) we will The nonsymmetric Lanczos algorithm, which belongs describe two possible simplifications in Artin’s repre- to the class of Krylov subspace methods, is increas- sentation: a reduction from the multiplicative cone ingly being used for model reduction of large scale sys- generated by the defining polynomials of the support tems in state space form, to exploit the sparse struc- set to the additive cone, and a reduction from real ture and reduce the computational burden. However, variables in an even dimensional space to complex a good approximation is, usually, achieved only with ones. An altenative, purely algebraic approach to the relatively high order reduced models. Moreover, the first simplification was recently obtained by T.Jacobi computational cost of the Lanczos algorithm is dom- and A.Prestel. inated by the full rebiorthogonalization procedure, which is necessary because the Lanczos vectors tend to lose their biorthogonality. A method based on lin- 17:50-18:10 Recent Progress in Polynomial Op- ear fractional transformations (LFTs) is proposed to timization, compute a reduced mth order model by applying k Ruchira Datta (UC Berkeley) small Lanczos algorithms with m/k steps each; thus No Abstract reducing the computational cost and storage require- ments. Applying this method, one can compute a tridiagonal similar realization of f(s) and when com- 18:10-18:30 Bounding Linear PDEs via Semidef- bined with conventional model reduction techniques, inite Optimization, a minimal or reduced realization. Constantine Caramanis, Dimitris Bertsimas (MIT) We apply some recent results of algebraic geometry, to show how the underlying geometry of an optimiza- Room: 208, Session: WP5 tion problem may be incorporated in a natural way, Chair: J. William Helton in a semidefinite optimization formulation. We dis- Title: Expressing Polynomials as Sums of cuss two examples; the first, an application to partial Squares Together with Applications differential equations, and the second, an application to deriving optimal inequalities in probability theory. 16:30-17:10 How to Write a Polynomial as a Sum Both illustrate the strength of the geometry induced of Squares of Polynomials, and Why You’d constraints. Want to Do So, Bruce Reznick (University of Illinois) A fundamental question about a real polynomial is whether it takes only non-negative values. If the poly- Room: 209, Session: WP6 nomial is a sum of squares of polynomials, then the Chair: Viswanath Ramakrishna answer to this question is ”yes”. Recent advances in Title: Quantum Engineering II implementing an old algorithm via semidefinite pro- gramming allow one to answer the ”sum of squares” 16:30-17:10 Optimal Control of Laser Cooling: A question rapidly. (This is connected to Hilbert’s 17th Theory of Purity Increasing Transformations, Problem.) The talk will be expository, light on ab- David Tannor, Shlomo Sklarz straction and filled with concrete examples. ( The Weizmann Institute) The powerful techniques of Optimal Control Theory (OCT), used in recent years for a variety of quan- 17:10-17:30 Applications of Our Newfound Fa- tum mechanical problems are applied to the problem cility in Expressing Polynomials as Sums of of laser cooling in molecules. This is a striking new Squares., mechanism in which the system does not lose energy Pablo A. Parrilo (ETH) at intermediate times, but grows more coherent; only No Abstract at the end of the process is the coherent or pure state manipulated into the lowest energy state. The key components of the theory – the definition of cooling as 17:30-17:50 Reduced Representations of Positive purity increase; the invariance of the purity to control Polynomials, fields; and a closed form for the optimal cooling tra- Mihai Putinar (UC Santa Barbara) jectory – correspond to the zeroth, second and third It is known from Emil Artin’s theory of ordered fields laws of thermodynamics, filling a longstanding need that a positive polynomial on a semi-algebraic set can for thermodynamic formulation of laser cooling. MTNS – 2002 72

17:10-17:30 Controllability of Pairs of Coupled torization explicitly. The results can be applied to the Quantum Dots, bang bang control with minimum number of switches Viswanath Ramakrishna of two level quantum systems. We also discuss the (University of Texas at Dallas) general problem of uniform finite generation of com- This article studies the controllability of a pair of cou- pact Lie groups. pled quantum dots being interrogated by an external electromagnetic field. It is shown that this system is controllable. However, in the limit of large spa- tial separation between the two dots the dynamical Lie algebra of the problem degenerates to the com- plex representation of so(4) in su(4). The question of which pure and mixed states for the system are acces- Thursday August 15, 2002 sible from the initial condition is then studied via an ab initio approach. This is achieved by finding an ex- 8:00-9:00 Room: 101 Plenary Talk plicit conjugation, within su(4), between the complex representation of so(4) and su(2) su(2), and then us- Eduardo Sontag (Rutgers University) ing the greater structure of the group⊗ SU(2) SU(2). On Systems Molecular Biology and Control ⊗ Theory We discuss some of the central topics in ”systems” 17:30-17:50 Constructive Control of Quantum (”post-genomic”) molecular biology, and how some of Systems, the basic concepts and ideas of control theory might Sonia Schirmer, A.D. Greemtree help in understanding basic issues such as feedback, (Open University, Milton Keynes) stability, systems identification, and switching. While The problem of explicit generation of unitary opera- problems in this new and exciting field ”sound” simi- tors for atomic systems with degenerate energy levels lar to standard problems in control theory, the techni- is considered. The Lie algebra structure is used to de- cal details are often different enough so as to present rive control schemes for the creation of arbitary super- totally new theoretical challenges. position states and selective population interchanges for a transition between two three-fold degenerate en- ergy levels.

17:50-18:10 Use of Wei-Norman Formulae and 9:00-10:00 Room: 126 Invited Talk Parameter Differentiation in Quantum Con- trol, Olof Staffans (Abo Akademi University), Claudio Altafini (SISSA-ISAS) Passive and Conservative Infinite-Dimensional For the unitary operator, solution of the Schrodinger Impedance and Scattering Systems (from a equation corresponding to a time-varying Hamilto- Personal Point of View) nian, the relation between the Magnus and the prod- A system is conservative if both the system itself and uct of exponentials expansions can be expressed in its dual system preservs energy. It is passive if it does terms of a system of first order differential equations not contain any internal energy sources. The words in the parameters of the two expansions, often referred scattering and impedance in the title describe how to as Wei-Norman formula. It is shown how to use the system interacts with its surroundings, i.e, it de- Wei-Norman formulae for the purposes of quantum fines the meaning of the word ”energy”. We present computing. the basic theory for discrete and continuous time lin- ear infinite-dimensional systems which are impedance or scattering passive or conservative. We also de- 18:10-18:30 Control of Quantum Mechanical scribe how to go from continuous time to discrete time Systems with Minimum Number of Switches, and back (using the Cayley transform), and from an Domenico D’ Alessandro (Iowa State University) impedance to a scattering setting (using a modified Given two linearly independent skew symmetric ma- feedback transform). We point out that every con- trices every rotation matrix can be written as the tractive analytic function in the right half-plane has product of alternate elements from the corresponding a continuous time scattering conservative realization, one dimensional subgroups. In this talk, we evaluate and that most positive (real) functions in the right the minimum number of factors required for the fac- half-plane have a continuous time impedance conser- torization of a given rotation matrix and provide an vative realization. We finally say a few words about algorithm to determine the factors in the optimal fac- lossless systems (i.e., systems which do not contain internal energy sources, sinks, or traps). MTNS – 2002 73

9:00-10:00 Room: 129 Invited Talk directly by the neighboring cells. Recent progress Wolfgang J. Runggaldier (LADSEB-CNR Padova) in genetics, recombinant DNA technology and single On Stochastic Control in Finance molecule imaging has resulted in elucidation of the In this talk we shall review various financial applica- structure of signaling pathways and regulatory mech- tions of stochastic control as well as appropriate solu- anisms used in gene expression. Biochemically based tion methodologies. These methodologies will further- mathematical and computational models are becom- more be discussed in comparison with one another. ing a common useful tool in understanding the sys- temic control characteristics of the coupled signaling and gene regulation networks. However, no attempt has been made so far to develop a description suitable for using the concepts of Shannon’s information the- 9:00-10:00 Room: 136 Invited Talk ory and to utilize this description of signaling events Matthias Heinkenschloss (Rice University) to follow the information flow. Here I will propose Domain Decomposition Approaches for the a way, in which a representation of information flow Optimization of Distributed Systems can be established and used in two well-defined biolog- Optimization of distributed systems in the context of ical settings involving regulation of gene expression in parameter estimation, optimal control, or optimal de- stress response and development. Implications of this sign plays an important role in science and engineer- analysis for a conceptually different understanding of ing. These optimization problems pose several compu- biological regulation will be discussed. tational difficulties arising, among other things, from the large number of variables and the conditioning of subproblems. Domain decomposition (DD) can be ap- plied at different levels (formulation of the distributed optimization problem, solution of optimization sub- Room: 126, Session: THA2 problems, and computation of PDE solutions) to de- Chair: Ruth Curtain, Olof Staffans sign efficient optimization algorithms for distributed Title: Distributed Parameter Systems: Theory systems. In this talk we study several DD approaches, Part I discuss relations among them, and present applica- 10:30-10:50 Some Results on the Theory of Lin- tions to selected problems. ear Time-Invariant Dissipative Systems with Hilbert and Pontryagin State Spaces, Damir Arov (South Ukranian Pedagogical University) Morning: New results on the similiraty of dissipative linear time- invariant systems and main operators of the systems Room: 102, Session: THA1 with Hilbert and Pontryagin state spaces will pre- Chair: Mark Alber sented.They retate to the systems with transfer func- Title: Complex Networks and Biological Appli- tions of generalized Schur, Caratheodory and Potafov cations 1 classes 10:30-11:10 The Spread of Infections on Social Networks, Mark Newman (University of Michigan) 10:55-11:15 Explicit Formulae for J-Spectral One of the primary motivations for the study of so- Factors for Well-Posed Systems, cial networks has been the role they play in the spread Ruth Curtain (University of Groningen), of infection. In this talk I will describe some recent Amol J. Sasane (University of Twente) work on the modeling of epidemic disease on net- The standard way to obtain explicit formulas for spec- works. Among other things, I will discuss what we tral factorization problems for rational transfer func- know about the structure of actual contact networks tions is to use a minimal realization and then obtain and describe some exact solutions we have found of formulae in terms of the generators A, B, C and D. epidemic models on networks of various kinds. It is only possible to extend this approach to spe- cial classes of infinite-dimensional systems. For the class of well-posed linear systems for which zero is in 11:10-11:50 Information Theory Aspects of Sig- the resolvent set of A we suggest another approach. nal Transduction and Gene Regulation, There are nice connections between well-posed linear Andrea Levchenko (Johns Hopkins University) systems and their reciprocal systems which allow one Intracellular signal transduction is the process of to translate a factorization problem for the well-posed transmission and processing of information supplied linear system into one for its reciprocal system, the by the state of the medium surrounding a cell or latter having bounded generating operators. We illus- MTNS – 2002 74 trate this general approach by giving explicit solutions Room: 129, Session: THA3 to the sub-optimal Nehari problem. Chair: J.M. (Hans) Schumacher Title: Systems and Control Theory in Finance and Insurance 1 11:20-11:40 A Riccati Equation Approach to 10:30-11:30 Control and Financial Engineering, the Standard Infinite-Dimensional H-Infinity J. M. (Hans) Schumacher (Tilburg University) Problem, The paper provides a review of the basics of financial Kalle M. Mikkola (Helsinki University of Technology), engineering, with a few examples. We emphasize con- Olof Staffans (Abo Akademi University) nections with control theory in a broad sense rather We study the standard (four-block) H-infinity prob- than with stochastic control theory in particular, and lem in the infinite-dimensional setting of regular well- the reader is not assumed to be versed in stochastic posed linear systems, and show that the standard processes. Two brief case studies are presented: the characterizations of the solvability of this problem in construction of an indexed bond, and the hedging of terms of either two independent Riccati equations and long-term contracts for delivery of oil. a spectral radius condition or two nested Riccati equa- tions are valid. We are also able to parameterize the set of all suboptimal solutions in the standard way. 11:30-12:00 Dynamic Risk Sensitive Asset Man- The exact formulation varies depending on the regu- agement With Nonnegative Multiple Factor larity assumptions that we make. Pure delays in the Constraints, impulse response are permitted in the most general Arunabha Bagchi (University of Twente), version, but these lead to certain correction terms in K. Suresh Kumar (University of Pennsylvania) the Riccati equations. We study the dynamic asset allocation problem with nonnegativity constraints on the economic factors. We first convert the dynamic asset allocation problem 11:45-12:05 Sub-optimal Hankel Norm Approx- into an equivalent stochastic differential game. We imation for the Wiener Class, then impose nonnegativity constraints on the game Orest Iftime, Amol Sasane problem. We then solve this new problem using some (University of Twente) recent results on such constrained differential games. For the Wiener class of matrix-valued functions we ob- For multifactor constraints, existence of solution of tain a simple frequency domain solution for the sub- the resulting HJB-equation is analysed in detail. optimal Hankel norm approximation problem. The approach is via J-spectral factorization. 12:00-12:30 A Filtered No-arbitrage Model for Term Structures from Noisy Data, 12:10-12:30 LQG Balancing in Infinite Dimen- Andrea Gombani (LADSEB-CNR), sions, Stefan R. Jaschke (Weierstrass-Institut, Berlin), Mark R. Opmeer, Ruth Curtain Wolfgang J. Runggaldier (University of Padova) (University of Groningen) We consider the problem of pricing in financial mar- Balanced realizations of a linear system are those for kets when agents do not have access to full informa- which the observability and controllability gramians tion. The particular problem concerns the pricing of are both equal and diagonal. Truncated balanced non traded or illiquid bonds on the basis of the obser- realizations have good approximation properties and vations of the yields of traded zero-coupon bonds. The are often used for controller design. Of course, bal- approach being used gives an example of how stochas- anced realizations only exist for stable transfer func- tic filtering techniques, in particular the Kalman filter, tions and it has been proposed that LQG-balanced can be usefully applied to pricing under incomplete realizations could prove a useful alternative. LQG- information. balanced realizations of a system are those for which the solutions of the associated control and filter Ric- cati equations are equal. While there exists a com- plete theory of balanced realizations (and truncations ) for infinite-dimensional systems, the theory for the Room: 136, Session: THA4 LQG versions has yet to be developed. In this pa- Chair: David Nicholls per we solve this problem for discrete-time infinite- Title: Fully Nonlinear, Three-Dimensional, dimensional systems. Surface Water Waves in Arbitrary Depth 10:30-11:00 Experiments on Deep-Water Waves with Two-Dimensional Surface Patterns, Diane Henderson (Penn State) MTNS – 2002 75

