Monitoring the Generation and Execution of Optimal Plans By

Total Page:16

File Type:pdf, Size:1020Kb

Monitoring the Generation and Execution of Optimal Plans By Monitoring the Generation and Execution of Optimal Plans by Christian Fritz A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy Graduate Department of Computer Science University of Toronto Copyright c 2009 by Christian Fritz Abstract Monitoring the Generation and Execution of Optimal Plans Christian Fritz Doctor of Philosophy Graduate Department of Computer Science University of Toronto 2009 In dynamic domains, the state of the world may change in unexpected ways during the generation or execution of plans. Regardless of the cause of such changes, they raise the question of whether they interfere with ongoing planning efforts. Unexpected changes during plan generation may invalidate the current planning effort, while discrepancies between expected and actual state of the world during execution may render the executing plan invalid or sub-optimal, with respect to previously identified planning objectives. In this thesis we develop a general monitoring technique that can be used during both plan generation and plan execution to determine the relevance of unexpected changes and which supports recovery. This way, time intensive replanning from scratch in the new and unexpected state can often be avoided. The technique can be applied to a variety of objectives, including monitoring the optimality of plans, rather then just their validity. Intuitively, the technique operates in two steps: during planning the plan is annotated with additional information that is relevant to the achievement of the objective; then, when an unexpected change occurs, this information is used to determine the relevance of the discrepancy with respect to the objective. We substantiate the claim of broad applicability of this relevance-based technique by developing four concrete applications: generating optimal plans despite frequent, un- expected changes to the initial state of the world, monitoring plan optimality during execution, monitoring the execution of near-optimal policies in stochastic domains, and ii monitoring the generation and execution of plans with procedural hard constraints. In all cases, we use the formal notion of regression to identify what is relevant for achieving the objective. We prove the soundness of these concrete approaches and present empir- ical results demonstrating that in some contexts orders of magnitude speed-ups can be gained by our technique compared to replanning from scratch. iii Dedication To my parents. iv Acknowledgements First and foremost I want to thank my supervisor, Sheila McIlraith, who, from all possible perspectives, is the best supervisor I could imagine. This regards not only the scientific guidance she gave me, which was always characterized by a maximal degree of objectivity and rationality, but also the mentoring I have received from her. Sheila was always an active—and pro-active—supporter of my, and her other students’, best interests, constantly looking out for and setting up new opportunities for us, both in research and in academic networking. I greatly appreciate her flexibility regarding the topics I wanted to work on, and her support and encouragement when pursuing my own ideas or collaborating with other students. It is due to what I have learned from her that I now feel ready to pursue independent research. I thank Jorge Baier for the endless discussions, white board sessions, and collabora- tions. Jorge’s analytical skills and his clear and calm way of pondering a problem have always been inspiring to me. Moreover, I greatly value his friendship. Besides Jorge, I was fortunate to have a number of other collaborators whom I would like to thank for working with me: Meghyn Bienvenu, for our collaborations on planning with preferences; Richard Hull and Jianwen Su for giving me the chance to work with them at Bell Labs and beyond, and for broadening my horizon and triggering my interest in work-flow related issues. For their extremely valuable feedback, comments, and suggestions I thank the mem- bers of my supervisory committee, Hector Levesque, Fahiem Bacchus, and Craig Boutilier. The presented thesis has greatly benefited from the thorough review and constructive crit- icism they have provided me with over the course of my program at the department. This also applies to my external examiner, Sven Koenig, who has done an extremely thorough and critical review of this thesis, providing a lot of useful comments and suggestions for future work. I thank Gerhard Lakemeyer, who, as the supervisor of my Master’s research in Aachen, v had given me the chance to apply my algorithms in the highly-dynamic RoboCup domain. This first-hand experience highlighted the need for and triggered my interest in