Unmanned Aerial Vehicles: an Overview Maria De Fátima Bento

working papers Unmanned Aerial Vehicles: An Overview Maria de Fátima Bento

A Raven B Mini UAV, during system checks before its maiden flight over Kirkuk,Iraq

U.S. Air Force Photo/Senior Airman Jeremy McGuffin Imagine yourself in the middle of a battlefield with only one truly compelling objective: to maneuver yourself from one point to another and execute your mission — with the reward of your own survival. One eye on the threat, one eye on the horizon! Tension perhaps a deep fear seizes you as you confront mortal danger. This is your last shot! Wouldn’t you rather make it while seated behind a desk at a mission control station far from the raging conflict, directing an aerial vehicle without a human on board?A powered, aerial vehicle that can do more for than you could personally on the battlefield yourself?

nce we tried to Google “UAV” and commercial tasks as data and image The purpose of this column is to and got more than two million acquisition of disaster areas, map build- give the reader an overview of the large citations on the Internet. ing, communication relays, search and number of existing UAV systems and O Try to find the definition rescue, traffic surveillance, and so on. R&D projects as well as the practical of unmanned aerial vehicle (UAV) and A UAV can be remotely controlled, challenges facing UAV designers and you’ll uncover a welter of choices in the semi-autonomous, autonomous, or a applications. literature. So, let’s just say that a UAV combination of these, capable of per- is an aerial vehicle capable of sustained forming as many tasks as you can imag- UAV Classification flight without the need for a human ine, including saving your life. Nowa- Several different groups have proposed operator onboard. days, UAVs perform a variety of tasks creation of reference standards for the Although unmanned aerial vehicles in both military and civil/commercial international UAV community. The (UAVs) are mostly used in military appli- markets. Indeed, many different types European Association of Unmanned cations nowadays, the UAVs can also of UAVs exist with different capabilities Vehicles Systems (EUROUVS ) has perform such scientific, public safety, responding to different user needs. drawn up a classification of UAV systems

54 InsideGNSS january/february 2008 www.insidegnss.com working papers

based on such parameters as flight alti- ing UAVs with examples of their usual smallest platforms that also fly at lower tude, endurance, speed, maximum take missions. altitudes (under 300 meters). Designs off weight (MTOW), size, and so forth. Table 1 identifies four UAVs main for this class of device have focused on We should stress the fact that EUROU- categories: micro/mini UAVs (MAV/ creating UAVs that can operate in urban VS did not create this classification for Mini), tactical UAVs (TUAVs), strategic canyons or even inside buildings, flying certification purposes, but rather with UAVs, and the special task UAVs where along hallways, carrying listening and the main purpose of compiling a uni- only the Decoy and Lethal are currently recording devices, transmitters, or min- versal catalog of UAVs categories as well flying. Let’s take a closer look at each of iature TV cameras. as their associated acronyms. these. The U.S. Defense Advanced Research Table 1, adapted from a EUROUVS Micro and Mini UAVs. Micro and mini Projects Agency (DARPA) has developed publication, presents the various exist- UAVs comprise the category of the a set of criteria with which to distinguish