Experiments are conducted in a three-dimensional Room: 208, Session: THA5 wave basin with a wavemaker system comprising 32 side-by-side paddles for which there is precise control. Two types of wavemaker forcings are used to create two-dimensional surface patterns: (1) two symmetric carrier waves interacting at an oblique angle and (2) Chair: Krzystof Galkowski, Eric Rogers, Victor Vin- a single carrier wave propagating in the x-direction nikov with a Jacobi elliptic, sn-function modulation in the Title: Multidimensional Systems 5 y-direction. Data are presented from overhead pho- tographs and from time series obtained by traversing 10:30-11:00 Robust Stability and Stabilization of a wave-gage through the patterns. Two parameters n-D Systems, are systematically varied: the horizontal aspect ratio Jiang-Qian Ying (Gifu University, Japan), of the cells comprising the surface pattern and the Li Xu (Akita Prefectural University, Japan), measure of nonlinearity of the input wavefield. Masayuki Kawamata (Tohoku University, Japan) The stability margin and stabilizability margin of n-D systems are first considered. The stabilizability mar- 11:00-11:30 Instability of Bounded Solutions of gin is defined for the first time to be the largest sta- the 2-D Cubic Nonlinear Schrodinger Equa- bility margin that a closed-loop feedback system can tion, reach. A general computational algebraic procedure is John Carter (Seattle University) then proposed for both computation of the stability/ In 1974, Zakharov and Rubenchik established that the stabilizability margins and construction of a stabiliz- soliton solutions of the two-dimensional cubic nonlin- ing compensator which can reach the stabilizability ear Schrodinger equation (NLSE) are unstable with margin as close as desirable. respect to long-wave transverse perturbations. We es- tablish that all bounded solutions of the NLSE are unstable with respect to long-wave transverse pertur- 11:00-11:30 Succesive stabilization of a class of bations. All possible sign choices in the NLSE are 2D systems, considered, though we focus on the NLSE as a model Krzysztof Galkowski, Bartek Sulikowski of waves on deep water. (University of Zielona Gora), Eric Rogers (University of Southampton), David H. Owens (University of Sheffield) 11:30-12:00 Computing (quasi) Periodic Waves The subject area of this paper is the application of sys- in Shallow Water, tems theory developed for linear repetitive processes, Bernard Deconinck (Colorado State University) a distinct class of 2D linear systems, to linear itera- The Kadomtsev-Petviashvili (KP) equation governs tive learning control schemes. A unique feature is the the evolution of waves in shallow water. It has a fam- inclusion of experimental results obtained from the ily of (quasi) periodic solutions which are implicitly application of control laws designed using this theory parametrized by compact, connected Riemann sur- to an experimental rig in the form of a chain conveyor faces. This presents a large obstacle for the use of system. these solutions. I will discuss black box implementa- tions aimed at making the computation of these solu- tions effective. 11:30-12:00 Optimal Control for a Class of Dif- ferential Linear Repetitive Processes, Eric Rogers (University of Southampton), 12:30-13:00 Mathematical Models of Deep- S. Dymkou (Ballarat University, Australia), Water Waves with two-Dimensional Surface M. Dymkov (National Academy of Sciences of Be- Patterns, larus, Minsk), Harvey Segur, (U. of Colorado) K. Galkowski (University of Zielona Gora), Another speaker in this session presents intrigu- D. H Owens (University of Sheffield) ing experiments on waves in deep water, with two- Differential repetitive processes are a class of dimensional surface patterns that are either periodic continuous-discrete 2D linear systems of both systems or nearly so. These wave patterns persist for relatively theoretic and applications interest. The feature which long times in the experiments. This talk presents makes them distinct from other classes of such sys- preliminary attempts to find a suitable mathematical tems is the fact that information propagation in one model for these observed waves. of the two independent directions only occurs over a finite interval. In this paper, we develop new results on optimal and sub-optimal control of an important sub-class of these processes. MTNS – 2002 76

12:00-12:30 Relation between Eigenvalues and 12:00-12:30 Enlarging the Stable Region of Im- Singular Values in the Problem of Stability age Based Control by Path Planning, Maintenance of Ellipsoidal Estimates, Youcef Mezouar (Columbia University) Taalaibek A. Akunov, Anatoly V. Ushakov No Abstract (Technical University, St. Petersburg, Russia) Problem of defining of relation between eigenvalues and singular values of matrix and matrix-valued func- tions of the matrix is considered. The problem arises Room: 210, Session: THA7 at the interfaces between methods of systems synthe- Chair: Bill Helton, Andre Ran, Leiba Rodman sis with use of generalized modal control and quality Title: Matrix and Operator Equations II evaluation of these systems with the help of ellipsoidal 10:30-11:00 Noncommutative Convexity of quality estimates. Solution is oriented on maintenance Functions and Sets, of stability of ellipsoidal quality estimates by means J. William Helton (University of California, San of stabilisation of state matrix eigenvalues and eigen- Diego) vectors. The type of convexity one sees in many linear sys- tems applications is ”matrix convexity”. Namely, one has a noncommutative rational function and requires it be ”convex” when matrices are plugged in. The talk concerns work with Scott McCullough charac- terizing such functions and also characterizing those Room: 209, Session: THA6 ”noncommutative regions” which can be represented Chair: Koichi Hashimoto with Linear functions (LMI). The tools include a new Title: Globally Stable Robust Visual Servoing noncommutative Positivestellensatz. 10:30-11:00 Keeping Features in the Camera’s Field of View: a Visual Servoing Strategy, 11:00-11:30 Symmetry Groups, Semidefinite K. Hashimoto ( University of Tokyo), Programming,and Sums of Squares, Graziano Chesi, D. Prattichizzo, A. Vicino Pablo A. Parrilo (ETH) (University of Siena) No Abstract A new visual servoing strategy ensuring global con- vergence and all points in the field of view without re- quiring any knowledge on the 3D model of the scene is 11:30-12:00 The Symmetric Linear Matrix Equa- proposed. Moreover, the camera trajectory length is tion, minimized in the rotational space and, for some cases, Martine C. B. Reurings also minimized in the translational space. Robustness (Vrije Universiteit, Amsterdam) against uncertainties on the camera intrinsic parame- In this talk the matrix equation X + A1∗XA1 + + ters and optical axis direction is also guaranteed. ··· Am∗ XAm = Q is considered, where Q is positive defi- nite and the Aj, j = 1, . . . , m are arbitrary matrices. For m = 1 the equation becomes X + A1∗XA1 = Q. 11:00-11:30 Binocular Visual Servoing with a Note the difference with the Stein equation. It is Limited Field of View, known that the equation for m = 1 has a unique pos- Noah Cowan (University of California at Berkeley) itive definite solution if A1∗QA1 < Q . In the general This paper presents a simple, globally conver- case a similar condition is sufficient for the existence of gent, binocular image-based visual servoing algorithm a unique positive definite solution, but we cannot use which accounts for the limited field-of-view of a CCD the same techniques as were used in the case m = 1. camera system. Recourse to Navigation Functions, a Our proof depends on the notion of the Kronecker refined notion of artificial potential functions, leads product of two matrices. to controllers for first order, quasi-static as well as second order, dynamic plants. Underlying these con- structions is a global account of the transformation 12:00-12:30 Investigating Duality on Stability between the task and image space. Conditions, Mauricio de Oliveira (UNICAMP) This paper is devoted to investigate the role played 11:30-12:00 Visual Servoing with Dynamics: by duality in stability analysis of linear time-invariant Control of an Unmanned Blimp, systems. We seek for a dual statement of a re- Jim Ostrowski (University of Pennsylvania) cently developed method for generating stability con- No Abstract ditions, which combines Lyapunov stability theory MTNS – 2002 77 with Finsler’s Lemma. This method, developed in networks, which are characterized by diverging con- the time domain, is able to generate a set of (primal) nectivity fluctuations, exhibit the lack of an epidemic equivalent stability tests involving extra multipliers. threshold and always show a finite fraction of infected The resulting tests have very attractive properties. individuals. This particular weakness is found also in Stability is characterized via Linear Matrix Inequali- immunization strategies that can be successfully de- ties and we use optimization theory to obtain the du- veloped only by taking into account the connectivity als. The dual problems are given a frequency domain patterns of scale-free networks. The understanding interpretation. of epidemics in scale-free networks might deliver new insights in the spread of information and diseases in biological and technological networks that often ap- pear to be characterized by complex heterogeneous architectures. Middle: Room: 102, Session: THM1 Chair: Mark Alber Title: Complex Networks and Biological Appli- Room: 126, Session: THM2 cations 2 Chair: Ruth Curtain, Olof Staffans 14:00-14:40 Synchronization of Oscillators in Title: Distributed Parameter Systems: Theory Small World Systems, Part II Lou Pecora (Naval Research Laboratory) 14:00-14:30 Zeros of SISO Infinite-Dimensional Smallworld networks are regular lattices of connected Systems, nodes to which random connections have been added. Kirsten Morris (University of Waterloo), Recently, Watts and Strogatz showed that only a few Richard Rebarber (University of Nebraska) random, long-range connections are necessary to cre- We define the zeros of SISO infinite-dimensional sys- ate paths that are very short even between diametri- tems with bounded control and observation operators cally opposed nodes. An interesting question regard- B and C respectively in terms of the spectrum of an ing smallworld systems is, how do such added, ran- operator on an invariant subspace. If the range(B) is dom connections affect dynamics on the lattice if the not in ker(C), we calculate the operator K such that nodes are oscillators? In particular, is the intuitive the spectrum of A+BK on ker(C) is the system ze- view that additional connections, even random ones, ros. Also, A+BK generates a semigroup on ker(C). If would help stabilize a synchronous state correct? We the range(B) is in ker(C), a number of situations may show that we can use the recently formulated mas- occur, depending on the nature of B and C. ter stability approach to synchronization stability to probe synchronization in smallworld systems in a di- rect and generic way. The results are often not what 14:30-15:00 Stabilizability of Systems with Sig- one expects. In some cases added connections can nals in `2(Z), both help and hurt synchronization. The opposite Birgit Jacob (University of Dortmund) case of the saturated world is equally interesting and We consider a system as a (possibly unbounded) lin- argues for a symmetry-breaking approach to study ef! ear operator from `2(Z) to `2(Z). In this setup even a fects at both ends of the connection distribution. We simple causal convolution system is not stabilizable is also compare the results for smallworld, random net- the usual sense of the term. We discuss this problem, works to highly regular networks/graphs. The small- and we “solve” the problem by adapting the definition world networks do as well as the best regular networks. of stabilizability. Finally, we compare the obtained controller with the one we obtain by restricting our system to `2(N0). 14:40-15:20 Intracellular Signaling is Dependent on the Cytoskeleton., Gabor Forgas (University of Missouri Columbia) 15:00-15:30 Stability and Boundedness of Con- No Abstract tinuous and Discrete-Time Systems, Hans Zwart, B.Z. Guo (University of Twente) 15:20-16:00 The Role of Scale-free Connectivity We investigate the relation between discrete and con- Patterns in Spreading Phenomena, tinuous systems on a Hilbert space. More precisely, we Alessandro Vespignani (ICTP) investigate the stability properties of the semigroup n We show that the large connectivity fluctuations usu- generated by the operator A, and the sequence Ad , ally found in scale-free networks strengthen consider- where n is a positive integer, and A and Ad are re- ably the incidence of epidemic outbreaks. Scale-free lated via Ad = (I + A)(I A)−1, or equivalently − MTNS – 2002 78

A = (Ad I)(Ad +I)−1. One of our result states that Mediods (PAM) clustering algorithm of Kaufman and if A and− its inverse generate a bounded semigroup, Rousseeuw(1990) is used to restrict the initial set of n then Ad is uniformly bounded. stocks. We find that PAM is effective in its ability to specify nonuniform stock series from the entire uni- verse. We are pleasantly surprised that the algorithm 15:30-16:00 Coprimeness Conditions for Pseudo- eliminated the bankrupt Enron and Federal Mogul rational Transfer Functions, stock series, without our intervention. We use outlier Yutaka Yamamoto (Kyoto University) analysis to define two separate active trading strate- This paper surveys and discusses some coprimeness is- gies. The outliers within a time series are determined sues for the class of pseudorational transfer functions. by the use of a Kalman Filter/Smoother model devel- They are effective in capturing various system prop- oped by de Jong and Penzer(1998). Weekly trading in erties for systems with bounded-time memory, e.g., stocks with an initial $30,000 with a closed stock port- delay systems. Various notions of coprimeness are folio from 1993 to 2001, we obtained a 17.8% annual known to be non-equivalent. We show, however, that return on a cash surrogate passive strategy, 18.1% on a under some conditions spectral coprimeness implies passive strategy using all the stocks in our restricted exact coprimeness (existence of Bezout identity). asset universe, 20.2% on a combined cash protected and outlier active strategy, and 23.3% using the out- lier active strategy only. Comparing these results to the passive strategy being entirely invested in the S&P Room: 129, Session: THM3 500 Large Cap index with at 9.9% return, we find that Chair: J.M. (Hans) Schumacher under this stock portfolio any of our strategies are su- Title: Systems and Control Theory in Finance perior to that of a purely passive index strategy. and Insurance 2 14:00-15:00 Ruin Probabilities Minimization and Dividend Distribution Optimization in Diffu- sion Models, Michael Taksar Room: 136, Session: THM4 This will serve as an introduction to the stochastic control models in insurance. We will consider a model Chair: Paul Van Dooren of an insurance company which has different modes of Title: Robust Control and Linear Matrix In- risk and financial control. Different types of reinsur- equalities ance correspond are the risk reduction techniques of 14:00-14:30 Linear Matrix Inequalities in Robust the insurance, while financial control is a more famil- Control: A Brief Survey, iar portfolio rebalancing. There are different objec- Venkataramanan Balakrishnan (Purdue University) tives which the company pursues. One is the classi- Control system models must often explicitly incorpo- cal minimization of ruin probabilities. Another one rate in them uncertainties or perturbations. Robust is dividend pay-out maximization. The latter merges control deals with the analysis of and design for such with the classical finance issue of a small investor pi- uncertain control system models. This paper provides oneered by Merton. Diffusion approximation enables a brief survey to some robust control techniques that one to get a closed form solution to many problems are based on numerical convex optimization over Lin- and see early the structure of the optimal policy. ear Matrix Inequalities (LMIs).