the topic of this thesis. The chance to learn about the entire spectrum from sophisticated, state-of- the-art theory to pragmatic, assumptions-defying practice, has made a lasting impression on me and has always been a source of motivation for my research. As a member of the Knowledge Representation Group, I have benefited numerous times from the feedback provided by its other members, and I thank them for attending and critiquing my practice talks and presentations of early stage ideas. I have always appreciated the friendly, constructive, yet critically objective atmosphere that has been characteristic for our group meetings. My gratitude also extends to the Department of Computer Science itself, for creating a stimulating environment and for hosting a long list of excellent and inspiring guest speakers. Finally, I thank my girlfriend Tanya for her loving and understanding support in hard times. vi Contents 1 Introduction 1 1.1 Motivation................................... 1 1.2 Framework................................... 4 1.2.1 Comparison To Existing Frameworks . 7 1.3 OutlineandContributions . 9 1.3.1 Contributions ............................. 9 2 Background 13 2.1 TheSituationCalculus . 13 2.1.1 Basic Action Theories . 14 2.1.2 The Frame Problem and A Solution for Deterministic Actions . 15 2.2 Regression................................... 17 2.3 NotationandDefinitions ........................... 19 3 Monitoring Plan Validity 21 3.1 Introduction.................................. 21 3.2 Characterizing Existing Approaches . 22 3.2.1 Integrated Planning and Execution . 23 3.2.2 Expectation-Based Monitoring . 27 3.2.3 StateEvaluation ........................... 32 3.3 AnAbstractMonitoringApproach . 33 vii 3.3.1 Other Uses of Relevance in Planning . 34 3.4 Monitoring Plan Validity in the Situation Calculus . ........ 36 4 Monitoring Plan Optimality During Execution 40 4.1 Introduction.................................. 40 4.1.1 Contributions ............................. 41 4.2 A∗ SearchBasedPlanning .......................... 41 4.3 A Sufficient Condition for Optimality . 43 4.4 Algorithm ................................... 47 4.4.1 Annotation .............................. 47 4.4.2 ExecutionMonitoring . 51 4.4.3 ExploitingtheSearchTree. 56 4.5 AnIllustrativeExample ........................... 57 4.6 EmpiricalResults............................... 59 4.7 Discussion................................... 61 5 Generating Optimal Plans in Highly Dynamic Environments 65 5.1 Introduction.................................. 65 5.1.1 Contributions ............................. 66 5.1.2 About Optimality in Dynamic Environments . 68 5.2 Algorithm ................................... 68 5.2.1 Regression-Based A∗ planning .................... 70 5.2.2 Recovering from Unexpected Changes . 72 5.3 EmpiricalResults............................... 76 5.4 Discussion................................... 82 6 Monitoring Policy Execution in Stochastic Domains 85 6.1 Introduction.................................. 85 6.1.1 Contributions ............................. 86 viii 6.2 Background .................................. 87 6.2.1 RepresentingMDPs.......................... 87 6.2.2 SolvingMDPsthroughSearch . 89 6.3 Algorithm ................................... 91 6.3.1 Annotation .............................. 92 6.3.2 ExecutionMonitoring . 95 6.4 AnIllustrativeExample ........................... 98 6.5 Analysis .................................... 101 6.5.1 SpaceComplexity. 101 6.5.2 OptimalityConsiderations . 102 6.5.3 EmpiricalResults. 105 6.6 Discussion................................... 106 7 Generating and Executing Plans with Procedural Control 109 7.1 Introduction.................................. 110 7.1.1 Contributions ............................. 111 7.2 Background .................................. 112 7.2.1 GologandConGolog . 112 7.3 Compiling ConGolog into Basic Action Theories . 115 7.4 Analysis .................................... 124 7.4.1 TheoreticalMerits . 124 7.4.2 PracticalMerits............................ 126 7.5 Monitoring the Execution of ConGolog Programs . 131 7.6 RelatedWork ................................. 135 7.7 Discussion................................... 137 8 Related Work 138 8.1 Replanning .................................. 138 ix 8.1.1 PlanRepair .............................. 140 8.1.2 Backtracking ............................. 147 8.1.3 Learning................................ 149 8.2 ContingencyPlanning ............................ 153 8.3 StateEstimation ............................... 156 8.4 ExternalMonitoring ............................. 162 8.5 Meta-reasoning ................................ 164 8.6 ControlTheory ................................ 165 9 Conclusion 166 9.1 Summary ................................... 166 9.2 Contributions ................................