Category Maximum Take Maximum Flight Endurance Data Link Example (acronym) Off Weight (kg) Altitude (m) (hours) Range (Km) Missions Systems Micro/Mini UAVs Micro (MAV) 0.10 250 1 < 10 Scouting, NBC sampling, Black Widow, MicroStar, Microbat, surveillance inside buildings FanCopter, QuattroCopter, Mos- quito, Hornet, Mite Mini < 30 150-300 < 2 < 10 Film and broadcast industries, Mikado, Aladin, Tracker, DragonEye, agriculture, pollution Raven, Pointer II, Carolo C40/P50, measurements, surveillance Skorpio, R-Max and R-50, Robo- inside buildings, communica- Copter, YH-300SL tions relay and EW Tactical UAVs Close Range (CR) 150 3.000 2-4 10-30 RSTA, mine detection, search Observer I, Phantom, Copter 4, & rescue, EW Mikado, RoboCopter 300, Pointer, Camcopter, Aerial and Agricultural RMax Short Range (SR) 200 3.000 3-6 30-70 BDA, RSTA, EW, mine detec- Scorpi 6/30, Luna, SilverFox, tion EyeView, Firebird, R-Max Agri/ Photo, Hornet, Raven, phantom, GoldenEye 100, Flyrt, Neptune Medium Range 150-500 3.000-5.000 6-10 70-200 BDA, RSTA, EW, mine detec- Hunter B, Mücke, Aerostar, Sniper, (MR) tion, NBC sampling Falco, Armor X7, Smart UAV, UCAR, Eagle Eye+, Alice, Extender, Shadow 200/400 Long Range (LR) - 5.000 6-13 200-500 RSTA, BDA, communications Hunter, Vigilante 502 relay Endurance (EN) 500-1.500 5.000-8.000 12-24 > 500 BDA, RSTA, EW, communica- Aerosonde, Vulture II Exp, tions relay, NBC sampling Shadow 600, Searcher II, Hermes 450S/450T/700 Medium Altitude, 1.000-1.500 5.000-8.000 24-48 > 500 BDA, RSTA, EW weapons Skyforce, Hermes 1500, Heron TP, Long Endurance delivery, communications MQ-1 Predator, Predator-IT, Eagle- (MALE) relay, NBC sampling 1/2, Darkstar, E-Hunter, Dominator Strategic UAVs High Altitude, 2.500-12.500 15.000-20.000 24-48 > 2.000 BDA, RSTA, EW, communica- Global Hawk, Raptor, Condor, Long Endurance tions relay, boost phase Theseus, Helios, Predator B/C, (HALE) intercept launch vehicle, Libellule, EuroHawk, Mercator, airport security SensorCraft, Global Observer, Pathfinder Plus, Special Task UAVs Lethal (LET) 250 3.000-4.000 3-4 300 Anti-radar, anti-ship, anti- MALI, Harpy, Lark, Marula aircraft, anti-infrastructure Decoys (DEC) 250 50-5.000 < 4 0-500 Aerial and naval deception Flyrt, MALD, Nulka, ITALD, Chukar Stratospheric TBD 20.000-30.000 > 48 > 2.000 - Pegasus (Strato) Exo-strato- TBD > 30.000 TBD TBD - MarsFlyer, MAC-1 spheric (EXO) Source: Adapted from “UVS-International-“UAV System producers & Models: All UAV Systems Referenced,” 2006