15:00-15:30 Continuous-Time Mean-Variance 14:30-15:00 Periodic Multirate Systems, nu-Gap Portfolio Selection with Markov-Modulated and Robust Stabilization, Market Parameters, Li Qiu, Li Chai Xun Yu Zhou (Chinese University Hong Kong (Hong Kong University of Science & Technology) No Abstract A new discription of multirate systems, called mul- tirate periodic system, is given using the concept of periodic time-varying input-output spaces. We then 15:30-16:00 Stock Selection Based on Cluster define the ν-gap metric of two multirate periodic sys- and Outlier Analysis, tems and study the robust stabilization with this met- Steven Craighead, Bruce Klemesrud ric. The optimal robust stabilization margin is explic- (Nationwide Financial, Columbus, Ohio) itly computed and an obsever-form suboptimal con- In this paper, we study the selection and active trad- troller is given. The solution amounts to solving two ing of stocks by the use of a clustering algorithm and discrete-time algebraic Riccati equations and an ex- time series outlier analysis. The Partitioning Among tended Parrot problem. MTNS – 2002 79

15:00-15:30 Spectral Factorization and Sums of neighborhood of the boundary. Squares via Semidefinite Programming, Hugo Woerdeman (College of William and Mary) It is known that a matrix (trigonometric) polynomial 15:00-15:30 Regular Implementability nD Be- that is nonnegative definite on the real axis (on the haviors, unit circle) allows an outer factorization. In this pa- Paula Rocha (University of Aveiro, Portugal) per we investigate the different possibilities how the We consider the problem of implementability and reg- outer factor may be determined via semidefinite pro- ular implementability of a nD behavior by intercon- gramming. In the dual formulation this leads to a nection through a control variable. In the 1D case, question what possible barrier functions there exist for the second property is equivalent to the regular im- classes of positive semidefinite structured block ma- plementability of the desired system variable behavior trices. The multivariable version, i.e., finding sums of by full interconnection. A similar result is not valid in squares representations, is also explored. Appropriate the nD case. However we obtain an alternative char- definitions for outer in this context also come to pass. acterization in terms of the regular implementability of a suitable control variable behavior.

15:30-16:00 Robustness Analysis via Stability Radii, Spectral Value Sets and mu-Functions, 15:30-16:00 Cones of Trajectories as Subsets of Michael Karow (Technical University of Berlin) Linear Systems: the Autonomous Case, This is an introductory talk on robustness problems Andrea Morettin (Universit`adi Padova) concerning the stability of uncertain finite dimen- In most pratical applications the designer has to face sional linear systems. Our tools are stability radii, physical and technological constraints that lead to a spectral value sets (pseudospectra) and µ-function for set of inequalities involving the variables of the model. structured matrix perturbations. We discuss the re- In order to provide a formal framework for dealing lationship between these quantities and give explicit with this situation, we introduce the concept of coni- formulas for several perturbation structures. In par- cal behaviour as a time-invariant cone of real trajec- ticular, we regard real perturbations and pseudospec- tories which is complete. We describe such a set in tra of composite systems. relation with the minimal vector space that contains it. In particular, we afford this problem by assuming that the conical behaviour is autonomous, i.e. a tra- jectory note in finite time window is totally defined. Room: 208, Session: THM5 If a conical behaviour is polyhedral it admits a non- Chair: Maria Elena Valcher negative state representation. In a dual look a conical Title: The Behavioral Approach to Dynamic behaviour may be the set of trajectories which satis- Systems fies a model of a finite number of inequality constraints that act on a finite time window of the trajectory it- 14:00-14:30 Deterministic Kalman Filtering, self. Jan C. Willems (University of Leuven, Belgium) A deterministic interpretation of the Kalman filter- ing formulas is given, using the principle of least squares estimation. The observed signal and the to- Room: 209, Session: THM6 be-estimated signal are modeled as being generated as outputs of a finite-dimensional linear system driven by Chair: Naomi Leonard an input disturbance. Postulating that the observed Title: Control and Dynamics of Mechanical Sys- signal is generated by the input disturbance that has tems I minimal least squares norm leads to a method of com- 14:00-14:20 Composition of Dirac Structures puting an estimate of the to-be-estimated output. and Control of Port-Hamiltonian Systems, Arjan van der Schaft (University of Twente), J. Cervera (Universidad de Murcia, Spain) 14:30-15:00 Over-Determined Systems, Key feature of Dirac structures (as opposed to Poisson Eva Zerz (University of Kaiserslautern) or symplectic structures) is the fact that the compo- Over-determined systems are a special type of au- sition of Dirac structures again defines a Dirac struc- tonomous systems. They are algebraically character- ture. This implies that any power-conserving inter- ized by kernel representation matrices that are minor connection of port-Hamiltonian systems is a port- right prime. The systems theoretic significance of the Hamiltonian system itself. Furthermore, the com- notion is closely related to the extension problem, an posed Dirac structure determines the closed-loop al- over-determined boundary value problem in which the gebraic constraints, as well as its Casimir functions. values of the desired function are prescribed in a whole In this paper the set of achievable composed Dirac MTNS – 2002 80 structures for a given plant port-Hamiltonian system 15:40-16:00 Extremal Flows on Stiefel Mani- is characterized. folds, and Riemannian Potatoes, Peter Crouch (ASU), Anthony M. Bloch (University of Michigan) 14:20-14:40 Hamiltonian Attitude Dynamics for In this paper, we consider an optimal control prob- a Spacecraft with a Point Mass Oscillator, lem, or constrained variational problem, that evolves Craig Woolsey (Virginia Tech) on a general Stiefel Manifold, posed in both continu- Two noncanonical Hamiltonian models are presented ous and discrete time. At one extreme this problem for the dynamics of a rigid body with an constrained yields a symmetric realization of the generalized rigid moving point mass. One of these models is used to body system, and at the other extreme reduces to the analyze stability of steady principal axis rotation of a geodesic flow on an ellipsoid. We work out the Hamil- rigid body with a spring-mass oscillator. The analysis tonian structure of the general extremal flow using the gives necessary and sufficient conditions for stability optimal control formalism revealing its complexity. of steady major axis rotation, as well as sufficient con- ditions for instability of intermediate and minor axis rotation.

14:40-15:00 Controllable Kinematic Reductions Room: 210, Session: THM7 for Mechanical Systems: Concepts, Computa- Chair: Jan van Schuppen tional Tools, and Examples, Title: Control and Algebra Andrew Lewis (Queen’s University), Francesco Bullo (University of Illinois at Urbana- 14:00-14:30 Control and Algebra - An Introduc- Champaign), tion, Kevin M. Lynch (Northwestern University) Jan H. van Schuppen (CWI, Amsterdam) This paper introduces the novel notion of kinematic The concept of bisimulation has been developed in reductions for mechanical systems and studies their computer science by R. Milner for the study of equiv- controllability properties. We focus on the class of alences of transition systems. It has been generalized simple mechanical control systems with constraints to coalgebra and to category theory. In control and and model them as affine connection control systems. system theory the concept of bisimulation is useful for We present a comprehensive treatment of local con- the study of relations of dynamic systems, in particu- trollability properties of mechanical systems and their lar of observationally equivalent systems. This short kinematic reductions. paper is an introduction to the session ‘Control and Algebra’.

15:00-15:20 Matching and Stabilization of Linear Mechanical Systems, 14:30-15:00 Towards an Algebraic Systems The- Dimitri Zenkov (North Carolina State University) ory of Hybrid Systems, We consider linear controlled mechanical systems George J. Pappas (University of Pennsylvania) and show that controllability enables one to use the The fundamental notion of bisimulation has inspired method of controlled Lagrangians for feedback stabi- various notions of system equivalences in concurrency lization of equilibria. theory. Many notions of bisimulation for various dis- crete systems have been recently unified in the ab- stract category theoretical formulation of bisimula- 15:20-15:40 Matching and Stabilization of Con- tion due to Joyal, Nielsen and Winskel. In this pa- strained Systems, per, we adopt their framework and unify the notions Guido Blankenstein (EPFL) of bisimulation equivalences for discrete, continuous In this paper we discuss the stabilization by means dynamical and control systems. This shows that our of structure preserving feedback laws (i.e., match- equivalence notion is on the right track, but also con- ing) of constrained systems described as implicit port- firms that abstract bisimulation is general enough to controlled Hamiltonian systems. The theory is ap- capture equivalence notions in the domain of contin- plied to underactuated mechanical systems with kine- uous systems. We believe that the unification of the matic constraints. bisimulation relation for labelled transition systems and dynamical systems under the umbrella of abstract bisimulation, as achieved in this work, is a first step towards a unified approach to modeling of and rea- soning about the dynamics of discrete and continuous structures in computer science and control theory. MTNS – 2002 81

15:00-15:30 The Category of a Affine Connection better identification of the initial conditions that lead Control Systems, to certain steady states. I propose this type of analysis Andrew Lewis (Queen’s University) as a first, qualitative step in understanding complex The category of affine connection control systems is networks. The success of a Boolean representation of one whose objects are control affine systems whose a network strongly suggests that the topology of the drift vector field is the geodesic spray of an affine con- network is correctly taken into account, and a more nection, and whose control vector fields are vertical quantitative approach can be used. lifts to the tangent bundle of vector fields on configu- ration space. This class of system includes a large and important collection of mechanical systems. The mor- 17:00-17:30 Modeling Mesenchymal Condensa- phisms (feedback transformations) in this category are tion during Limb Chondrogenesis, investigated. Gilberto Tomas (University of Notre Dame) We study a quasi-2D micromass cell culture of chick limb-bud cells which undergoes chondrogenesis that 15:30-16:00 Coalgebra and Supervisory Control qualitatively resembles chondrogenesis in vivo. Dis- with Partial Observations, associated mesenchymal cells, after plating at high Jan Komenda (CWI, Amsterdam) density, form local condensations, which then differ- Coalgebraic techniques are applied to the supervisory entiate into cartilage. A similar process in the intact control of discrete-event systems with partial obser- embryonic limb leads to the formation of skeletal pri- vations. Classical notions from concurrency theory mordia. Among several possible underlying mecha- are specialized to control theory.The concept of weak nisms of mesenchymal condensation, we concentrate transitions enables the relational characterization of on local cell-cell adhesion and cell-extracellular ma- observability and gives rise to a coalgebraic formula- trix interaction and propose a stochastic pattern for- tion of the necessary and sufficient conditions for the mation mechanism for cell clustering. We implement existence of a supervisory control which achieves a a 2D extended Potts Model simulation and we discuss considered legal language. how this mechanism could be an important part of mesenchymal condensation in limb chondrogenesis in vitro and in vivo.

Afternoon: 17:30-18:00 Classification of scale-free networks, Room: 102, Session: THP1 Byungnam Kahng (Seoul National University) Chair: Mark Alber While the emergence of a power law degree distribu- Title: Complex Networks and Biological Appli- tion in complex networks is intriguing, the degree ex- cations 3 ponent is not universal. Here we show that the be- tweenness centrality displays a power law distribu- 16:30-17:00 Connections Matter: A Boolean tion with an exponent η which is robust and use it Model for the Segment Polarity Network of to classify the scale-free networks. We have observed Drosophila Melanogaster, two universality classes with η 2.2 and 2.0, respec- Reka Albert (University of Minnesota) tively. Real world networks for the≈ former are the pro- The Drosophila segment polarity genes form a tein interaction networks, the metabolic networks for complex network of cross-regulatory interactions. eukaryotes and bacteria, and the co-authorship net- Through the functioning of this network the initial work, and those for the latter one are the Internet, expression of the segment polarity genes is stabilized the world-wide web, and the metabolic networks for and maintained throughout several stages of embry- archaea. Distinct features of the mass-distance rela- onic development including two rounds of cell divi- tion, generic topology of geodesics and resilience un- sions. In this talk I propose a Boolean representation der attack of the two classes are identified. Various of this network that assumes that genes and proteins model networks also belong to either of the two. are either ON or OFF, and their interactions can be formulated as logical functions. This model is deeply rooted in the topology of the segment polarity net- 18:00-18:30 Prediction of Protein Essentiality work, while it disregards the biochemical details of Based on Genomic Data, the interactions. The model is able to reproduce the Hawoong Jeoong (Korea Advanced Inst. of Science wild type expression pattern of the segment polarity and Technology Taejon), genes, as well as the ectopic expressions obtained for Zoltan N. Oltvai (Northwestern University Medical gene mutation experiments. In addition, the Boolean School) representation allows for a more complete analysis of Albert-Laszlo Barabasi (University of Notre Dame) the possible steady states as experiments, and for a MTNS – 2002 82

A major goal of pharmaceutical bioinformatics is to 17:00-17:30 H-infinity Control of Acoustic Noise develop computational tools for systematic in silico in a Duct with a Feedforward Configuration, molecular target identification. Here we demonstrate Kirsten Morris (University of Waterloo) that in the yeast, Sacharomyces cerevisiae, the pheno- A mathematical model of sound propagation in a duct typic effect of single gene deletions simultaneously cor- derived from physical principles is described. Exper- relate with fluctuations in mRNA expression profiles, imental results validate the model. Theoretical lim- the functional categorization of the gene products, its of noise reduction are examined using this model. and their connectivity in the yeast’s protein-protein In many feedforward configurations, the optimal con- interaction network. Building on these quantitative troller can reduce the noise at a point to almost zero. correlations we developed a computational method for However, no noise reduction is obtained at points be- predicting the phenotypic effect of a given genes func- low the performance point, and at many such points tional disabling or removal. Our subsequent analyses the noise level is increased. were in good agreement with the results of system- atic gene deletion experiments, allowing us to predict the deletion phenotype of a number of untested yeast 17:30-18:00 Positivity and Dissipativity of Oscil- genes. The results underscore the utility of large ge- lating Diffusive Filters, Application to the Sta- nomic databases for in silico systematic drug target bility of Coupled Systems, identification in the postgenomic era. G. Dauphin, Denis Matignon (ENST/TSI & CNRS, URA 820) Oscillating diffusive filters such as Gegenbauer filters have a slowly decreasing impulse response. They are a continuous aggregation of positive oscillating filters; hence, they are positive and have a dissipative realiza- tion, which helps prove both the external and internal Room: 126, Session: THP2 stabilities of some coupled systems involving a ratio- Chair: Kirsten Morris, Olof Staffans nal filter and such an oscillating diffusive filter in the Title: Distributed Parameter Systems: Stabi- feedback loop. lization and Control, Part I 16:30-17:00 Reciprocals of Regular Linear Sys- 18:00-18:30 Can Positive Pseudo-Differential tems: a Survey., Operators of Diffusive Type Help Stabilize Ruth Curtain (University of Groningen) Unstable Systems?, This paper proposes a novel approach to studying reg- Denis Matignon (ENST/TSI & CNRS, URA 820) ular linear systems via their reciprocal systems. Un- Diffusive representations of positive pseudo- der the generic assumption that A has a bounded differential operators can often be used in the analysis inverse, a regular linear system possesses a recipro- of coupled systems, in which their dissipative realiza- cal system with four bounded generating operators. tion plays a major role. Now, some coupled systems Many system theoretic problems for regular linear sys- involving a negative PDO can still be stable. Con- tems can be translated into equivalent problems for versely, some unstable systems can be stabilized by their reciprocal system. Due to the bounded nature of positive PDOs, thus requiring some more analytical the generators, the problems for the reciprocal system knowledge: such striking examples will be presented, are easier to solve and these solutions can be trans- either in continuous time or in discrete time. lated back to solutions for the original regular linear system. Properties of reciprocal systems are reviewed and the success of this approach is illustrated with the LQ control problem and the existence of (pseudo-) co- prime factorizations. Another very successful applica- Room: 129, Session: THP3 tion of reciprocal systems is to obtain explicit formu- Chair: Tyrone Duncan las for solutions to spectral factorization problems in Title: Stochastic Theory and Applications terms of the original unbounded operators A,B,C,D 16:30-17:00 An Approach to Stochastic Inte- (see Curtain and Sasane, this conference). The re- gration for Fractional Brownian Motion in a ciprocal approach has also great potential for other Hilbert Space, problems for regular linear systems, for example, H- Tyrone Duncan, B. Pasik-Duncan, infinity control, sampling and the robust stability ra- (University of Kansas) dius. Finally, we remark that similar conclusions hold J. Jakubowski (University of Warsaw) if we replace the assumption that A has a bounded A Hilbert-valued stochastic integration is defined for inverse by the assumption that for some real number an integrator that is a cylindrical fractional Brown- a, aI-A does. ian motion in a Hilbert space. Since the integrator is MTNS – 2002 83 not a semimartingale for the fractional Brownian mo- from closing data of the Standard and Poor’s 500 tions considered, a different definition of integration is (S&P500) stock index during the prior decade 1992- required. Both deterministic and stochastic operator- 2001. The jump-diffusion model combines the usual valued integrands are used. The approach to inte- log-normal diffusion and a log-uniform jump distribu- gration has an analogue with Skorokhod integrals for tion. The results are applied to an optimal portfolio Brownian motion by the basic use of a derivative of application in a computational companion paper. some functionals of Brownian motion. An Ito formula is given for some processes obtained by this stochastic integration. Room: 136, Session: THP4 Chair: Georg Heinig, Vadim Olshevski 7:00-17:30 A Class of Tractable Partially Ob- Title: Computational Methods for Structured served Discrete Stochastic Games, Matrices and Applications William McEneaney (University of California at San Diego) 16:30-17:00 Split Algorithms for Toeplitz and Stochastic games under partial information are typi- Toeplitz-plus-Hankel Matrices, cally computationally intractable even in the discrete- Georg Heinig (Kuwait University) time/discrete-state case. We consider a problem New algorithms for Toeplitz and Toeplitz-plus-Hankel where one player has perfect information. A chief diffi- matrices are presented that are in the spirit of the culty is that the information state for the other player split algorithms of Delsarte/Genin. It is shown that is infinite-dimensional. However, in the problem type the split algorithms are related to ZW-factorizations here, the information state and state-feedback value like the classical algorithms are related to LU- functions are finite-dimensional. Thus computational factorizations. Special attention is paid to skewsym- tractability is greatly enhanced. metric Toeplitz, centrosymmetric, and general T+H matrices.