Recommended publications
  • 4:30 PM 120 Clemens
    Department of Computer Science and Engineering Presents Sheila McIlraith, University of Toronto Semantic Web Services Two trends are emerging in the World Wide Web (WWW). The first is the proliferation of Web Services -- self-contained, Web-accessible software applications and associated distributed systems architectures. The second is the emergence of the "Semantic Web", the vision for a next-generation WWW that is computer interpretable. Today's Web was designed primarily for human use. The vision of Semantic Web Services is to enable reliable, large-scale automation of Web service discovery, invocation, interoperation, composition, monitoring and verification. This will be achieved by describing the properties and capabilities of Web services in an unambiguous, machine-understandable form. In this talk we briefly overview efforts to develop languages and ontologies for describing Web services, focusing on OWL-S, an OWL (Ontology Web Language) ontology for Web services. We follow this with a brief discussion of computational techniques for automating the composition of Web service. Bio: Sheila McIlraith is an Associate Professor in the Department of Computer Science at the University of Toronto. Prior to joining the faculty at U of T in 2004, Dr. McIlraith was a research scientist at Stanford University for six years, and a postdoctoral fellow at Xerox PARC for one. Her current research spans three areas: knowledge representation and reasoning for Web services on the Semantic Web; automated planning and diagnosis; and efficient automated reasoning. Dr. McIlraith has authored over 50 refereed papers. She is an associate editor of the journal Artificial Intelligence and past chair of a number of scholarly meetings including the Third International Semantic Web Conference.
    [Show full text]
  • Bringing Semantics to Web Services with OWL-S
    World Wide Web (2007) 10:243–277 DOI 10.1007/s11280-007-0033-x Bringing Semantics to Web Services with OWL-S David Martin & Mark Burstein & Drew McDermott & Sheila McIlraith & Massimo Paolucci & Katia Sycara & Deborah L. McGuinness & Evren Sirin & Naveen Srinivasan Received: 17 July 2005 /Revised: 2 March 2006 / Accepted: 30 March 2007 / Published online: 3 July 2007 # Springer Science + Business Media, LLC 2007 Abstract Current industry standards for describing Web Services focus on ensuring interoperability across diverse platforms, but do not provide a good foundation for automating the use of Web Services. Representational techniques being developed for the Semantic Web can be used to augment these standards. The resulting Web Service specifications enable the development of software programs that can interpret descriptions This work was performed while Paolucci was at the Robotics Institute, Carnegie Mellon University, Sirin was at the University of Maryland, College Park, and Srinivasan was at the Robotics Institute, Carnegie Mellon University. D. Martin (*) Artificial Intelligence Center, SRI International, Menlo Park, CA, USA e-mail: [email protected] M. Burstein Intelligent Distributed Computing Department, BBN Technologies, Cambridge, MA, USA D. McDermott Computer Science Department, Yale University, New Haven, CT, USA S. McIlraith Department of Computer Science, University of Toronto, Toronto, Canada M. Paolucci DoCoMo Communications Laboratories Europe, Munich, Germany K. Sycara Robotics Institute, Carnegie Mellon University, Pittsburgh, PA, USA D. L. McGuinness Knowledge Systems, Artificial Intelligence Laboratory, Stanford University, Stanford, CA, USA E. Sirin Clark+Parsia, LLC, Arlington, VA, USA N. Srinivasan webMethods, Inc., Fairfax, VA, USA 244 World Wide Web (2007) 10:243–277 of unfamiliar Web Services and then employ those services to satisfy user goals.