TABLE 1. UAV Classification

www.insidegnss.com january/february 2008 InsideGNSS 55 working papers

Micro UAVs Requirements: of vertical take off and landing (VTOL) these purposes as a result of their large Specification Requirement — in the near future micro UAVs are dimensions and high capabilities. Size < 15 cm expected to become more practical and Strategic UAVs. As our discussion to Weight 100 g prevalent. Thus, the prospects are good this point suggests, at higher altitudes Payload 20 g for micro and mini UAVs to become UAVs tend to be heavier platforms with intelligent “aerial robots,” that is, fully longer range and endurance. Indeed, Range 1-10 Km autonomous thinking machines. that makes sense, because big platforms Endurance 60 min Tactical UAVs. This category includes can carry a larger payload and, in order Altitude < 150 m heavier platforms flying at higher alti- to reach greater distances while fly- Speed 15 m/s tudes (from 3,000 to 8,000 meters). ing for longer time, they require more Source: “Challenges facing future micro air vehicle Unlike micro and mini UAVs, which are energy. Thus, big platforms are usually development”, D.J. Pines & F. Bohorquez AIAA Journal mostly used for civil/commercial appli- used for high altitude, long endurance of Aircraft, April 2006 cations, tactical UAVs primarily support and long range purposes as is the case TABLE 2. Mav requirements military applications. of the High Altitude Long Endurance Referring again to Table 1 criteria, (HALE) UAVs, which comprise the micro UAV. These criteria are presented tactical UAVs can be divided in six heaviest UAVs. in Table 2. subcategories: Close range (CR), Short HALE platforms are strategic UAVs With the DARPA requirements in Range (SR), Medium Range (MR), Long with a MTOW varying from 2.500 mind and referring to some of the sys- Range (LR), Endurance (EN), and, kilograms up to 12.000 kilograms and a tems and performance specifications finally, Medium Altitude Long Range maximum flight altitude of about 20,000 presented in Table 1, we can see that (MALE) UAVs. meters. They are highly automated, with certain small UAVs systems identified in The lack of satellite communications takeoffs and landings being performed the literature don’t actually satisfy all the (Satcom) systems limits the distances automatically. At any time during its DARPA requirements yet. These include over which close-, short-, and medium- mission the ground control station the Black Widow and Microbat from range UAVs can operate. The absence of (GCS) can control the HALE UAV. Aerovironment, the FanCopter by EMT, Satcom equipment is mainly due to the Northrop Grumman’s military UAV, the and the MicroStar by BAESystems. size, weight, and cost of antennas for this Global Hawk, with 35 hours of endur- The ambitious performance goals type of UAV. ance is probably the most well-known represented in Table 2 are still unat- Long-range UAVs, however, must HALE UAV and offers truly remarkable tained, perhaps due to the fact that not use more advanced technology in order performance. all technologies are scalable yet and also to achieve their missions. Usually, this An example of a non-military HALE to UAV designers’ inability to overcome means incorporating a satellite link or is the electric/solar-powered Helios from environmental constraints with avail- another platform acting as a relay, in Aerovironment operated by NASA. The able technology. order to overcome the communication Helios uses solar panels to power elec- The development of micro-elec- problem between the ground station and trically driven propellers and has set tromechanical systems (MEMS) in a UAV caused by the earth’s curvature. an altitude record of about 30.000 kilo- recent years may help overcome these Medium-range UAV platforms fea- meters. This UAV’s design offers many constraints by enabling the produc- ture more advanced aerodynamical attractive features for civil tasks, such tion of small, highly functional naviga- designs and control systems due to their as Earth observation augmenting and tion hardware (MEMS accelerometers high operational requirements, as exem- complementing remote sensing satel- and piezoelectric rate gyros), room plified by the AAI Corporation’s Shadow lites. Other HALE UAV applications temperature infrared (IR) sensors and 200 and 400 aircraft. include communications, mapping, and charge-coupled device (CCD) cameras As for MALE UAVs, many read- atmospheric monitoring. arrays. (We will return to the subject ers have probably already heard about of MEMS later.) Propulsion and energy the MQ-1 Predator designed and built VTOLs storage technologies still remain critical by U.S. General Atomics Aeronautical A feature of particular interest in rotary research challenges for micro UAVs. Systems. The Predator can operate for wing UAVs is the capability for vertical Turning to the category of Mini UAVs up to 40 hours at a maximum range of take off and landing (VTOL). Although as described in Table 1, we can define 3,704 kilometers and has seen extensive different in weight and configurations, these as any UAV under 30 kilograms service in Kosovo, Afghanistan, and in the VTOL UAVs can be found in the flying at altitudes between 150 and 300 other areas of conflict that puts human mini, CR, SR, MR, and MALE catego- meters, with an endurance of about two pilots at risk. Predators can carry and ries. Indeed, the use of VTOL-capable hours of operation. Although mini UAVs release precision guided missiles, how- UAVs is rapidly increasing due to their are currently predominant — especially ever, endurance UAVs are considered ability to hover over specific sites and fly rotary wing mini UAVs with capabilities the most sophisticated UAV systems for at low altitudes in urban areas.