17:30-18:00 Hybrid Stock Models and Parame- ter Estimation, 17:00-17:30 Structured LDPC Codes, George Yin (Wayne State University), Amin Shokrollahi (EPFL) Q. Zhang (University of Georgia), No Abstract K. Yin (University of Minnesota) In this work, we study a class of hybrid models for the stock market to account for the coexistence of 17:30-18:00 Efficient Matrix Computations in continuous dynamics and discrete events. Different Wideband Communications, from the original geometric Brownian motion mod- Patrick Dewilde (Delft University of Technology), els, both the rate of return and the volatility in the Lang Tong (Cornell University) hybrid model depend on a continuous-time Markov Alle-Jan van der Veen chain. This model can deal with random volatility (Delft University of Technology) by incorporating market trend with other economic Modern telecommunications put increasing demands factors. To use the models requires being able to on the efficient use of bandwidth in a channel. An estimate the values of elements of the generator of important general technique used in wideband com- the underlying Markov chain. We develop a stochas- munications is code division multiple access, and for tic approximation-based algorithm for the estimation large bandwidths, it uses a long, time-varying code to task. The asymptotic properties including conver- achieve the spreading. Optimal and optimally efficient gence and rates of convergence of the algorithm are signal recovery can be achieved using modern meth- proved. Using the estimated generator, one can then ods of time-varying system theory. A nice example of proceed to make equity liquidation decisions. system theory applied to modern communications.

18:00-18:30 Jump-Diffusion Stock Return Mod- 18:00-18:30 Stable Factorization of Hankel and els in Finance: Stochastic Process Density with Hankel-like Matrices, Uniform-Jump Amplitude, Vadim Olshevsky (Georgia State University), Floyd B. Hanson (University of Illinois at Chicago), Michael Stewart J. J. Westman (UCLA) This talk will outline stable Schur-type algorithms for The stochastic analysis is presented for the parameter the factorization of Hankel and Hankel-like matrices. estimation problem for a weighted least squares fitting The algorithm factors the matrix by applying sym- a theoretical jump-diffusion model to the log-returns plectic transformations, instead of hyperbolic trans- formations, to generators. The new algorithms can be MTNS – 2002 84 proven to be backward stable with a particular choice (Kyoto University) of symplectic transformations. This paper is concerned with a study on the varia- tional systems and their adjoints of Hamiltonian con- trol systems and its application to iterative learning control, which is applicable to electro-mechanical sys- Room: 209, Session: THP5 tems. The proposed learning method does not require Chair: Naomi Leonard either the knowledge of physical parameters of the tar- Title: Control and Dynamics of Mechanical Sys- get system nor the time derivatives of the output sig- tems II nals. 16:30-16:50 On the Ball and Beam Problem: Regulation with Guaranteed Transient Perfor- 17:50-18:10 Controllability of Mechanical Sys- mance and Tracking Periodic Orbits, tems with Constraints and Symmetry, Romeo Ortega (LSS-Supelec) Jorge Cortes (University of Twente), Fabio Gomez-Estern, Javier Aracil, Sonia Mart´inez Francisco Gordillo (Consejo Superior de Investigaciones Centif´icas), (Escuelo Superior de Ingenieros) Jim P. Ostrowski (University ov Pennsylvania), Despite the large number of controllers presented in Hong Zhang (Rowan University) the literature for the stabilization of the ball and beam We develop tools within the affine connection formal- system, only a few results are available on the tran- ism for the control of underactuated mechanical sys- sient performance problem. The aim of this paper tems evolving on a principal fiber bundle. We present is to propose an asymptotically stabilizing controller reduced formulations of the Levi-Civita and the non- that ensures that, for a well defined set of initial con- holonomic affine connections in the presence of sym- ditions, the ball remains on the bar during the tran- metries and nonholonomic constraints. Specialized sient. Tuning the controller, the set can be extended controllability tests are developed, and the notion of to include any initial position with zero velocity. The fiber configuration controllability is introduced. The controller is a nonlinear static state feedback that is results are illustrated in a planar blimp. derived using the interconnection and damping assig- ment energy-shaping controller design methodology. As an extension of the regulation problem we also 18:10-18:30 The Use of Information in Swarm propose in this paper a controller that forces the ball Motions of Autonomous Vehicles, and beam to oscillate. This is achieved, within the John Baillieul (Boston University) framework of energy-shaping, by assigning an energy We discuss recent work on the distributed coordinated function that attains its mimimum along the desired control of a system (swarm) of mobile robots whose periodic orbit. A key step in our design is the im- members possess distinct for sensing their position mersion of the oscilator system into the fourth order and environment. Analysis is carried out in which it is system dynamics. assumed that some of the robots have global position knowledge (within some work domain) while other robots are only able to locate themselves only in a 16:50-17:10 Reduction of Controlled Lagrangian local sense with respect to obstacles and other robots. Systems with Symmetries, Formation motions and leader-follower strategies are Dong Eui Chang (Cal Tech) called for, and it is shown that care must be taken to We extend the theory of controlled Lagrangian sys- keep the followers from colliding with obstacles and tems to include systems with symmetry and the La- each other. grangian reduction theory. This extension is crucial to study examples such as spacecraft control, under- water vehicle control, etc.

Room: 210, Session: THP6 17:10-17:30 Constrained Mechanical Systems with Impacts, Chair: Jan Willem Polderman Patrick Hagerty (University of Arizona) Title: New Approaches to Adaptive Control No Abstract 16:30-16:50 Cautious Hierarchical Switching Control of Stochastic Linear Systems, Marco Campi (University of Brescia), 17:30-17:50 Adjoints of Hamiltonian Systems Jaoa Hespanha and Iterative Learning Control, (University of California, Santa Barbara), Kenji Fujimoto, Toshiharu Sugie M. Prandini (University of Brescia) MTNS – 2002 85

In this paper, we propose a supervisory switching logic 17:50-18:10 Geometry of Adaptive Control, that takes into account the uncertainty on the esti- Part II: Optimization and Geodesics, mated process model when performing controller se- Felipe Pait, Diego Colon lection. If the candidate controller set is hierarchically (University of Sao Paulo) structured, the supervisor automatically selects the Two incompatible topologies appear in the study of controller that appropriately compromises robustness adaptive systems: the graph topology in control de- and performance, given the actual level of uncertainty sign, and the coefficient topology in system identifi- on the process description. Randomized algorithms cation. Their incompatibility is manifest in the stabi- make the approach feasible. lization problem of adaptive control. We argue that this problem can be approached by changing the ge- ometry of the sets of control systems under consider- 16:50-17:10 Strong Robustness in Multi-Phase ation: estimating np parameters in an np-dimensional Adaptive Control: the Basic Scheme, manifold whose points all correspond to stabilizable Maria Cadic, Jan Willem Polderman systems. One way to accomplish this is using the (University of Twente) properties of the algebraic Riccati equation. Parame- The general structure of adaptive control systems ter estimation in such a manifold can be approached based on strong robustness is introduced. This adap- as an optimal control problem akin to the determin- tive approach splits into two phases. The first phase istic Kalman filter, leading to algorithms that can be focuses on identification until enough information is used in conjunction with standard observers and con- obtained to design a controller stabilizing the actual trollers to construct stable adaptive systems. system. This is achieved if the input sequence is such that the uncertainty on the system parameters to be controlled decreases sufficiently fast. Then, in the sec- 18:10-18:30 Two Scale High Gain Adaptive Con- ond phase, emphasis is shifted to control. trol, Jan Willem Polderman (University of Twente), Iven Mareels (University of Melbourne) 17:10-17:30 Near Optimal LQR Performance for Simple adaptive controllers based on high gain out- Uncertain First Order Systems, put feedback suffer a lack of robustness with respect to Daniel Miller (University of Waterloo), bounded disturbances. Existing modifications achieve Li Luo (IBM Canada Toronto Laboratory) boundedness of all solutions but introduce solutions In adaptive control, the objective is to provide stabil- that, even in the absence disturbances, do not achieve ity and acceptable performance in the face of signifi- regulation. In this paper a new modification that cant plant uncertainty. However, often there are large achieves the desired robustness without the side-effect transients in the plant output and the control signal of undesirable solutions is proposed. can become excessively large. Here we consider the first order case with the plant parameters restricted to a compact set; we show how to design a (linear time-varying) adaptive controller which provides near optimal LQR performance.

17:30-17:50 Self-Tuning Control for Polynomial Friday August 16, 2002 Systems: an Algorithmic Perspective, Iven Mareels (University of Melbourne) A two stage adaptive or self-tuning control algorithm 8:00-9:00 Room: 101 Plenary Talk applicable to systems described by transition maps that are polynomial in state, input and parameter Anthony Bloch (University of Michigan), variables is discussed. The feedback is defined only Conservative and Dissipative Dynamics in on the basis of past input and output measurements. Classical and Quantum Systems. In a first finite time stage the system to be controlled In this talk I will discuss dissipative behavior in me- is identified together with its state trajectory. In a chanical systems which preserve energy. The talk will second stage a local observer, is used in conjunction encompass both the classical and quantum domains. with a receding horizon control scheme to effectuate In the classical context I will consider Hamiltonian the control objective. We discuss briefly the computa- systems and almost Poisson systems such as nonholo- tional complexity aspects of this approach to adaptive nomic systems. I will also discuss both classical and or self-tuning control. quantum systems of oscillators interacting with fields and the phenomena of decoherence and dissipation in this setting. MTNS – 2002 86