    [Show full text]
  • Bringing Semantics to Web Services with OWL-S
    World Wide Web (2007) 10:243–277 DOI 10.1007/s11280-007-0033-x Bringing Semantics to Web Services with OWL-S David Martin & Mark Burstein & Drew McDermott & Sheila McIlraith & Massimo Paolucci & Katia Sycara & Deborah L. McGuinness & Evren Sirin & Naveen Srinivasan Received: 17 July 2005 /Revised: 2 March 2006 / Accepted: 30 March 2007 / Published online: 3 July 2007 # Springer Science + Business Media, LLC 2007 Abstract Current industry standards for describing Web Services focus on ensuring interoperability across diverse platforms, but do not provide a good foundation for automating the use of Web Services. Representational techniques being developed for the Semantic Web can be used to augment these standards. The resulting Web Service specifications enable the development of software programs that can interpret descriptions This work was performed while Paolucci was at the Robotics Institute, Carnegie Mellon University, Sirin was at the University of Maryland, College Park, and Srinivasan was at the Robotics Institute, Carnegie Mellon University. D. Martin (*) Artificial Intelligence Center, SRI International, Menlo Park, CA, USA e-mail: [email protected] M. Burstein Intelligent Distributed Computing Department, BBN Technologies, Cambridge, MA, USA D. McDermott Computer Science Department, Yale University, New Haven, CT, USA S. McIlraith Department of Computer Science, University of Toronto, Toronto, Canada M. Paolucci DoCoMo Communications Laboratories Europe, Munich, Germany K. Sycara Robotics Institute, Carnegie Mellon University, Pittsburgh, PA, USA D. L. McGuinness Knowledge Systems, Artificial Intelligence Laboratory, Stanford University, Stanford, CA, USA E. Sirin Clark+Parsia, LLC, Arlington, VA, USA N. Srinivasan webMethods, Inc., Fairfax, VA, USA 244 World Wide Web (2007) 10:243–277 of unfamiliar Web Services and then employ those services to satisfy user goals.
    [Show full text]
  • Eric Miller Interview
    Volume 1 Issue 2 SIGSEMIS July 2004 Bulletin Theme: “SW Challenges http://www.sigsemis.org for KM” The Official Bimonthly Newsletter of AIS Special Interest Group on Semantic Web and Information Systems SIGSEMIS © 2004 Inside Bulletin EDITORIAL… Editorial................................1 For the second time AIS SIGSEMIS Bulletin is on air. We would like to thank SIGSEMIS Activities ...........2 Eric Miller Interview...........3 you for your warm receipt of the first issue which gives us a motive to IJSWIS Announcement ....11 continue our volunteer hard work towards more activities and services for our Dialogue Column..............18 research community. Our portal site at www.sigsemis.org had 1650 unique Special Issue Theme..........23 visitors (and same number of AIS SIGSEMIS Bulletin 1(1) downloads) from Research in Progress over 50 countries. Inside this issue you can find many interesting things. A Column...............................68 call for papers for our SIG Official Peer Reviewed Journal which will be Special Section: DERI........73 published by IDEA Group (inaugural issue 1/2005). I invite you to consider Chris Bussler Interview....74 this publication outlet as an interesting and high quality journal in which we Regular Columns ..............93 will put all of our efforts in order to meet your high standards for quality and Semantic Search to communicate high impact research. Technologies ......................93 SW Technologies ...............99 Methodologies for SW....103 The special theme that is discussed in this issue is Semantic Web Challenges SW Basics..........................113 for Knowledge Management: towards the Knowledge Web. We would like to SW Calendar ....................115 thank all the contributors of the short articles and especially the directors and SW Research Centers......119 Professors Christoph Bussler and Dieter Fensel of Digital Enterprise Research Projects Corner ................123 Institute (DERI) who contributed to the special section of DERI presentation.