56 InsideGNSS january/february 2008 www.insidegnss.com These capabilities make them favor- situation awareness, and a UAV’s inter- units, they are not currently capable ites for civil and commercial applications action with the physical world and prob- of meeting UAV requirements for such as surveillance and reconnaissance ably represents a more preferable tech- accurate and precise and navigation in urban canyons and indoors. Conse- nology for these purposes than either due to their inherent measurement quently, several institutes are working GPS or INS. noise. If MEMS inertial accuracy can with this type of platform to aid their On the other hand, however, data be improved by integrating them with research projects, frequently present- from sensors such as the GPS, IMU, and other sensors while simultaneously ing their results at events such as the magnetometer, when combined with the developing improved estimation algo- International Aerial Robotics Compe- information obtained from sequences rithms, however, adequate solutions tition organized by the Association for of images, can significantly increase the may be found. Unmanned Vehicle systems Interna- situation awareness of the vehicle and its Although external sensors, such as tional (AUVSI). operator. So, visual techniques are used vision/radar, are widely used in autono- Because VTOL platforms are mostly for positioning in several UAV projects, mous UAVs, the GPS/INS integration is used by research institutes and agencies, and computer vision plays the most still the most commonly used option in tactical or strategic UAVs are unsuitable important role in the environmental such applications. The best estimate is as test platforms for their experiments. sensing accomplished by UAVs. obtained by combining both INS and Thus, an interest and need persists for GPS measurements using one of the designing micro and mini UAVs with Inertial Sensors existing GPS/INS integration methods. the same degree of autonomy as their Inertial navigation systems — or, more Typically, these techniques rely upon larger counterparts. precisely, inertial measurement units filtering techniques to achieve accurate (IMUs) — are widely used as sensors state estimation. Technology and UAV Autonomy Aerovironment’s Helios strategic The operational requirements of small UAV modified for use by NASA in UAVs — such as flying close to the environmental monitoring ground and inside buildings with a lot of obstacles — introduces problems for a simplistic application of technolo- gies used in larger UAVs. For instance, GNSS-based navigation is successfully used in tactical and strategic UAVs, but is less suitable for smaller UAVs operating close to the ground or around obstacles. This has helped stimulate the application of MEMS technology in highly integrat- ed and light-weight navigation systems, as well as the development of miniature sensors such as microcontrollers and autopilots. Neither inertial nor GNSS naviga- tion systems can provide guidance or collision avoidance for autonomous close-proximity flight, however. These NASA Tom Tschida Tom NASA operations require an accurate position estimation of the UAV relative to sur- for position estimation. Current IMU State Estimation Methods rounding objects. For their part, radar technologies come in many shapes, sizes, One of the most important capabilities technologies are too large and heavy to and costs, depending on the application of an autonomous UAV is its state esti- use for in smaller UAVs. However, the and performance required. Small, light- mation or localization. In fact, reliable development of CMOS cameras and weight, power-efficient, and low-cost localization is an essential component improved digital imaging has promot- MEMS inertial sensors and microcon- of every autonomous land, underwater, ed the development and application of trollers available in the market today and air vehicle. “camera images” for these purposes in help reduce the instability of such plat- The most common algorithm used small, autonomous UAVs. forms making them easier to fly. for state estimation is based on theo- Visual sensing can provide a source Although MEMS inertial sensors retical estimation algorithms such as of data for relative position estimation, offer affordable, appropriately scaled the Kalman filter (KF) and the extend- www.insidegnss.com january/february 2008 InsideGNSS 57 working papers

ed Kalman filter (EKF). Indeed, the KF M. St-Pierre and D. Gingras, cited in cal models in the system dynamics and offers an efficient, iterative means of the Additional Resources section at measurements. Although these methods combining information from several the end of this article) showed that the are considered to be simpler to imple- sensors in order to provide the best esti- EKF exhibited some weakness in com- ment in terms of design, they present mate of the state of the vehicle. parison to the UKF. In fact, because some limitations such as the fact that no Currently, two methods are most the sensor model used in the filter is statistical information is used as input. commonly used to estimate the VTOL strongly nonlinear, the UKF’s estima- As a result, they do not present statistics UAV state. The first is to use a