9:00-10:00 Room: 101 Invited Talk The success for the practical realization of such op- Raffaello D’Andrea (Cornell University), timal control problem relies on the succinct use of A State Space Approach to Control of Spatially control-theoretic structure as well as on efficient nu- Interconnected Systems merical techniques. We shall highlight the receding In this talk a state space approach to controlling sys- horizon technique as an example for the former and tems with a highly structured interconnection topol- semi-smooth Newton methods as an example for the ogy is presented. It is shown that by capturing these latter. Receding horizon control is well-known for systems as fractional transformations on temporal and suboptimal control of ordinary differential equations. spatial operators, many standard results in control – Its investigation for optimal control of infinite dimen- such as the bounded real lemma, H-infinity optimiza- sional systems has only recently been initiated. We tion, and robustness analysis – can be generalized ac- propose a receding horizon technique for a class of cordingly. The state space formulation yields condi- control-dissipative systems, which is applicable, for tions that can be expressed as linear matrix inequal- instance, to the Navier-Stokes equations. The tech- ities. Applications of this control design methodol- nique will be justified by means of the system sta- ogy to several hardware test-beds, including formation bilization property. The construction is based on a flight of autonomous vehicles, is also discussed. proper choice of control Liapunov functions. We shall also address the numerical treatment of control and state constraints. While simple in structure they in- troduce non-differentiabilty into the optimal control formulation and cause the solution to suffer a loss of 9:00-10:00 Room: 102 Invited Talk regularity. These features significantly influence nu- Allen Tannenbaum (Georgia Institute of Technology), merical algorithms. We propose the primal-dual ac- Controlled Active Vision in Image Guided tive set strategy as an efficient technique to incorpo- Surgery and Therapy. rate constraints in optimal control problems. We fur- We will describe use biomedical engineering methods ther establish their relationship to semi-smooth New- that can be integrated into complete therapy delivery ton methods and analyze local as well as global con- systems. Such systems will support more effective de- vergence properties. livery of many image-guided procedures: biopsy, min- imally invasive surgery, and radiation therapy, among others. To understand the extensive role of imaging in the therapeutic process, and to appreciate the current usage of images before, during, and after treatment, Morning: we will focus our talk on four main components of Room: 102, Session: FA1 image-guided therapy (IGT)and image guided surgery Chair: Reinhard Laubenbacher (IGS): localization, targeting, monitoring and con- Title: Genetic Networks trol. Specifically, we will describe robust algorithms for 1. segmentation automated methods that cre- 10:30-11:00 Biochemistry by Numbers: Model- ate patient-specific models of relevant anatomy from ing, Signaling and Genetic Networks, multi-modal imagery. 2. registration automated Pedro Mendes, methods that align multiple data sets with each other Alberto de la Fuente, Paul Brazhnik, Stefan Hoops and with the patient. All of the methods will be illus- (Virginia Tech) trated from medical imagery from various modalities Genomics has revolutionized research in biological sci- including MRI, CT, PET, and ultrasound. ences. We have gone from a single-molecule approach to global views of the entire cellular machinery (e.g. with microarrays). The availability of large data sets creates an opportunity to model biochemical networks using a top-down approach, instead of the traditional 9:00-10:00 Room: 126 Invited Talk bottom-up. We describe the two approaches, their Karl Kunisch (University of Graz), pros and cons, and the strategies that we are devel- From Viscoelastic Fluids to Constrained Opti- oping to approach this exciting problem. mal Control To solve large scale optimal control problems gov- erned by partial differential equations, efficient quan- 11:00-11:30 Designer Gene Networks, titative and numerical techniques are required. This Mads Kaern, James J. Collins is the case, for example, for thermal control of visco- (Boston University) elastic fluids. We present our numerical results on Many fundamental cellular processes are governed by this optimal control problem, which is governed by a genetic programs which employ protein-DNA interac- set of partial differential equations of formidable size. tions in regulating function. Owing to recent tech- MTNS – 2002 87 nological advances, it is now possible to design syn- Room: 126, Session: FA2 thetic gene regulatory networks, and the stage is set Chair: Belinda King, Kirsten Morris for the notion of engineered cellular control at the Title: Distributed Parameter Systems: Stabi- DNA level. Theoretically, the biochemistry of the lization and Control, Part II feedback loops associated with protein-DNA interac- tions often leads to nonlinear equations, and the tools 10:30-10:50 An Example of Output Regulation of nonlinear analysis become invaluable. In this talk, for Distributed Parameter Systems with Infi- we describe how techniques from nonlinear dynamics nite Dimensional Exosystem, and molecular biology can be utilized to model, design David Gilliam (Texas Tech University), and construct synthetic gene regulatory networks. We Christopher I. Byrnes present examples in which we integrate the develop- (Washington University, St. Louis), ment of a theoretical model with the construction of Jeff B. Hood,(North Carolina State University) an experimental system. We also discuss the implica- Victor I. Shubov (Texas Tech University) tions of synthetic gene regulatory networks for gene In this short paper we present an example of the geo- therapy, biotechnology, biocomputing and nanotech- metric theory of output regulation applied to solve a nology. tracking problem for a plant consisting of a boundary controlled distributed parameter system (heat equa- tion on a rectangle) with unbounded input and output 11:30-12:00 Function, Design, and Gene Cir- maps and signal to be tracked generated by an infinite cuitry, dimensional exosystem. The exosystem is neutrally Michael A. Savageau (University of Michigan) stable but with an infinite (unbounded) set of eigen- Mathematically Controlled Comparisons have proved values distributed along the imaginary axis. For this useful for the elucidation of biological function and reason the standard methods of analysis do not apply. design. This method will be used to characterize al- ternative patterns for the coupling of expression in gene circuits as well as the switching times of inducible 10:50-11:10 Control of Systems with Infinitely gene circuits. Finally, I will show how the analysis of Many Unstable Modes and Strongly Stabiliz- gene circuits can be used to direct construction of a ing Controllers Achieving a Desired Sensitiv- gene circuit that will produce a circadian clock in the ity, ¨ bacterium Escherichia coli. Suat G¨um¨u¸ssoy,Hitay Ozbay (The Ohio State University) In this paper we consider a class of linear time in- 12:00-12:30 Comparative analysis of mathemat- variant systems with infinitely many unstable modes. ical models of intracellular networks, By using the parameterization of all stabilizing con- Vassily Hatzimanikatis, Amit Mehra, Michael Beste trollers, we show that H-infinity controllers for such (Northwestern University) systems can be computed using the techniques de- Intracellular networks, such as mRNA and protein ex- veloped earlier for infinite dimensional plants with pression networks, signal transduction networks, and finitely many unstable modes. We illustrate connec- metabolic networks, are complex systems that involve tions between the problem solved here and an indirect large number of components and nonlinear interac- method for strongly stabilizing H-infinity controller tions. Mathematical modeling and analysis has been design for systems with time delays. invaluable in analyzing, understanding, and redesign- ing intracellular networks. There is a number of alter- native mathematical representations associated with 11:10-11:30 Receding Horizon Control and every intracellular network. However, the alternative Reduced-Order Methods, representations of a given network are not equivalent Ito Kazufumi (North Carolina State University) with respect to their ability to capture the properties No Abstract of the network. We will compare alternative mathe- matical representations of protein expression networks with respect to their differences and similarities in 11:30-11:50 Some Problems of Control for Non- capturing dynamic and steady state responses to en- linear Partial Differential Equations, vironmental and genetic perturbations. David Russell (Virginia Tech) We intend in this talk to explore several problems in control of nonlinear partial differential equations, fore- most among which will be the problem of controlling the transverse buckling modes of a nonlinear elastic strip moving at constant velocity between two sets of clamping rollers. Control is exercised by rotation of MTNS – 2002 88 one of the roller assemblies. by using the dynamic programming technique, which allows for several possible generalizations that are also considered. 11:50-12:10 Global Stabilization of Systems of Partial Differential Equations Using Finite Di- mensional Controllers, 11:10-11:30 An Addendum to the Problem of Igor Mezic (UCSB) Stochastic Observability, No Abstract Vasile Dragan, Teodor Morozan (Romanian Academy, Bucharest) In this paper the problem of stochastic observability 12:10-12:30 Ouput Regulation of Nonlinear Sys- for a class of linear stochastic systems subject to mul- tems with State Delay, tiplicative white noise and Markovian jumping is in- Emilia Fridman (Tel Aviv University) vestigated. Necessary and sufficient conditions which Regulator equations are derived, which generalize guarantee the stochastic observability are given. Sev- Francis-Byrnes-Isidori equations to the case of re- eral examples are provided to show that stochastic tarded type nonlinear systems. It is shown that, under observability does not implies always, stochastic de- standard assumptions,the regulator problem is solv- tectability. able if and only if these equations are solvable. In the linear case, the solution of these equations is reduced to linear matrix equations. An example of a delayed 11:30-11:50 Combined Optimization of Portfolio Van der Pol equation illustrates the efficiency of the and Risk Exposure of an Insurance Company, results. Daniel Cajueiro, Takashi Yoneyama (Instituto Tecnologico De Aeronautica) This paper presents a model for an insurance com- pany that controls its risk and is allowed to invest in Room: 129, Session: FA3 a financial market with just two assets - a risk free Chair: Wolfgang Kliemann asset and a stock. The new feature of this paper is Title: Stochastic Control and Estimation to consider that the potential profit of this company depends on the dynamical state of the economy. The 10:30-10:50 Algebraic Optimization Techniques aim is to maximize the reserve of an insurance com- for the Estimation of Zero-Beta Pricing Mod- pany whose manager is risk averse. els, Bernard Hanzon (Free University Amsterdam) In the area of quantitative financial modelling the 11:50-12:10 On a Unitary Model for Two-Time Capital Asset Pricing Model (CAPM) and the Ar- Parameter Stationary Processes, bitrage Pricing Theory (APT) play a prominent role. Dan Emanuel Popovici In one subclass, the so-called zero-beta pricing mod- (University of the West Timisoara) els, the least-squares and maximum likelihood estima- The basic problem in prediction theory is to estimate tion problem lead to a non-linear optimization prob- a certain, desired, behavior of the phenomenon un- lem. Classically the non-linearity is handled by using der study (in our case a stationary process) using the an iterative algorithm that tries to find the optimum, information already obtained about it. In order to ac- however there are apparently no guarantees that the complish this we should separate completely the de- global optimum is found in this way. Here we present terministic part from the part corrupted by noises. an algebraic solution of the problem which guarantees This is done via a Wold-type decomposition. In the that the global optimum is found. case of a two-time parameter stationary process we characterize its different parts in some certain Wold- type decompositions in terms of a unitary model given 10:50-11:10 Trajectory Planning Under a by Berger, Coburn and Lebow for the commuting iso- Stochastic Uncertainty, metric pair associated to the process. Ulf J¨onsson(KTH), Clyde Martin (Texas Tech University), Yishao Zhou (Stockholm University) A trajectory planning problem for linear stochastic differential equations is considered in this paper. The Room: 138, Session: FA4 aim is to control the system such that the expected Chair: Patrick Dewilde value of the output interpolates given points at given Title: Stability and Numerics times while the variance and an integral cost of the 10:30-10:50 Parameter Dependent Extremal control effort are minimized. The solution is obtained Norms for Linear Parameter Varying Systems, MTNS – 2002 89

Fabian Wirth (University of Bremen) 11:50-12:10 Pole Placement Under Output Feed- We study families of parameter varying linear systems back: A Simplification of the Problem, with restrictions on the derivative of the parameter Michael Schilmoeller, Joyce O’Halloran variation. A method to construct exact parameter- (Portland State University) ized Lyapunov norms for a wide class of such sys- For a given system defined by the matrix triple tems is presented. This may be used to derive (locally (A,B,C), which polynomials are characteristic poly- Lipschitz) continuous dependence of the exponential nomials of A+BKC as K varies? When is this set of growth rate on the systems data. Furthermore, the polynomials dense in the space of all monic polyno- maximal exponential growth rate may be approxi- mials of degree n? We show that a polynomial occurs mated using only periodic parameter variations. as a characteristic polynomial of A+BKC for some matrix K if and only if it occurs as a characteristic polynomial of A’+B’K’C’ for some matrix K’, where 10:50-11:10 On the Sensitivity of Algebraic Ric- (A’,B’,C’) is related to (A,B,C) via an equivalence cati Equations, relation. Regarding the question of the density of Ji-guang Sun (Umea University) the set of characteristic polynomials, our approach al- Consider the continuous-time algebraic Riccati equa- lows previously known necessary (and generically suf- tion (CARE) and the discrete-time algebraic Riccati ficient) conditions to be rewritten in terms of the fol- equation (DARE) which arise in linear control and lowing two conditions on the matrices B and C: (rank system theory. Appropriate assumptions on the coef- B)(rank C) ¿= n and CB is not the zero matrix. ficient matrices guarantee the existence and unique- ness of symmetric positive semidefinite stabilizing so- lutions. In this paper, we apply the theory of condi- 12:10-12:30 To the Problem of Construction of tion developed by Rice to define condition numbers Liapunov Functions for Continuous Large Scale of the CARE and DARE in the Frobenius norm, and Systems, derive explicit expressions of the condition numbers Vitaliy Slyn’ko, Anatoliy Martynyuk in a uniform manner. (National Academy of Sciences of Ukraine) No Abstract