    [Show full text]
  • AAAI-15 / IAAI-15 Conference Program January 25 – 30, 2015 Hyatt Regency Austin, Austin, Texas, USA
    Twenty-Ninth AAAI Conference on Artificial Intelligence Twenty-Seventh Conference on Innovative Applications of Artificial Intelligence AAAI-15 / IAAI-15 Conference Program January 25 – 30, 2015 Hyatt Regency Austin, Austin, Texas, USA Sponsored by the Association for the Advancement of Artificial Intelligence Cosponsored by the National Science Foundation, AI Journal, Baidu, Infosys, Microsoft Research, BigML, Disney Research, Google, NSERC Canadian Field Robotics Network, University of Southern California/Information Sciences Institute, Yahoo Labs!, ACM/SIGAI, CRA Computing Community Consortium, and David E. Smith In cooperation with the University of Texas at Austin Computer Science Department, IEEE Robotics and Automation Society, and the RoboCup Federation MORNING AFTERNOON EVENING Sunday, January 25 Student Newcomer Lunch Tutorial Forum Tutorial Forum Workshops / NSF Workshop Workshops / NSF Workshop AAAI/SIGAI DC AAAI/SIGAI DC Monday, January 26 Tutorial Forum Tutorial Forum Speed Dating Workshops Workshops Opening Reception AAAI/SIGAI DC AAAI/SIGAI DC Open House Open House Invited Talks Robotics / RoboCup Robotics / RoboCup Tuesday, January 27 AAAI / IAAI Welcome / AAAI Awards Invited Talks: Bagnell and Etzioni Shaky Celebration AAAI / IAAI Technical Program AAAI / IAAI Technical Program Poster / Demo Session 1 Funding Information Session Senior Member Blue Sky / Doctoral Consortium, What’s Hot Talks / RSS Talks Virtual Agent Demos Robotics / RoboCup / Exhibits Robotics / RoboCup / Exhibits Robotics Demos General Game Playing
    [Show full text]
  • Invited Talks
    Invited Talks VSL/KR Keynote Talk, Supported by ECCAI scription logics, modal logics, ontologies) and auto - mated deduction (term rewriting, unification theo - Ontology-Based Monitoring ry, combination of decision procedures). He was of Dynamic Systems program chair of the KI’01, CADE’03, LPAR’04, and Franz Baader, TU Dresden RTA’07 conferences, is an editorial board member of several journals in AI and logic in computer Ss - Our understanding of the notion “dynamic system” cience, and has published more than 200 refereed is a rather broad one: such a system has states, articles in major journals and conferences. He is a which can change over time. Ontologies are used to member of the Academia Europaea and an ECCAI describe the states of the system, possibly in an in - fellow, and was for twelve years president of CADE complete way. Monitoring is then concerned with Inc., the organization that runs the International deciding whether some run of the system or all of Conference on Automated Deduction. its runs satisfy a certain property, which can be ex - pressed by a formula of an appropriate temporal logic. I consider different instances of this broad Knowledge Representation Meets framework, which can roughly be classified into two Computer Vision: From Pixels to cases. In one instance, the system is assumed to be a black box, whose inner working is not known, but Symbolic Activity Descriptions whose states can be (partially) observed during a Tony Cohn, University of Leeds run of the system. In the second instance, one has (partial) knowledge about the inner working of the While the fields of KR and computer vision have di - system, which provides information on which runs verged since the early days of AI, there have been of the system are possible.
    [Show full text]
  • Vector Institute AI Master's Summit & Career Fair Program
    AI Master’s Summit & Career Fair September 19th 2019 | 10am to 3:30pm Design Exchange, Toronto AGENDA AGENDA 11:00 am Networking Coffee Break 10:00 am Opening Remarks & Acknowledgement of 11:30 am Industry Panel Discussion: Scholarship Recipients Careers in AI Dr. Garth Gibson, President & CEO, Vector Institute Explore the dynamics of working in AI across a range of industries and learn more about the career paths of these highly experienced individuals as they share their insights and advice for students joining the AI workforce. 10:15 am Keynote Presentations Panel moderated by Melissa Judd, Director, Academic Reflecting on their own career journeys, Marcus Partnerships, Vector Institute. Panelists include: Brubaker and Sheila McIlraith offer thoughtful insights on some of the issues, challenges and opportunities in Mike Cloutier, Canada Chief Data Scientist, Accenture AI and what they mean for master’s graduates entering Peter Danforth, Vice President, Data Science, Loblaw a rapidly changing and dynamic workforce. Marcus Steve Kludt, Chief Technology Officer & Brubaker explores some of the lessons learned through Founding Partner, Canvass Analytics his experiences applying AI techniques to problems Stephany Lapierre, CEO , Tealbook outside of traditional spheres. Sheila McIlraith unpacks Reuben Pereira, Data Scientist , Air Canada the complexity of decision making and the implications it has on responsible AI. Having Impact with AI 12:30 pm Networking Lunch & Career Fair Marcus Brubaker, Faculty Affiliate, Vector Institute; - 3:30 pm Enjoy an afternoon of conversation across two floors. Research Director, Borealis AI; Assistant Professor, You will be designated a floor to begin your networking Department of Electrical Engineering and Computer and career exploration as the afternoon begins and Science, York University will have the opportunity to spend time on each floor connecting with industry sponsors and master’s students Doing the Right Thing: from across Ontario.