state-space tion performance is better. Moreover, associated with the solution as output, helicopter model from which a KF-based implementation of this sampling-based which plays an important role in post- estimator can be built. Typically, the filter is simpler than that of the EKF processing applications. dynamics of the helicopter is described due to the fact that no derivatives of the Other AI-based approaches for deal- using a conventional six-degrees-of-free- state equations need to be calculated. For ing with the navigation and control of dom (6 DOF) rigid body model: three application in small UAVs, this point is the UAVs include non-linear adaptive degrees for translational motion (along very important given the limitations of control and fuzzy logic. the three body axes) and three degrees on-board computer power. for rotational motion. Thus, the equa- Wan Van der Merwe has proposed an UAVs Challenges tions that describe the translational and interesting approach to state estimation With this background, we can now iden- rotational movement of the helicopter involving the possibility of implement- tify the main issues facing the develop- can be derived (from the Newton-Euler ing the UKF in its square root form, ment of improved UAV capability which equations for a rigid body) and then which has been demonstrated to present are autonomous micro and mini UAVs, expressed by a state-space model. improved numerical stability along with autonomous VTOL applications, and The second approach is to use a sen- reduced computational complexity. (See vision-based systems for navigation sor model and forget the complex heli- his paper cited in Additional Resources and control of autonomous vehicles. In copter model, which is the preferred for further details.) addition to that, collision avoidance, the option. A research group from Stanford development of simultaneous localiza- A survey regarding the estima- University has also presented a method tion and mapping algorithms (SLAM) tion methods for integrated navigation for navigating a small UAV through an and multi-vehicle systems are also chal- systems has identified three common unsurveyed environment. (See the paper lenges that need to be overcome. approaches: the linearized Kalman fil- by J. Langelaan and S. Rock in Addition- UAV Autonomy. Some of the intended ter (LKF) or the extended Kalman filter al Resources.) The authors address the applications proposed for small UAVs — (EKF); sampling based filters, such as problem of estimating the state of an air such as flying close to the ground, inside particle filters and the unscented Kalman vehicle in its environment given the used buildings, or around a lot of obstacles filter (UKF); and artificial intelligence– of limited sensors. The results of the sim- — introduces problems for a simplis- or AI-based methods. The latter category ulations conducted in two dimensions tic application of technologies used in include such techniques as artificial neu- show that while an EKF implementation larger UAVs. For instance, GNSS-based ral networks (ANN) or adaptive neural diverges, the UKF implementation gen- navigation is successfully used in tactical fuzzy information systems (ANFIS). erates consistent estimates of the state of and strategic UAVs, but is less suitable The LKF or EKF has seen extensive the vehicle. for smaller UAVs. use in the design of navigation software. Another sampling-based filter is the Most micro UAVs are too small for Given the simplicity and low computa- particle filter, also known as sequential remote control instrumentation, such tional demand of LK filters, they have Monte Carlo filter, which has been wide- as traditional stability-and-control or been very attractive for low-cost UAV ly developed for nonlinear/non-Gauss- navigation aids. Thus, the autonomy of applications. The EKF can provide a ian processes based on the Bayesian fil- flight of such vehicles is an issue of great more accurate solution but is more tering theory. These methods have been importance. complex than the LKF. It also requires a put aside mainly due to the lack of com- Neither inertial nor GNSS naviga- higher computational overhead. puting power. However, recent work has tion systems can provide guidance or Both of them have been used in a sought to apply these methods in some collision avoidance for autonomous variety of applications and have distinct practical problems such as communica- close-proximity flight, which requires advantages and disadvantages; so, the tions, computer vision, and target track- an accurate position estimation of the choice of which one to use will depend ing for radar. UAV relative to surrounding objects. on the particular situation at hand. Finally, we have AI-based estimation For their part, radar technologies are too The EKF has been successfully methods that are well differentiated from large and heavy to use in smaller UAVs. applied to helicopter-state estimation the other types presented previously. The application of MEMS technology in problems. However, a 2004 compari- AI-based estimation differs from other highly integrated and light-weight nav- son analysis discussed in the paper by methods in its lack of any mathemati- igation systems will help the situation,