11:10-11:30 A Numerically Reliable Method for a Neglected but Unsolved Problem: State Feedback Decoupling with Stability for (A, B, Room: 208, Session: FA5 C, D) Quadruples, Chair: Harry Trentelman Delin Chu (National University of Singapore) Title: A Behavioral Approach to Systems, Con- The noninteracting control problem with stability for trol and Coding Theory (A, B, C, D) quadruples has been studied for a long time. But, there are no numerically verifiable solv- 10:30-10:50 A Behavioral Approach to List De- ability conditions and no numerically implementable coding, methods for solving it in the existing literatures. Jan Willem Polderman (University of Twente) Hence, indeed it is still an unsolved problem from the- Margreta Kuijper (The University of Melbourne) oretical and numerical points of view. In this talk, we List decoding may be translated into a bivariate in- develop a numerically reliable method for solving this terpolationproblem. The interpolation problem is to unsolved problem. The main tool that we use is nu- find a bivariate polynomial of minimal weighted de- merical linear algebra technique. gree that interpolates a given set of pairs taken from a finite field. We present a behavioral approach to this interpolation problem. With the data points we 11:30-11:50 Large Stability Property of Solu- associate a set of trajectories. For this set of tra- tions of Large-Scale Discrete-Time Systems, jectories we construct the Most Powerful Unfalsified Tanya Lukyanova, Anatoliy Martynyuk Model. The bivariate polynomial is then derived from (National Academy of Sciences of Ukraine) a specific representation of the MPUM. The aim of this paper is to present one approach to solution of stability problem for discrete-time system based on hierarchical Lyapunov function. The ex- 10:55-11:15 Linear Hamiltonian systems, ample showing the proposed approach efficiency are Paolo Rapisarda (University of Maastricht), given. H.L. Trentelman (University of Groningen) We study linear Hamiltonian systems using bi- linear and quadratic differential forms. Such a representation-free approach allows to use the same concepts and techniques to deal with systems isolated MTNS – 2002 90 from their environment and with systems subject to Particles Undergoing Phase Transitions, external influences, and allows to study systems de- A. Stephen Morse (Yale University) scribed by higher-order differential equations, thus No Abstract dispensing with the usual point of view in classical mechanics of considering first and second-order differ- ential equations only. 10:50-11:10 Stability Properties of Intercon- nected Vehicles, Vijay Kumar, Herbert Tanner, George Pappas 11:20-11:40 Approximate Time-Controllability (University of Pennsylvania) versus Time-Controllability, The paper presents a methodology for analyzing the Amol Sasane (University of Twente), stability of formations of interconnected vehicles that M.K. C¸amlibel (University of Groningen) are based on leader-follower relations. The methodol- This paper studies the notions of approximate time- ogy exploits input-to-state stability properties of ba- controllability and (exact) time-controllability of be- sic leader-follower interconnections and builds on the haviours. In the 1-D case, we show that these two propagation of these properties throughout the net- notions are equivalent and in the n-D case we give an work to establish global stability bounds. This is for- example of a behaviour which is approximately time- malized using the notion of (formation ISS), a weaker controllable, but not time-controllable. Finally, we form of stability than string or mesh stability, which discuss time-controllability of the heat equation. It relates leader input(s) to formation state errors. In turns out that it is time-controllable with respect to this paper we focus on cyclic interconnections of vehi- the so-called Gevrey class of second order. cles and show how the ISS framework can be extended to include these structures. This is the first such re- sult for cyclic graphs that represent formations based 11:45-12:05 On a Class of Time-Varying Behav- on leader-follower controllers. iors, Madhu Belur (University of Groningen), M.K. C¸amlibel (University of Groningen), 11:10-11:30 Formations with a Mission: Stable A.J. Sasane (University of Twente), Coordination of Vehicle Group Maneuvers, J.C. Willems (University of Leuven) Naomi Leonard (Princeton University), We study a class of time-varying systems that we en- Petter Ogren¨ (KTH), counter when we look into decomposition of behav- Edward Fiorelli (Princeton University) iors. This class is the set of behaviors that are them- We present a stable coordination strategy for vehicle selves polynomials in time, with coefficients as time- formation missions that involve group translation, ro- invariant behaviors. Operators that have such behav- tation and expansion. The underlying coordination iors as their kernels are studied. We are led to the framework uses artificial potentials and virtual lead- study of skew polynomial ring as the underlying ring ers. Symmetry in the is exploited to partially decouple for such operators. the mission control problem into a formation manage- ment subproblem and a maneuver management sub- problem. The coordination strategy is illustrated in 12:10-12:30 Synthesis of Strictly Dissipative Sys- the context of adaptive gradient climbing missions. tems and the Strictly Suboptimal State Space H-infinity Control Problem, Harry. L. Trentelman (University of Groningen) 11:30-11:50 Coordinated Control Strategies for In this short paper we study the problem of existence Networked Vehicles: An Application to Au- of a controlled behavior that is strictly dissipative with tonomous Underwater Vehicles, respect to a quadratic supply rate. The relation be- Joao Sousa, Fernando Pereira tween strictness and the rank of a suitable qudratic (Universidade do Porto) differential form that couples the dissipativity prop- The specification and design of coordinated control erties of the hidden behavior and the orthogonal com- strategies for networked vehicles and systems is dis- plement of the plant behavior is analyzed. cussed. A strategy to find the local minimum of an oceanographic scalar field with networked au- tonomous underwater vehicles (AUV) is presented. The strategy consists in coordinating the motions of Room: 209, Session: FA6 the AUVs to implement a modified version of the Chair: Naomi Leonard simplex optimization algorithm. In the original al- Title: Coordinated Control of Vehicle Networks gorithm, the scalar field is given by a function. In the modified version, the scalar field is given by the phe- 10:30-10:50 Stability of Systems of Self-Driven MTNS – 2002 91 nomenon itself. The AUVs sample the phenomenon simple SISO format and develop a systematic design to calculate the directions of descent, and to mini- algorithm based on standard LTI control techniques mize the phenomenon along each direction of descent. to satisfy the stability test. We also supply an ana- The strategy is discussed in the more general context lytical quantification of the level of design difficulty in of coordination and control of networked vehicles and terms of the number of parameters and in terms of the systems. shape of the unknown equilibrium map. Moreover, we remove the requirement of slow forcing for plants with strictly proper output dynamics (and consequent slow 11:50-12:10 Group Shape Feedback Control, convergence) present in earlier works. Raffaello D’Andrea (Cornell University) In this talk we present a method for controlling the shape of a group of vehicles based on absolute and rel- 11:30-12:00 Stabilization of Sets Parametrized ative measurements. The control objective is a user by a Single Variable: Application to Ship Ma- adjustable combination of the shape of the group and neuvering, the absolute position of the group. The techniques Roger Skjetne are applied to a group of sixteen mobile robots in the (Norwegian University of Science and Technology), Cornell Autonomous Vehicles Laboratory. Andrew R. Teel, Petar V. Kokotovic (University of California, Santa Barbara) We consider the problem of stabilizing sets 12:10-12:30 Hamiltonian Structures for Interact- parametrized by a single variable. Using the solu- ing Satellites, tion to the maneuvering problem, the state of the P.S. Krishnaprasad (University of Maryland) system is driven to a path that coincides with the No Abstract set, where the particular location on the path is de- termined dynamically. Attractivity is obtained, and within the control structure, we induce a separation of time scales that allows us to achieve near forward Room: 210, Session: FA7 invariance of the path from a large range of initial Chair: Mrdjan Jankovic conditions. Title: Nonlinear Control and Applications 10:30-11:00 Application of Nonlinear Lyapunov- 12:00-12:30 Nonlinear Control and Automotive based Controllers and Observers to Gasoline Engine Applications, Direct Injection Engine Charge and Torque Mrdjan Jankovic (Ford Research Laboratory Control, Nonlinear control theory has undergone a rapid devel- Ilya Kolmanovsky (Ford Research Laboratory) opment in the past decade. Some of these results and The paper discusses several Lyapunov-based con- ideas have found application in automotive powertrain troller and observer techniques that enhance the con- control. In this paper, we illustrate a particular appli- ventional Speed-Gradient control design approach. cation where the disturbance decoupling paradigm is Our developments cover the treatment of the deriva- used to design a controller that coordinates the elec- tive action, time-delays, backstepping via dynamic tronic throttle and variable cam timing actuators to surface control, and feedforward governor to satisfy achieve a desired transient engine performance. Some pointwise-in-time constraints. Some of these enhance- aspects of this problem relevant to practical applica- ments have been already proposed in the prior litera- tion are also discussed in the paper. ture in the context of specific automotive applications but are generalized here; other developments such as a feedforward governor are new. We also review several concrete applications of these techniques to control of gasoline direct injection engines, including torque, Middle: air-to-fuel ratio and charge control. Room: 102, Session: FM1 Chair: Martin Haenggi 11:00-11:30 Multivariable Extremum Seeking Title: Mathematical Theory of Networks and Feedback: Analysis and Design, Circuits Kartik B. Ariyur, Miroslav Krstic 14:00-14:20 On Switched Hamiltonian Systems, (University of California, San Diego) Arjan van der Schaft (University of Twente), The paper provides a multivariable extremum seek- Maurice Heemels, Karin Gerritsen ing scheme, the first for systems with general time- (Eindhoven University of Technology) varying parameters. We derive a stability test in a In this paper we study the well-posedness and stability MTNS – 2002 92 of a class of switched linear passive systems. Instru- 15:40-16:00 Interconnection Structures in Phys- mental in our approach is the result, also of interest ical Systems: a Mathematical Formulation, in its own right, that any passive linear system with Goran Golo, Orest V. Iftime, Arjan van der Schaft positive definite storage function can be represented (University of Twente) as a port-Hamiltonian system. The power-conserving structure of a physical system is known as interconnection structure. This paper presents a mathematical formulation of the intercon- 14:20-14:40 Parameter Influence on the Zeros of nection structure in Hilbert spaces. Some properties Network Determinants, of interconnection structures are pointed out and their Sven Feldmann (University of Kaiserslautern) three natural representations are treated. The devel- No Abstract oped theory is illustrated on two examples: electrical circuit and one-dimensional transmission line.

14:40-15:00 Canonical Realizations of Linear Time-Varying Systems, Fred Neerhoff, P. van der Kloet (Delft University of Technology) Room: 126, Session: FM2 In this article, general scalar linear time-varying sys- Chair: Belinda King, Kirsten Morris tems are addressed. In particular, canonical realiza- Title: Distributed Parameter Systems: Appli- tions with integrators, multipliers and adders are pre- cations and Computation, Part I sented. Essentially, it is shown that the well-known configurations for constant systems can be generalized 14:00-14:30 Performance Enhancement of Con- to the time-varying context by replacing the conven- trolled Diffusion Processes by Moving Actua- tional eigenvalues by the earlier introduced dynamic tors, eigenvalues. However, it is also shown that at least Michael Demetriou, Nikolaos Kazantzis one configuration is not suitable for such a generaliza- (Worcester Polytechnic Institute) tion. The present research work deals with the systematic development and implementation of a practical algo- rithm for an actuator activation and control policy 15:00-15:20 In Search of Sensitivity in Network through a scheme of moving actuators for systems Optimization, governed by parabolic partial differential equations Mike Chen, Charuhas Pandit, Sean Meyn (PDEs). An illustrative example with simulation re- We consider the average-cost optimization problem sults of an 1-D diffusion process is included to support for a family of stochastic fluid models with Gaussian the paper’s theoretical findings. statistics. It is shown that optimal switching curves scale linearly, while second order sensitivity vanishes as network variability increases. Under general condi- 14:30-15:00 Equilibrium Profiles of Tubular Re- tions, the switching curve is approximately affine, and actor Nonlinear Models, in this case an affine policy obtained from considera- M. Laabissi, M. E. Achhab tion of the deterministic fluid model is approximately (Universit´eChouaib Doukkali), optimal. Joseph Winkin (University of Namur (FUNDP)), D. Dochain (Universit´eCatholique de Louvain) The multiplicity of the equilibrium profiles is ana- 15:20-15:40 Dynamic Eigenvalues for Scalar Lin- lyzed for axial dispersion nonisothermal tubular re- ear Time-Varying Systems, actors described by Arrhenius type nonlinear models. Pieter Van der Kloet, F.L. Neerhoff It is reported that there is at least one steady state (Delft University of Technology) among the physically feasible states for such models. In this paper, an algorithm is derived for comput- Moreover physically meaningful conditions which en- ing the earlier introduced eigenvalues of scalar vary- sure the multiplicity of equilibrium profiles are given. ing systems. These new types of eigenvalues are key quantities for describing the dynamic behavior of such systems. They generalize the conventional antipodes 15:00-15:30 Control of Electronic Material, pertaining to constant systems. Essentially, the al- Katherine Kime (University of Nebraska at Kearney) gorithm performs successive Riccati transformations This talk will consider efforts to control solid-state that gradually triangularize the accompanying time- structures at the quantum level. Such structures are varying system matrix. described by the Schrodinger equation and control is exercised by a time-dependent potential term. A ba- sic type of potential is a modulated barrier, and one MTNS – 2002 93 physical example of control via such a barrier arises in The Widrow’s quantization theorem is analyzed. This quantum cellular automata. We will discuss schemes theorem gives the conditions to be satisfied by the for numerical approximation and also recent work in probability density function of a random signal for its quantum feedback. perfect reconstruction after the quantization process. The disadvantages of this theorem (it’s hypotheses are very restrictive) are envisaged and some solutions to 15:30-16:00 Active Sound Field Attenuation via decrease the effects of these disadvantages, in the case Acoustic Arrays, of different classes of input signals, used in practice, H.T. Banks are presented. No Abstract

15:40-16:00 Functions of System and Their Per- turbations, Room: 129, Session: FM3 Alexey (Olexiy) Tikhonov Chair: William Helton (Taurida National University) Title: Operator Theoretic Methods We introduce the notion of function for a conservative linear discrete-time-invariant system. We show that 14:00-14:20 A Nehari Theorem for Continuous- functions of systems naturally appear for problems re- Time FIR Systems, lated to functional models, allow an informative defi- Gjerrit Meinsma (University of Twente), nition of the transfer function, have interesting appli- Leonid Mirkin (Technion), Zhong (Imperial College, cations to perturbation theory, scattering theory, and London) problems related to factorizations of operator-valued Explicit formulae are derived for Nehari extensions of functions. continuous time FIR systems

14:25-14:45 Optimal Approximation of Linear Room: 138, Session: FM4 Operators: a Singular Value Decomposition Approach, Chair: David Nicholls Siep Weiland (Eindhoven University of Technology), Title: Nonlinear Surface Water Waves: Theory, Hardy Siahaan (Eindhoven University of Technology), Computation and Experiment Anton Stoorvogel (Delft University of Technology) 14:00-14:30 Numerical Simulation of Blow-up We present generalization of singular values and sin- Solutions of the Vector Nonlinear Schr¨odinger gular value decompositions to operators defined on Equation, spaces equipped with the p-norm, where p is arbi- Catherine Sulem (University of Toronto) trary. The problem of optimal rank approximation of We present numerical simulations of blow-up solutions linear operators is investigated in this context. We of the Vector Nonlinear Schr¨odingerequation, which give sufficient conditions for the existence of optimal arises as the subsonic limit of the vectorial Zakharov rank approximants in the p-induced norm and discuss system. In the course of our calculations, we observed an application of general-ized singular values for the the phenomenon of splitting of the solution profile. To identification of dynamical systems from data. capture the structure of the solution, we developed a dynamic mesh refinement method based on a itera- tive grid distribution method introduced by Ren and 14:50-15:10 Geometrical and Spectral Properties Wang (J. Comput. Phys., 159(2000), 246) of the Time-Varying Riccati Difference Equa- tion, Nevio Carpanese (University of Padova) 14:30-15:00 Existence Theory for Traveling Wa- Operator theoretic methods are used to study time- ter Waves in Three Dimensions, varying systems. The focus is on the spectral proper- Walter Craig (McMaster University) ties in order to give a geometrical parameterization of This talk describes the recent results of W. Craig square summable symmetric solutions of time-varying and D. Nicholls on the existence of traveling capilary- Riccati difference equations. gravity water waves in three dimensions. This is a bifurcation problem, with parameters the two di- mensional phase velocity of the solutions, and the 15:15-15:35 A Generalization of the Widrow’s result in the nonresonant case is a construction of Quantization Theorem, two-dimensional bifurcation surfaces of solutions. In Alexandru Isar, Dorina Isar resonant cases there is a higher multiplicity of solu- (“Politehnica” University, Timisoara) tions, and the proof in this case is related to the reso- MTNS – 2002 94 nant Lyapunov center theorm of A. Weinstein and J. a dynamical system then some interesting new phe- Moser. nomena occur. We discuss charactarization of the dy- namic situation with the help of Lyapunov functions and some applications. 15:00-15:30 Numerical Simulation of Traveling Water Waves, David Nicholls (University of Notre Dame) 15:00-15:30 A Relaxation Theorem for Differen- The author will discuss numerical methods for the tial Inclusions with Applications to Stability simulation of two and three dimensional traveling Properties, gravity water waves, and present results on observ- Yuan Wang (Florida Atlantic Univerity), able wave-forms and their stability. If time permits a Eduardo Sontag (Rutgers University), discussion of numerical stability issues in these calcu- Brian Ingalls (Cal Tech) lations will also be presented. The fundamental Filippov–WaˇzwskiRelaxation The- orem states that the solution set of an initial value problem for a locally Lipschitz inclusion is dense in the 15:30-16:00 Similarities between the Quasi- solution set of the same initial value problem for the Bubble and the Generalized Wave Continuity corresponding relaxation inclusion on compact inter- Equation Solutions to the Shallow Water Equa- vals. In our recent work, a complementary result was tions, provided for inclusions with finite dimensional state John H. Atkinson, Joannes Westerink spaces which says that the approximation can be car- (University of Notre Dame) ried out over non-compact or infinite intervals pro- The author will discuss numerical methods for the vided one does not insist on the same initial values. simulation of two and three dimensional traveling This note extends the infinite-time relaxation theo- gravity water waves, and present results on observ- rem to the inclusions whose state spaces are Banach able wave-forms and their stability. If time permits a spaces. To illustrate the motivations for studying such discussion of numerical stability issues in these calcu- approximation results, we briefly discuss a quick ap- lations will also be presented. plication of the result to output stability and uniform output stability properties.

