Robots in Urban Search and Rescue Operations

David Greer, Phillip McKerrow, Jo AbrantesSchool of Information Technology and Computer Science

University of Wollongong, Wollongong 2522.{djg07,phillip,jo@uow.edu.au}

AbstractHelicopters were used to assist in the location ofpersons in inaccessible parts of the Thredbolandslide disaster site. Unfortunately their usehad to be abandoned because the updraughttossed loose material around the site and thevibrations from the rotor blades shook the soilgenerating a risk of further land slippage [Hand,2000]. In this paper, we describe the search andrescue environment, the application of robots inurban search and rescue, an outline of roboticcompetitions in simulating a real rescueenvironment and describe the advantages of ahovering search and rescue robot.

1 IntroductionDisasters can affect anyone without warning. They candisrupt the economic and social balance of society. Urbanand industrial areas are more disaster prone. Such areasare usually heavily populated and have undergonesignificant civil development. The urban environment isliable to experience a variety of disasters that can becategorised into natural disasters and human induceddisasters. Natural disasters include floods, storms,cyclones, tsunami, storm surges, bushfires andearthquakes. The urban environment is often far moresusceptible to human induced disasters such astransportation accidents, industrial accidents, major fires,incidents stemming from hazardous substances in storageor transit, and acts of terrorism, sabotage or war [Cape,1980].

A number of definitions feature the inability to copewith disaster conditions. A disaster is an event over whichthe community can exercise little effective control[Gilmore, 1980]. The lack of trained rescue personnel,and the risks involved in search and rescue operations, arebecoming major problems for emergency serviceorganisations [Murphy et al., 2001]. Specialised robotsmay help alleviate these problems.

As the population changes, so does the potential threatof natural or human induced disasters (Figure 1). AnAustralian Bureau of Transport and Economics reportsupported this trend indicating the number of Australiannatural disasters per year is increasing as populationincreases [Bureau of Transport Economics, 2001].

Figure 1 People Affected by Disaster [Nijhoff, 1994]

People are dependent on the urban environment. It isthe central hub of the economy and interruption can havea catastrophic effect. These factors make the urbanenvironment a prime target for terrorism and sabotage.The World Trade Center (WTC) had been the target ofseveral terrorist attacks before the events of September 112001.

Illustrated in Figure 2, the trend in the number ofsignificant disasters world-wide between 1963 and 1992indicates that the number of deaths, significant damageand persons affected is increasing over time.

Figure 2 Deaths, Damage & Persons Affected by Disaster

[Department of Humanitarian Affairs, 1994]

Proc. 2002 Australasian Conference on Robotics and Automation Auckland, 27-29 November 2002

Copyright © ARAA 2002 25

2 The Search and Rescue EnvironmentAn understanding of the disaster environment is necessaryin order to determine how specialised robots may assist insearch and rescue. A disaster can be divided into fivephases: pre-disaster, pre-impact, impact, emergency, andrecovery [Crane, 1980]. The emergency and the recoveryphases of a disaster are very time consuming. Fordisasters that occur with no warning there is no pre-disaster phase.

Figure 3 Five Phases of a Disaster [Crane, 1980]

One of the initial tasks of the emergency phase is toidentify which areas to target first. This task, known assite assessment and reconnaissance, is a mandatoryprocess to determine the extent of the search and potentialrisks [National State & Territory Emergency Service,1999]. The goal of this phase is to obtain an accurateassessment of the number and location of casualties, anydangerous situations such as gas leaks, live wires,overhanging walls, unsafe structures or anything elsewhich might endanger rescue personnel or survivors[National State & Territory Emergency Service, 1999b].Reconnaissance involves determining access to casualties,the extent and type of damage, the appropriate servicesand support agencies required and an estimate of the timerequired to complete the phase. Unfortunately, this is atime consuming process and a search operation cannotcommence until it is complete.

Once this information is established, rescue teams canrecover surface or easily accessible casualties, explorelikely survival points, begin supporting and removingtrapped persons, and assist in the recovery of the dead.

2.1 Prominent Urban DisastersAustralian urban disasters, the 1989 Newcastle earthquakeand the 1997 Thredbo landslide will be used inconjunction with the 2001 WTC disaster to help illustratethe problems with traditional response activities in UrbanSearch and Rescue (USAR).

1989 Newcastle Earthquake

The Newcastle earthquake is one of the most significanturban disasters in Australia. Occurring at 10.37am onThursday 28th December 1989 and measuring 5.6 on theRichter scale, it was the first time in Australia that anearthquake has caused death and destruction. 13 peopledied, more than 100 people sustained injuries and over10,000 buildings in Newcastle suffered modest tosubstantial damage [Lewis, 1995/96]. Figure 4 illustrates

the Newcastle Workers Club, one of the worst hit areas.

Figure 4 Newcastle Workers Club [Newcastle Regional Library,1999]

1997 Thredbo Landslide

Another prominent urban disaster in Australia was the1997 Thredbo landslide. The event occurred atapproximately 11.30pm on 30 July 1997 at ThredboVillage in Kosciusko National Park [Hand, 2000].Approximately ten thousand tonnes of embankment fillsupporting a road above the village collapsed impactingtwo ski lodges and killing eighteen of the nineteen peoplesleeping in the two lodges. The Thredbo landslideremains the worst natural disaster in Australian history.Figure 5 is an aerial photograph of the site the day afterthe disaster. It illustrates the environment and conditionsthe rescuers were faced with.