58 InsideGNSS january/february 2008 www.insidegnss.com along with the development of minia- ture sensors such as microcontrollers and autopilots. Other factors that need to be addressed to achieve autonomy for small UAVs, as reflected in the DARPA crite- ria in Table 2, include: aerodynamics at low Reynolds numbers; miniaturization of the airframe, components, and pay- load; collaborative control; new concepts in inertial and remote sensing, propul- sion, and energy storage; integration of air traffic management, and robust com- munications (secure and unjammable). Vision-Based Systems. Visual sensing can provide a source of data for relative position estimation, situation aware- AAI’s Aerosonde, an endurance UAV

ness, and a UAV’s interaction with the Corporation AAI physical world and probably represents a preferable technology for these purposes cameras on board, however, it makes niques for simultaneous localization and than either GPS or INS. sense to use those as an alternative. mapping using only visual data. (See the At the same time, data from sen- Now, with cameras available the referenced paper by J. Langelaan.) Other sors such as GNSS, IMUs, and magne- real challenge becomes one of creating Stanford researchers (see K. Jonghyuk et tometers — when combined with the a simple, efficient, robust, and computer alia) have developed a passive GPS-free information obtained from sequences vision algorithm that can convert images navigation for small UAVs operating in of images — can significantly increase into a real-time guidance and obstacle- areas where GPS signals are jammed or the situation awareness of the vehicle avoidance system practical for use in obscured by natural or man-made fea- and its operator. So, visual techniques small UAVs. tures. The navigation method is based on are used for positioning in several UAV Simultaneous Localization and Mapping only an IMU and a monocular camera, projects, and computer vision plays the (SLAM). SLAM algorithms are used to with SLAM providing the cornerstone most important role in the environmen- develop landmark-based, terrain-aided of this work. tal sensing accomplished by UAVs. navigation systems, with the capability Multi-Vehicle Systems. In many The development of CMOS cameras for online map building. Such systems applications, the active cooperation of and improved digital imaging has pro- simultaneously use the generated map several different vehicles such as UAVs, moted the development and application to bind the errors in the inertial naviga- unmanned ground vehicles (UGVs), of “camera images” for these purposes tion system. and airships, has important advantages. in small, autonomous UAVs. Another SLAM techniques have been widely Indeed, over the last few years, research area of considerable interest involves used for ground robot navigation but on the coordination and cooperation of emulating the optic-flow vision of fly- only a few are based on vision sensors. multiples vehicles not only research but ing insects, which employ it to maneu- Usually these techniques are associated competitive activities such as DARPA’s ver through regions with dense obstacles with GPS/INS sensors. A 2002 paper Urban Challenge for UGVs have focused fields. Flying insects don’t have GPS or by S. Lacroix et alia (see Additional on coordination of multiple vehicles. IMUs to perform tasks such as collision Resources) presents a new concept of Recently, research supported by the avoidance, altitude control, take off and autonomous UAV navigation based on a Office of Naval Research and the Uni- landing. Thus, in the last few years, new SLAM algorithm and applied on a 6 DOF versity of California, Berkeley, led to navigation and collision-avoidance tech- airborne platform that the group’s work a collaborative UAV design based on niques have been developed modeled on demonstrated the potential for applying in-the-air task allocation and conflict flying insects. SLAM-augmented, low-cost GPS/INS resolution. Each UAV used in that proj- Obstacle Avoidance. Sensing technolo- systems to UAVs in GPS-denied situa- ect has onboard software that provides gies such as laser range finders or radar tions, such as urban canyons, indoors, low-altitude, vision-based control, task are available, but only for medium and or even underwater. selection and negotiation, and aircraft- large UAVs. These technologies are unre- Another example is the perception to-aircraft communication. alistic for micro and mini UAVs due to system designed for the Karma airship This combination allows for a high their heavy weight and excessive power that applies stereo vision, interest point degree of autonomy at both the individ- requirements. Because most UAVs carry matching, and Kalman filtering tech- ual and group level, requiring a mini- www.insidegnss.com january/february 2008 InsideGNSS 59 working papers