Room: 210, Session: FM5 15:30-16:00 Characterization of the Non- Chair: Lars Gruene, Fabian Wirth Uniform in Time ISS Property and Applica- Title: Input-to-State Stability, Part II tions, Iasson Karafyllis, J. Tsinias 14:00-14:30 Input-to-state stability of pulse (National Technical University of Athens) width modulated control systems, For time-varying control systems various equivalent Andrew Teel L. Moreau, D. Nesic characterizations of the non-uniform in time ISS prop- (University of California, Santa Barbara), erty are established. These characterizations enable Results on stability and input-to-state stability in us to derive sufficient conditions for ISS concerning pulse-width modulated (PWM) control systems are composite time-varying systems. One of the main presented. The results are based on a recent gener- results generalizes the well-known small-gain theo- alization of two time scale stability theory to differ- rem due to Jiang-Teel-Praly for autonomous systems ential equations with disturbances. In particular, av- under the presence of uniform in time ISS. eraging theory for systems with disturbances is used to establish the results. The nonsmooth nature of PWM systems is accommodated by working with up- per semicontinuous set-valued maps, locally Lipschitz inflations of these maps, and locally Lipschitz param- Room: 126, Session: FP2 eterizations of locally Lipschitz set-valued maps. Chair: Belinda King, Ruth Curtain Title: Distributed Parameter Systems: Appli- 14:30-15:00 ISS for Dynamic Inputs, cations and Computation, Part II Fabian Wirth (University of Bremen) 16:30-16:50 POD Based Control of Beam Vibra- The standard definition of ISS consists of a growth tions: Methodology and Experimental Imple- condition in terms of the size of the input and the ini- mentations, tial condition for arbitrary measurable inputs. This Gregory P. Hicks, Brian Lewis property can be characterized by Lyapunov functions. (North Carolina State University) If the class of inputs is restricted to the outputs of In this talk we discuss real time implementation of MTNS – 2002 95 control methodologies for the attenuation of beam vi- 17:30-17:50 The Effect on Control Design of a brations in a smart structure paradigm caused by a Stabilized Finite Element Approximation for narrow-band exogenous force. By narrow-band ex- Burgers’ Equation, ogenous force we mean a periodic force over a narrow Belinda King (Virginia Tech) frequency band or a particular harmonic. In particu- The Galerkin approximation of Burgers equation is lar, a central part of this focus is the Proper Orthogo- numerically unstable for small viscosity coefficients nal Decomposition (POD) reduction technique and its epsilon. By implementing the Galerkin Least Squares application to real-time control of beam vibrations. method, terms can be added stabilizing the approxi- mation. We will use cubic splines in our approxima- tion and will apply linear feedback control law devel- 16:50-17:10 A Comparison of Balancing Tech- oped for the non-stabilized problem. Our results will niques for Reduced Order Controllers for Sys- show that the functional gains for the non-stabilized tems of PDEs, code and the stabilized code for small viscosity coeffi- Belinda King, Katie A. E. Camp cients epsilon will be nearly identical while the func- (Virginia Tech) tional gains for code that use larger viscosity coeffi- Reduced order controllers are often needed to achieve cients will vary considerably. real time control for physical systems described by PDEs. Two methods of reduced order control design are balanced truncation and LQG balancing. Bal- 17:50-18:10 Functional Gain Computations for anced truncation involves reducing the approximating a 1D Parabolic Equation Using Non-Uniform finite-dimensional system and then designing a control Meshes., for the low order system. LQG balancing involves con- John Burns, Belinda B. King, Lizette Zietsman trolling the full order approximating system and then (Virginia Tech) reducing the control. This presentation will describe We consider a numerical algorithm for computing both balanced reduction methods and give numerical functional gains that define optimal feedback laws for results. Dirchlet boundary control of parabolic equations. The focus is on using non-uniform meshes to improve con- vergence of finite element schemes. Since boundary 17:10-17:30 Modeling and Control Issues Asso- control problems of this type often lead to functional ciated with Atomic Force Microscopy, gains with support near the boundary, uniform meshes Ralph Smith (NCSU) are not optimal. Numerical examples are presented Atomic force microscopes employ cylindrical or to illustrate the effectiveness of using a non-uniform stacked piezoceramic actuators to obtain angstrom mesh concentrated near the boundary. level resolution. However, even at the low drive lev- els used to achieve these tolerances, the actuators exhibit hysteresis and certain constitutive nonlineari- 18:10-18:30 A Continuous Control Design ties. These effects must be accurately quantified and Method, controlled to maintain the resolution, robustness and Jeff Borggaard (ICAM,Virginia Tech) speed of both current instruments and future gener- It is clear that initial successes in challenging dis- ations of devices based on this technology. In this tributed parameter control problems will be achieved talk, a variety of techniques for modeling hystere- as researchers continue to exploit the underlying sis in piezoceramic materials will be discussed, and structure of these problems. In this talk, we look at inverse compensation techniques for linear and non- strategies to exploit two features of these problems: linear control design in the microscopes will be pre- the sparsity of finite element approximations and the sented. Finally reduced-order models based on proper case where the dimension of the control inputs and orthogonal decompositions (POD) will be employed to outputs are small, both of which make Chandrasekhar provide controllers which can be implemented in real equations an effective computational tool. time. MTNS – 2002 96

Index

A Brazhnik, Paul FA1 Cory, David G. TUA6 Absil, P.-A. WM4 Brockett, Roger M-Invited Costello Jr., Daniel TUP1 Achhab, M.E. FM2 Broucke, Mireille MP6 Cowan, Noah THA6 Akunov, Taalaibek A. THA5 Buchot, Jean-Marie TUM5 Craig, Walter FM4 Al-Towlem, Tarek TUM3 Bullo, Francesco MP6 Craighead, Steven THM3 Alber, Mark WM1 Bullo, Francesco THM6 Crouch, Peter THM6 Albert, Reka THP1 Burns, John FM6 Curtain, Ruth THA2 Alpay, Daniel MM4 Byrnes, C. I. TUM4 Curtain, Ruth THA2 Alpay, Daniel WA5 Byrnes, C.I. FA2 Curtain, Ruth THP2 Altafini, Claudio MA6 Curtiss, Michael M. MP6 Altafini, Claudio WP6 Antoulas, Thanos WA4 C Antoulas, Thanos WP4 D Antsaklis, Panos WM6 Cadic, Maria THP6 Aracil, Javier THP5 Cajueiro, Daniel FA3 D’Alessandro, Domenico WP6 Ariyur, Kartik B. FA7 Calafiore, Giuseppe TUA7 D’Alessandro, Domenico MA6 Arov, Damir THA2 Calistru, Catalin Nicolae TUP6 D’Andrea, Raffaello F-Invited Atkinson, J. H. FM4 Calkin, Neil MM1 D’Andrea, Raffaello FA6 Aydin, Nuh TUM1 C¸amlibel, M.K. FA5 Datta, Biswa N. WP4 C¸amlibel, M.K. FA5 Datta, Ruchira WP5 Camp, Katie A. E. FM6 Dauphin, G. THP2 Campbell, Stephen TUA7 Davis, Mark WM1 B Campi, Marco THP6 Day, Martin MP2 Bagchi, Arunabha THA3 Caponetto, Riccardo WA6 Dayawansa, Wijesura P. TUA2 Bai, Zhaojun WA4 Caponetto, Riccardo TUA5 De Cock, Katrien MM3 Baillieul, John THP5 Caramanis, Constantine WP5 De Koning, W.L. TUP4 Baker, Brian M. WA1 Carpanese, Nevio FM3 De Koning, Willem L. WA7 Balakrishnan, V. THM4 Carter, John THA4 de la Fuente, Alberto FA1 Ball, Joseph A. TUA3 Casal, Arancha WM1 De Leenheer, Patrick MA5 Ball, Joseph A. TUM4 Cernea, Aurelian MP2 De Moor, Bart MM3 Ball, Joseph A. WM3 Cervera, J. THM6 de Oliveira, Mauricio THA7 Banerjee, Adrish TUP1 Chahlaoui, Younes WA4 de Snoo, Henk TUM4 Banks, H.T. FM2 Chai, Li THM4 Deconinck, Bernard THA4 Barabasi, Albert-Laszlo WAINV2 Chaichanavong, Panu WM2 Deistler, Manfred TUM6 Barhorst, Alan TUA2 Chakraborty, Arup K. WM1 Del Buono, N. WM4 Barry, J. WM2 Chang, Dong Eui THP5 Delattre, Cedric TUP5 Bastin, Georges MA5 Chatterjee, Sankar TUA2 Demetriou, Michael FM2 Basu, Sankar WP3 Chaves, Madalena TUA4 Devanathan, Rajagopalan MM2 Bauer, Peter WA2 Chen, Mike FM1 Dewilde, Patrick WA5 Belur, Madhu FA5 Chesi, Graziano MP2 Dewilde, Patrick THP4 Belyi, Sergey TUM4 Chesi, Graziano THA6 Dey, Subhrakanti TUA5 Bertsimas, Dimitris WP5 Chiuso, Alessandro WM7 Dieci, Luca WM4 Blankenstein, Guido THM6 Chu, Delin FA4 Digailova, Irina TUM6 Bloch, Anthony M. THM6 Climent, Joan-Josep WP2 Dijksma, Aad MM4 Bloch, Anthony M. F-Plenary Cohen, Nir MA4 Dijksma, Aad MP4 Blondel, Vincent WA7 Cole, A. C. MM2 Dochain, Denis FM2 Bolotnikov, Vladimir TUM4 Collins, James J. FA1 Dochain, Denis TUP5 Borggaard, Jeff FM6 Colon, Diego THP6 Dougal, Roger MA2 Bortolin, Gianantonio MA2 Colonius, Fritz TUA4 Dragan, Vasile TUP6 Bose, Nirmal WP3 Conchello, Jose Angel TUM2 Dragan, Vasile FA3 Boulant, Nicolas TUA6 Coombs, Dan WP1 Duncan, Tyrone THP3 Branicky, Michael MP6 Cortes, Jorge THP5 Dustin, Michael TUP2 MTNS – 2002 97

Dym, Harry MM4 Galkowski, K. THA5 Hatzimanikatis, Vassily FA1 Dymkou, S. THA5 Galkowski, Krzysztof TUM3 Havel, Timothy TUA6 Dymkov, M. THA5 Galkowski, Krzysztof THA5 Havel, Timothy F. WA1 Gallivan, K. WA4 Heemels, Maurice FM1 Garrido, R. TUM6 Heinig, Georg THP4 Gayer, Tobias WA6 Heinkenschloss, M. TH-Invited E Geman, Oana TUP6 Helmke, Uwe MA7 Edidin, Michael WM1 Gerritsen, Karin FM1 Helmke, Uwe MM6 Eidelman, Yuli WA5 Gilliam, David FA2 Helmke, Uwe WM4 El Ghaoui, Laurent TUA7 Gluesing-Luerssen, Heide MP5 Helton, J. William W-Plenary Elia, Michele TUM1 Gluesing-Luerssen, Heide TUP1 Helton, J. William THA7 Elia, Nicola MP1 Gluesing-Luerssen, Heide TUP1 Henderson, Diane THA4 Enquist, Per MM3 Gluesing-Luerssen, Heide WA3 Henkel, Heather WP2 Eremenko, Alex MM5 Gohberg, Israel WA5 Hespanha, Jaoa THP6 Evans, Robin J. TUA5 Goldstein, Byron WP1 Hespanha, Joao MP6 Evans, Robin J. MP1 Golo, Goran WM5 Hicks, Gregory P. FM6 Golo, Goran FM1 Hinrichsen, Diederich MP5 Gombani, A. MM3 Hofmann, Stefan TUM6 Gombani, Andrea THA3 Hood, Jeff B. FA2 F Gombao, Sophie TUM5 Hoops, Stefan FA1 Fabijonas, Bruce MA2 Gomez-Estern, Fabio THP5 Hoshi, Yoshikatsu WA7 Fagnani, Fabio MP1 Gordillo, Francisco THP5 Huang, Xianqing TUP4 Farina, Lorenzo MA5 Gosh, Bijoy TUA2 Huang, Xianqing TUP4 Fathpour, Nanaz WM5 Gough, N. E. MM2 Hueper, Knut W-Invited Feintuch, Avraham WA5 Gray, W. Steven MM2 Feldmann, Sven FM1 Greemtree, A.D. WP6 Greferath, Marcus TUM1 Fern`anez-Anaya, G. TUP6 I Ferrante, Augusto MA6 Greferath, Marcus TUM1 Ferrante, Augusto MM3 Gregor, Jiri TUP3 Idczak, Dariusz TUP3 Finesso, Lorenzo WM7 Grimble, Michael MA7 Iftime, Orest V. THA2 Fiorelli, Edward FA6 Grimble, Michael TUM5 Iftime, Orest V. FM1 Foote, Jefferson WA1 Gruene, Lars TUA4 Ingalls, B. FM5 Forgas, Gabor THM1 Guo, B.Z. THM2 Interlando, Carmelo TUM1 Fortuna, Luigi MM6 Gurvits, Leonid MP2 Isar, Alexandru FM3 Fortuna, Luigi TUA5 Gutman, Per-Olof MA2 Isar, Dorina FM3 Fortuna, Luigi WA6 Gutman, Per-Olof MA7 Fortunato, Evan M. TUA6 Schmale, Wiland TUP1 Frasca, Mattia TUA5 Frazho, A. MM4 J French, Mark MA7 H Jacob, Birgit THM2 French, Mark MA7 Jacobsen, Elling W. MA2 Fridman, Emilia FA2 Habets, Luc C.G.J.M. MP5 Jaimoukha, Imad M. WP4 Friedland, Shmuel MA1 Haenggi, Martin WA2 Jakubowski, J. THP3 Frossard, Pascal WA2 Hagerty, Patrick THP5 Jank, Gerhard MM6 Fuad, Yusuf TUP4 Hajek, Bruce M-Plenary Jankovic, Mrdjan FA7 Fuhrmann, Paul A. MA7 Han, Guangyue TUA1 Jaschke, Stefan R. THA3 Fuhrmann, Paul A. TUP1 Hanson, Floyd MA2 Jeoong, Hawoong THP1 Fujimoto, Kenji THP5 Hanson, Floyd B. WP4 Jibetean, Dorina MM6 Fujisawa, Yoshinori WP2 Hanson, Floyd B. THP3 J¨onsson,Ulf FA3 Funahashi, Yasuyuki TUM3 Hanzon, Bernard MM6 Jonckheere, Edmond A. WM5 Fuwa, Yasushi WP2 Hanzon, Bernard TUM6 Jordan, Jens WM4 Hanzon, Bernard TUP6 Jordan, Michael I. WM7 Hanzon, Bernard FA3 Hashimoto, K. THA6 G Hassi, Seppo TUM4 G¨um¨u¸ssoy, Suat FA2 Hassibi, Babak MP4 K Gabrielov, A. MM5 Hassibi, Babak TUA7 Kaczorek, Tadeusz TUP3 MTNS – 2002 98