    [Show full text]
  • Chapter 1 OWL-S and AGENT-BASED SYSTEMS
    Chapter 1 OWL-S AND AGENT-BASED SYSTEMS David Martin, Mark Burstein, Sheila McIlraith, Massimo Paolucci, Katia Sycara Members of the OWL-S Coalition Abstract: Over the last decade, research in agent-based systems (ABS) has spawned a multi-faceted field, addressing a broad range of challenges and generating a varied array of technical approaches. Web service technologies, in contrast, KDYHDULVHQLQDPRUHLQFUHPHQWDOIDVKLRQZLWKPRUHPRGHVWDLPV DOWKRXJK the vision statements associated with Web services sometimes overlap significantly with those of ABS. Work on Semantic Web services aims to provide richer specifications of services, so as to enable fuller, more flexible automation of service provision and use, support the construction of more powerful tools and methodologies, and promote the use of semantically well- founded reasoning about services. This chapter provides an overview of OWL- S, a Semantic Web services ontology, and discusses its connections with work on agent-based systems. We argue that OWL-S takes some significant, although limited, steps towards a foundation for the deployment of agent technologies on the Web. 1. INTRODUCTION Over the last decade, research in agent-based systems (ABS) has spawned a multi-faceted field, addressing a broad range of challenges and generating a varied array of technical approaches. ABS research topics may be divided, roughly, into those having to do with individual agents, and those having to do with multiagent systems (MAS). The agenda of ABS has been ambitious and, in many respects, visionary. ABS researchers have sought to endow individual agents with characteristics such as autonomy, proactivity, and cooperativeness. In the area of MAS, the focal areas have included agent communication languages, extended conversations between agents, shared knowledge, shared activities, interoperation frameworks, and various kinds of middle-agents.
    [Show full text]
  • Christian James Muise
    [email protected] Dr. Christian Muise www.haz.ca Education Ph.D., Computer Science [Feb. 2009 – Feb. 2014] University of Toronto, Ontario, Canada • Thesis: “Exploiting Relevance to Improve Robustness and Flexibility in Plan Generation and Execution” • Advisors: Professor Sheila A. McIlriath and Professor J. Christopher Beck • CGPA of 11.50 / 12.00 Masters, Computer Science [Sept. 2007 – Jan. 2009] University of Toronto, Ontario, Canada • Thesis: “Exploiting Modern #SAT-Solving Techniques to Generate Implicants” • Advisors: Professor Sheila A. McIlriath and Professor J. Christopher Beck • CGPA of 11.50 / 12.00 Bachelor of Computer Science: Software and Computing, Minor in Mathematics Carleton University, Ottawa, Canada [Sept. 2002 – Aug. 2007] • Thesis: “Audio Panning in Ardour” • Advisor: Dr. Louis Nel • CGPA of 11.19/12.00 Research Visits and Extracurricular Activities [2012/13] Presented PhD research at the Research In Action showcase [2010/11] Judge for University of Toronto ShowCase Competition for visiting high school students [2009] Studied with the planning group at the University of Strathclyde, Glasgow, Scotland [2005/06] Captain of Carleton’s ACM programming contest team [2004] Train-IT Exchange Program to Royal Institute of Technology (KTH), Stockholm, Sweden Honours and Awards [2013] Google Places API Developer Challenge Judge's Choice Award [2012] Top submission for the Graduate Student Symposium of the Canadian AI Conference [2012] Bell Graduate Scholarship [2012] Ray Reiter Graduate Award [2012] Random Hacks of
    [Show full text]