edge in such areas as coordination and control, task planning and execution control of multiple vehicles, and for fleets of UAVs. Led by FEUP, the so-called Asasf program seeks to gather the expertise acquired in airplane design with the existing expertise on the coordination and control of multiple vehicles in order to develop an innovative program on cooperative air vehicles. The main goal is the development of a system where air, ground, and aquatic vehicles perform coordinated actions in order to achieve a common goal. FEUP is also one of the national insti- tutes that participates in the “Projecto de I&T em veículos aéreos nao-tripulados” Hunter MQ-5B, medium altitude endurance undertaken by the Portuguese Air Force tactical unmanned aerial system Academy (PAFA) and Porto University NorthropGrumman photo (UP) with the support of the Ministério mum of human intervention. Future copters and airships. da Defesa Nacional. The main objective work will include the development of a The UAVs used in the experiments of this project is to promote PAFA R&D similar platform based on a larger air- are the helicopters Marvin (TU Berlin), activities in several aeronautics-related frame in order to provide greater weight the Heliv (Robotics, Vision and Control areas of interest for the Portuguese Air and power allowances to allow longer group at the University of Seville, Spain), Force (FAP), especially UAVs. flights and support a variety of sensing and the airship Karma developed by The PAFA UAV is designated and communication devices. For more LAAS (Laboratoire d’Architecture et ANTEX-M ((from the Portuguese information, see the article by A. Ryan et d’Analyse des Systèmes at Toulouse, acronym for “Aeronave Não-Tripulada alia cited in Additional Resources. France). The COMETS UAVs exploit the Experimental – Militar”).) This platform Another example is the BEAR complementarities of distributed sensors has to be able to demonstrate control sys- (BErkeley AeRobot) project, which on different platforms when they work tems, development of intelligent struc- was presented at the 2002 Internation- together in the same mission, employing tures for detection of defects in aircraft, al Conference on Intelligent Robots a vision-based, real-time image process- GNSS navigation system, and control of and Systems. BEAR uses a hierarchi- ing method for multi-UAV (VMUAV) autonomous team vehicles. cal multiagent system architecture for motion estimation. Other partners in the ANTEX-M coordinated team efforts (helicopters Another example of related academic project are the Observatório Astronómi- and ground vehicles), including vision- research in this field is the University of co da Faculdade de Ciências da Universi- based pose estimation of multiple UAVs Sydney, Australia, Autonomous Naviga- dade do Porto (OAFCUP) and Instituto and UGVs. Pursuit-evasion games and tion and Sensing Experimental Research de Engenharia Mecânica e Gestão Indus- learning have also been presented in that or ANSER project. ANSER demonstrates trial (INEGI). The international partners research effort. The main goal of these decentralized data fusion and SLAM in this project are the Institute of Geod- games is to put a team of UAVs and methods on multiple UAVs. esy and Navigation (IGN), University ground vehicles in pursuit of a second FAF Munich, Germany; the University team while concurrently building a map UAV Research in Portugal of California at Berkeley (UCB), Center in an unknown environment. Considerable expertise in the coordina- for Collaborative Control of unmanned Another good example of multi- tion and control of ground, underwater, Vehicles (C3UV) ;Swedish Defense vehicle systems is the COMETS project, and surface autonomous vehicles has Research Agency (FOI); Embraer, funded by the Information Society Tech- developed over the years in Portugal Empresas Brasileiras de Aeronáutica S.A nologies (IST) Program of the Europe- with which this author is familiar. Last (EMB), and Honeywell. an Commission. The main objective of year, two teams from the Faculty of The main focus of IGN’s portion of COMETS is to design and implement Engineering of the University of Porto the project is to develop an accurate a distributed control system for coop- (FEUP) designed and built two airplanes INS/GPS/Galileo navigation system as erative detection and monitoring using for the Portuguese Air Cargo Challenge. a source for feedback control in UAVs. heterogeneous UAVs, particularly heli- The teams developed extensive knowl- Ultimately, the ANTEX-M flight tests