Kaern, Mads FA1 Lemmon, M.D. WM6 McLaughlin, S. WM2 Kahng, Byungnam THP1 Lemonnier, D. WA4 Meerbergen, K. WA4 Kalyuzhniy-Verbovetzky, D. WM3 Leonard, Naomi FA6 Mehl, Christian MM6 Karafyllis, Iasson FM5 Levchenko, Andrea THA1 Meinsma, Gjerrit FM3 Karow, Michael THM4 Levy, Bernard TUA7 Mendes, Pedro FA1 Kavcic, Aleksandar WM2 Lewin, Paul TUM3 Meyn, Sean FM1 Kawamata, Masayuki THA5 Lewis, Andrew THM6 Mezic, Igor FA2 Kazantzis, Nikolaos WA6 Lewis, Andrew THM7 Mezouar, Youcef THA6 Kazantzis, Nikolaos FM2 Lewis, Brian FM6 Michaletzky, Gyorgy MM3 Kazufumi, Ito FA2 Lewkowicz, Izchak MA4 Michel, A.N. WM6 Kern, Daniel MA2 Li, Tien-Yien MM5 Micheli, Mario WM7 Khaneja, Navin WM2 Li, Yaqin MM2 Michtchenko, Anna WM5 Kheifets, Alexander TUM4 Liberzon, Daniel TUA4 Mikkola, Kalle M. THA2 Kime, Katherine FM2 Lin, Wei TUP4 Miller, Daniel THP6 Kimura, Hidenori TUP6 Lin, Wei TUP4 Minchenko, Leonid MM2 King, Belinda FM6 Lindquist, Anders TUM4 Mirkin, FM3 King, Belinda FM6 Lippuner, Dani WM2 Mirkin, Boris MA7 King, Belinda B. FM6 Liu, Jialing MP1 Mitter, Sanjoy MP1 Klemesrud, Bruce THM3 Loeliger, Hans-Andrea M-Invited Mitter, Sanjoy WM2 Kliemann, W. TUA4 Loeliger, Hans-Andrea WM2 Mohamedy, Alaa TUM5 Koetter, Ralf TUA1 Lomadze, Vakhtang MP5 Molchanov, A.P. WM6 Kokotovic, Petar V. FA7 Lomadze, Vakhtang TUA5 Monico, Christopher WP2 Kolesnikov, Alexander MA2 Lomadze, Vakhtang TUA5 Moore, Helen WM1 Kolesnikov, Alexander MA2 Lopez, L. WM4 Moreau, L. FM5 Kolesnikov, Anatoly MA2 Lototsky, Sergey MP3 Morettin, Andrea THM5 Kolesnikov, Anatoly MP2 Lukyanova, Tanya FA4 Mori, Kazuyoshi MP5 Kolmanovsky, Ilya FA7 Luo, Li THP6 Morozan, Teodor FA3 Komenda, Jan THM7 Lynch, Kevin M. THM6 Morris, Kirsten THM2 Kongprawechnon, Waree TUP6 Morris, Kirsten THP2 Koo, Jaehoon WP3 Morse, A. Stephen FA6 Koutsoukos, Xenofon WM6 Muscato, Giovanni MM6 Krajnik, Eduard WP3 M Kravaris, Costas WA6 Mahony, R. WM4 Kremer, Dirk MM6 Malakorn, Tanit TUM4 Kremer, Dirk MM6 Malinovsky, Vladimir TUA6 N Krener, Arthur J. WA6 Marcus, Brian WM2 Nabieva, Elena WM1 Krishnaprasad, P.S. FA6 Marcus, P. TUA2 Nagamune, Ryozo TUP6 Krstic, Miroslav FA7 Mareels, Iven THP6 Nair, Girish MP1 Kugi, Andreas WM5 Mareels, Iven THP6 Nair, Girish TUA5 Kuijper, Margreta FA5 Mart´inez, Sonia THP5 Nakaura, Shigeki WA7 Kumar, Vijay FA6 Martin, Clyde TUA2 Nayak, Aravind WM2 Kunisch, Karl F-Invited Martin, Clyde TUA2 Neerhoff, F.L. FM1 Kurien, James WM6 Martin, Clyde FA3 Neerhoff, Fred FM1 Kurzhanski, Alexander B. MP2 Martin, Peter MA7 Nesic, D. FM5 Kurzhanski, Alexander B. TUM6 Martinez-Garcia, Juan C. TUM6 Newman, Mark THA1 Kuzmenko, Andrew MP2 Martinez-Garcia, Juan C. TUP6 Nicholls, David FM4 Martynyuk, Anatoliy FA4 Nikoukhah, Ramine TUA7 Martynyuk, Anatoliy FA4 Nikoukhah, Ramine TUA7 Matignon, Denis THP2 L Matignon, Denis THP2 Laabissi, M. FM2 Maze, Gerard WP2 Lagonotte, Patrick TUP5 Mboup, Mamadou TUP5 O Langer, Heinz MP4 McClamroch, N. Harris MM2 O’Halloran, Joyce FA4 Langer, Matthias MP4 McCullough, Scott TUM4 O’Sullivan, Michael TUM1 Le Boudec, Jean-Yves WA2 McEliece, Robert J. TU-Invited Ober, Raimund J. TUM2 Le Boudec, Jean-Yves WA2 McEneaney, William MM2 Ober, Raimund J. TUA6 Lee, Peter L. WM1 McEneaney, William THP3 Ogren, Petter FA6 MTNS – 2002 99

Olivi, Martine TUP6 Q Schmale, Wiland TUP1 Olshevsky, Vadim THP4 Qian, Chunjiang TUP4 Schmidt, Henning MA2 Ooba, Tatsushi TUM3 Qiu, Li THM4 Schovanec, Lawrence TUA2 Opmeer, Mark R. THA2 Schrader, Cheryl TUM6 Ordys, Andrzej TUM5 Schr¨oder,Dierk TUM6 Ortega, Antonio MP6 Schuck, Peter TUP2 Ortega, Romeo THP5 R Schumacher, J. M. (Hans) THA3 Ostrowski, Jim THA6 Sebastian, Abu WA6 Raccanelli, Giorgio MA6 Ostrowski, Jim P. THP5 Segur, Harvey THA4 Radcliffe, James TUM3 Owens, D. H. THA5 Sepulchre, Rodolphe WM4 Ramakrishna, Viswanath WP6 Owens, David H. TUM3 Shang, Ying WM6 Ran, A. C. M. MA4 Owens, David H. THA5 Shen, Jinglai MM2 Rantzer, Anders WA7 Ozbay,¨ Hitay FA2 Shi, Yun Q. WP3 Rapisarda, Paolo FA5 Shokrollahi, Amin THP4 Raymond, Jean-Pierre TUM5 Shondin, Yuri MP4 Rebarber, Richard THM2 Shubov, V.I. FA2 Reddy, Tim WM1 P Siahaan, Hardy FM3 Reurings, Martine C. B. THA7 Silva-Ortigoza, Ramon WA7 Pait, Felipe THP6 Reznick, Bruce WP5 Sindano, H. MM2 Pandit, Charuhas FM1 Ribarits, Thomas TUM6 Sira-Ramirez, Hebert WA7 Papakos, Vasilios WP4 Ricardo, Sandra MM6 Sivergina, Irina TUP5 Pappas, George J. WM5 Roberson, Dawnlee TUM6 Skjetne, Roger FA7 Pappas, George J. THM7 Rocha, Paula TUP3 Sklarz, Shlomo WP6 Pappas, George J. FA6 Rocha, Paula THM5 Skoogh, Daniel WA4 Parrilo, Pablo A. TUA5 Rodman, Leiba MA4 Slyn’ko, Vitaliy FA4 Parrilo, Pablo A. WP5 Rogers, Eric TUM3 Smarandache, Roxana TUM1 Parrilo, Pablo A. THA7 Rogers, Eric THA5 Smarandache, Roxana TUP1 Pasik-Duncan, B. THP3 Rogers, Eric THA5 Smith, Hal MA5 Pavon, Michele MA6 Rosenthal, Joachim TUA1 Smith, Ralph FM6 Pecora, Lou THM1 Rosenthal, Joachim TUP1 Solomon, A. I. MA6 Peeters, Ralf TUP6 Rosenthal, Joachim WP2 Solyom, Stefan WA7 Pereira, Fernando FA6 Rovnyak, Jim MM4 Song, Guobiao MP6 Petersen, Mark A. MA4 Runggaldier, W. J. TH-Invited Sontag, Eduardo TUA4 Picci, Giorgio MM3 Runggaldier, Wolfgang J. THA3 Sontag, Eduardo TH-Plenary Picci, Giorgio WM7 Russell, David FA2 Picci, Giorgio WM7 Sontag, Eduardo FM5 Pinzoni, Stefano MM3 Sorensen, Dan WA4 Pivovarchik, Vjacheslav MP4 Soulier, Fabien TUP5 Plischke, Elmar MP5 S Sousa, Joao FA6 Polderman, Jan Willem THP6 Spitkovsky, I. M. MA4 Saito, Osami TUM3 Polderman, Jan Willem THP6 Srai, Manjit Singh MM2 Sakhnovich, Alexander L. MM4 Polderman, Jan Willem FA5 Staffans, Olof TH-Invited Sakhnovich, L. A. MA4 Polis, Michael TUP5 Staffans, Olof THA2 Sal`azar-Silva, G.H. TUM6 Polpitiya, A. TUA2 Stefan, Radu MM6 Salapaka, Murti WA6 Popov, Andrey MA2 Stern, Lawrence WP1 Sampei, Mitsuji WA7 Popovici, Adriana WP3 Stewart, Michael THP4 Sanei, Ahmad MA7 Popovici, Dan Emanuel WP3 Stockbridge, Richard H. MP3 Sanyal, Amit K. MM2 Popovici, Dan Emanuel FA3 Stoica, Adrian TUP6 Sarkissian, Daniil WP4 Porto, Domenico WA6 Stoorvogel, Anton FM3 Sasane, Amol J. FA5 Prajna, Stephen TUA5 Strang, Gilbert TU-Plenary Sasane, Amol J. THA2 Prandini, M. THP6 Striha, Melissa TUP1 Sasane, Amol J. FA5 Prattichizzo, D. THA6 Sugie, Toshiharu THP5 Sasane, Amol J. THA2 Pravia, Marco A. TUA6 Sulem, Catherine FM4 Savageau, Michael A. FA1 Premaratne, Kamal WA2 Sulikowski, Bartek THA5 Scanavino, Bartolo TUP1 Provost, A. MA5 Sumen, Cenk WM1 Schilmoeller, Michael FA4 Putinar, Mihai WP5 Sun, Ji-guang FA4 Schirmer, Sonia WP6 Sun, Ye WM6 Schirmer, Sonia G. MA6 Suresh Kumar, K. THA3 Schlacher, Kurt WM5 MTNS – 2002 100

T Vandendorpe, A. WA4 Wood, Jeff TU-Invited Tabuada, Paulo WM5 Vandendorpe, Antoine WA4 Woodburn, Cynthia WM3 Taksar, Michael THM3 Varaiya, Pravin MP2 Woolsey, Craig THM6 Talasila, Viswanath WM5 Vasudevan, Lavanya MP6 Wu, Mengnien MM5 Tannenbaum, Allen F-Invited Vatta, Francesca TUP1 Tanner, Herbert FA6 Verduyn Lunel, Sjoerd M-Invited Verriest, Erik WM5 Tannor, David WP6 X Tatikonda, Sekhar MP1 Verschelde, Jan MM5 Tatikonda, Sekhar WM2 Verscheure, Olivier WA2 Xiao, MingQing WA6 Tchobanou, Mikhail WM3 Veselov, Gennady MA2 Xibilia, Maria Gabriella MM6 Teel, Andrew FM5 Vespignani, Alessandro THM1 Xie, Guangming WA7 Teel, Andrew R. FA7 Vettori, Paolo TUP4 Xie, Min WA2 Teklemariam, Grum TUA6 Vicino, A. THA6 Xu, Li TUM3 Tesi, Alberto MP2 Vicino, Antonio MP2 Xu, Li THA5 Theys, Jacques WA7 Vinnikov, Victor TUA3 Xu, Xuping WM6 Thiran, Patrick WA2 Vladimirov, Alexander WA7 Thompson, Nancy L. TUP2 Vojnovic, Milan WA2 Tikhonov, Alexey (Olexiy) FM3 Volosevich, Aleksey MM2 Tomas, Gilberto THP1 Vontobel, Pascal O. TUA1 Y Tong, Lang THP4 Vontobel, Pascal O. WM2 Yamada, Minoru TUM3 Topchiev, Boris TUP4 Yamamoto, Yutaka THM2 Treichl, Thomas TUM6 Yang, Shaohua WM2 Trentelman, Harry L. FA5 W Yin, George MP3 Trentelman, Harry L. FA5 Yin, George THP3 Tretter, Christiane MP4 Wang, Dianhui MP5 Yin, K. THP3 Trumpf, Jochen MA7 Wang, Dianhui TUA5 Ying, Jiang-Qian THA5 Tsekanovskii, E. R. TUM4 Wang, Hui MP3 Yoneyama, Takashi FA3 Tsinias, J. FM5 Wang, Long WA7 Yoshizawa, Shintaro MM6 Turinici, Gabriel TUA6 Wang, Xiaochang MM5 Yu, Runyi MP5 Wang, Xiaochang WP2 Yu, Runyi TUA5 Wang, Xiaoshen MM5 Wang, Yijing WA7 U Wang, Yuan FM5 Ushakov, Anatoly V. THA5 Wang, Yusong MM5 Z Ushida, Shun TUP6 Ward, E. Sally TUM2 Zampieri, Sandro MP1 Weeks, William MM1 Zefran, Milos MP6 Weiland, Siep FM3 Zehetleitner, Kurt WM5 Westerink, Joannes FM4 Zenkov, Dimitri THM6 V Westman, J. J. WP4 Zerz, Eva WA3 Westman, J. J. THP3 Valcher, Maria Elena MA5 Zerz, Eva THM5 Westman, John MA2 van der Kloet, P. FM1 Zhang, Hong THP5 Willems, Jan C. FA5 Van der Kloet, Pieter FM1 Zhang, Jinsong WA2 Willems, Jan C. W-Invited van der Mee, Cornelis MP4 Zhang, Q. THP3 Willems, Jan C. WA3 van der Schaft, Arjan TU-Invited Zhang, Xi Min WP3 Willems, Jan C. THM5 van der Schaft, Arjan WM5 Zhao, Feng WM6 Winkin, Joseph TUP5 van der Schaft, Arjan THM6 Zhirabok, Aleksey WM5 Winkin, Joseph FM2 van der Schaft, Arjan FM1 Zhong, FM3 Wirth, Fabian MP5 van der Schaft, Arjan FM1 Zhou, X. Y. MP3 Wirth, Fabian FA4 van der Veen, Alle-Jan THP4 Zhou, Xun Yu THM3 Wirth, Fabian FM5 van der Woude, J.W. TUP4 Zhou, Yishao FA3 Wittenmark, Bjorn MP1 Van Dooren, P. WA4 Zietsman, Lizette FM6 Woerdeman, H. J. MA4 Van Dooren, P. WA4 Zirilli, Francesco TUM5 Woerdeman, Hugo THM4 van Schuppen, Jan H. MP5 Zwart, Hans THM2 Wofsy, Carla WP1 van Schuppen, Jan H. THM7 Wong, Wing Shing MP1