60 InsideGNSS january/february 2008 www.insidegnss.com helicopter research at Carnegie Mellon rootics [18] Ryan, A., Xiao Xiao, Rathiman, S., Tisdale, J., institute, Proceedings of Heli Japan, April 1998 Zennaro, M., Caveney, D., Sengupta, R., Hedrick, [3] ANTEX-M website: http://whale.fe.up.pt/ J.: Colloborative flight demonstration of three asasf/index.php/Antex-M unmanned aerial vehicles, AIAA Conference on [4] Brown, R.G., Hwang, P.Y.C: Introduction to Guidance, Navigation and Control, Keystone, CO, Random Signals and Applied Kalman Filtering 2006. with Matlab Exercises and Solutions, 3rd edition, [19] Sarris, Z.: Survey of UAV Applications in Civil John Wiley & Sons, 1997 Markets, Technical University of Crete, June 2001 ANTEX-M, Portuguese Air Force Academy UAV [5] Brown, T., Doshi, S., Jadhav, S., and Himmel- [20] Schumacher, C., Sahjendra, N.: Nonlinear stein, J.: Test bed for a wireless network on small control of multiple UAVS in close-coupled for- will take place at the GATE (Galileo Test UAVs, Proceedings of AIAA 3rd Unmanned Unlim- mation flight, AIAA Guidance, Navigation, and ited Technical Conference, pp. 20–23, September and Development Environment) facility Control Conference and Exhibit, Denver, CO, Aug. 2004 14-17, 2000 in Berchtesgaden, Germany, within the [6] Doitsidis, L., Valavanis, K.P., Tsourveloudis, [21] Sorenson, W., Ed.: Kalman Filtering Theory next two years. N., Kontitsis, M.: A Framework for Fuzzy Logic and Applications, Piscataway, NJ:IEEE Press, Based UAV Navigation and Control, Proceedings 1985 Conclusions of the IEEE, International Conference on Robotics [22] St-Pierre, M, Gingras, D.: Comparison A surprising and seemingly vast num- & Automatation, New Orleans, LA April 2004 between the unscented Kalman filter and the ber of different types of UAVs exist in [7] European Unmanned Vehicles Systems Asso- extended Kalman filter for the position estima- the literature, with different capabili- ciation, Voice of Forum for the Unmanned Vehicles tion module of an integrated navigation infor- ties responding to different user needs. Systems Community, EUROUVS, 1998 mation system, IEEE Proceedings, Intelligent We have reviewed the four main cat- [8] Gavrilets, V.: Autonomous Aerobatic Maneu- Vehicles Symposium, University of Parma, June egories: MAV/Mini UAVs; Tactical vering of Miniature, PhD thesis, Massachussetts 14-17, 2004 Institute of Technology, 2003. UAVs; Strategic and special task UAVS. [23] Taha, R., Noureldin, M. M., and El-Sheimy, [9] Godsill, S.J, Doucet, A., and West, M.: Monte N.: Improving INS/GPS Positioning Accuracy Dur- MAV/mini UAVs represent the smallest Carlo Smoothing for Nonlinear Time Series, In ing GPS Outages Using Fuzzy Logic, Proceedings class of UAVs and are mostly used for Symposium on Frontiers of Time Series Modeling, of the 16th annual Technical Meeting of Satellite civil applications. Strategic UAVs are Tokyo, Japan, Institute of Statistical Mathemat- Devision, Institute of Navigation (ION) GPS-GNSS the largest and mostly used in military ics, 2000 2003, Portland, Oregon, pp. 499-508, 2003 applications. Although the tactical and [10] Jonghyuk, K., Sukkarieh, S.: SLAM Aided [24] Van der Merwe, Wan: Sigma-point Kalman strategic UAVs are the more used, in the GPS/INS Navigation in GPS Denied and Unknown filters for integrated navigation, In: 60th Annual mean time MAVs and Mini UAVs will Environment, the 2004 International Symposium Meeting of the Institute of navigation, 2004. become more practical and prevalent. on GNSS/GPS, Sydney, Australia [25] Vidal, R., Sastry, S., Kim, J., Shakernia, O., Different kinds of UAV platforms [11] Jun, M., Roumeliotis, S.G.: State estimation and Shim, D.: The Berkeley Aerial Robot Project have different mission and applications. via sensor modeling for helicopter control using (BEAR). 2002 IEEE/RSJ International Conference an indirect Kalman Filter, In: IEEE/RSJ Interna- For instance, most research institute on Robots and Systems – IROS, Proceedings of the tional Conference on Intelligent Robots and Sys- workshop WS6 Aerial Robotics (pp. 1-10), 2002 prefers rotary wing UAVs with verti- tems, 1999 cal take off and landing capacities as [12] Lacroix, S., Jung, I.K., Soueres, P., Hygounenc, Acknowledgments test platforms for demonstrating their E., and Berry, J.P.: The autonomous blimp proj- The author would like to thank the Por- research subjects. International compe- ect of LAAS/CNRS- Current status and research tuguese Air Force Academy as well as titions such as the Aerial Robotic com- challenges, IEEE/RSJ International Conference petition organized by AUVSI are very on Intelligent Robots and Systems – IROS 2002, the Portuguese Calouste Gulbenkian important, not only as a good way to Proceedings of the workshop WS6 Aerial Robotics, Foundation which are supporting the promote and share research results but pp.35-42, 2002 Ph.D. studies of the author. [13] Langelaan, J., Rock, S.: Navigation of Small also to understand what is going on in Author the field of UAVs. UAVs Operating in Forests, Guidance, Navigation and Control Conference, August 16-19, 2004 As we said before, it is unreasonable Maria de Fátima Bento [14] Langelaan, J., Rock, S.: Passive GPS-Free to know all the ins and outs of UAVs. received her degree in Navigation for Small UAVs, IEEE Aerospace Con- Avionics Engineering, That is why one can say: Once we tried ference, Big Sky, Montana 2005 from the Portuguese Air to “Google” UAV, we are still Googling [15] Micro and Mini UAVs: International activity Force Academy (PAFA), UAV,” and . . . we haven’t found the end review, EUROUVS, 2005 Sintra, Portugal, and the yet! [16] Nikolos, Y., Doitsidis, L., Christopoulos, V., Master of Science degree Tsourveloudis, N. C.: Roll Control of Unmanned in Satellite Positioning and Navigation from the References Aerial Vehicles using Fuzzy Logic, WSEAS Trans- Faculty of Sciences, Porto University. She has been [1] Almeida, P., Bencatel, R., Gonçalves, G., Sousa, actions on Systems, vol. 4, no. 2, pp. 1039-1047, a lecturer at the PAFA since 2006 and is a Ph.D. J.: Multi-UAV Integration for Coordinated Mis- 2003 student at the Institute of Geodesy and Navigation sions, FEUP, 2006 [17] Ollero, A., Merino, L.: Annual reviews in Con- of the University FAF Munich. Her main subject is [2] Amidi, O., Kanade, T., Miller, J.R.: Autonomous trol 28, pp 167-178, 2004 sensor fusion for applications in UAVs. www.insidegnss.com january/february 2008 InsideGNSS 61