Figure 5 Aerial View of the Thredbo Landslide [Hand, 2000]

2001 World Trade Centre Collapse

On September 11th 2001, two fully fuelled Boeing 767aeroplanes deliberately collided with the World TradeCentre in an act of terrorism. The World Trade Centretowers sustained significant damage and subsequentlycollapsed. Many of the surrounding buildings either

Impact

Emergency

Pre-Impact

Pre-Disaster

Recovery

26

collapsed or suffered significant damage. Figure 6illustrates the World Trade Centre site.

Figure 6 World Trade Centre, Tower Two, External View[CRASAR, 2001]

2.2 Search & Rescue DifficultiesUSAR can be defined as:

An integrated multi agency response which isbeyond the capability of normal rescuearrangements to locate, provide initial medicalcare and remove entrapped persons fromcollapsed structures and other environments in asafe and expeditious manner. [EmergencyManagement Australia, 1996]

The definition of USAR introduces the challenges thatresponse teams have to overcome.

Efficient Response

Time is critical during search operations for urbandisasters. Rescue personnel do not have the luxury ofwaiting for a robot to be initialised and prepared for thefield. A robot needs to be deployed without delayotherwise the acceptance of the technology by rescuepersonnel could prove difficult.

It is estimated that for a typical urban search operation,emergency crews have approximately forty eight hours tofind trapped survivors in a collapsed structure, otherwisethe likelihood of finding victims still alive is very low[Sensor Based Planning Lab, 2001]. The demandsgenerated by disasters require an efficient and effectiveresponse.

Safety

Safety is one of the biggest concerns during search andrescue. It is widely accepted that the safety of personnelshould always be put ahead of the provision of assistance[National State & Territory Emergency Service, 1998].The site of the 1997 Thredbo landslide was so unstablethat it was considered unsafe for anyone to attempt arescue. Large slabs of concrete were still moving andwater and mud were running through the site and over thedebris [Hand, 2000]. People were prevented fromentering the site, which delayed and hampered rescueefforts.

Asbestos dust was identified as one of the biggest

safety issues during the rescue effort at the WTC [Nelson,2002]. Built in the 1960/70s, asbestos was extensivelyused in the construction. The collapse of the towergenerated massive quantities of asbestos dust that meantmuch of the equipment exposed was removed anddestroyed. The asbestos problem was so large that it madethe site un-insurable resulting in no liability insurance orcontracts signed.

Asbestos was not the only hazardous material thatrecovery workers had to contend with at the WTC disaster[Nelson, 2002]. Freon gas was extensively used in the airconditioning system. If the Freon gas cylinders were toignite, it would result in deadly hydrochloric and sulphuricacid clouds.

Occupational health and safety legislation require acertain level of dress and personal equipment.Inappropriate dress and equipment is a concern because itunnecessarily increases the risk of further injury to rescuerand casualty.

No matter how severe the situation, rescuers must haverest periods [Price, 1980]. Without rest, rescuers becomebewildered, disheartened and inefficient. Therecommended rest period is 8 hours off every 24 hours[National State & Territory Emergency Service, 1998].

The Environment

The urban disaster environment can be a crime scene sodisturbance must be kept to a minimum. In the case of theWTC disaster, every piece of material removed from thesite was inspected three times by the New York FireBrigade before it was transported to the recovery site[Nelson, 2002]. Locating a body resulted in machinerybeing moved to a different part of the site so the debrisclearing could be completed by hand. Collecting real timedata for research purposes during an emergency can bedifficult. If the disaster is determined to be criminal innature, then information gathering is restricted pending aninvestigation. The Thredbo landslide was another sitewhere data gathering was restricted.

Climatic conditions can also contribute to an unsafeenvironment for rescue personnel. While rain providedsome relief for fire fighting crews at the WTC disaster, itcreated slip and flood hazards for recovery crews trying tolocate survivors. The hostile climatic conditions at theThredbo disaster severely affected the capability ofrescuers and machinery to operate effectively [Hand,2000].

A briefing hosted by the Property Council of Australiain April 2002, highlighted some of the problemsassociated with the WTC environment. The Senior VicePresident of Bovis Lend Lease, Paul Ashlin, detailed thedifficult site terrain. Much of the equipment needed toshift the collapsed steel and concrete could not make it upthe ramps constructed to get to the lower basements.

Inappropriate Equipment & Resources

Having an abundant supply of resources does not ensureadequate disaster response. The unexpected nature of adisaster means that there is often a lack of trainedindividuals available to perform the multitude of tasksduring a rescue [Casper, 2002]. Untrained personnel andvolunteers were some of the problems identified by Bovis

27

Lend Lease at the WTC disaster. Although volunteers hadthe best intentions, many were ill equipped and lackedexperience presenting further danger to themselves andothers. The same problem occurred at the 1997 Thredbolandslide where untrained witnesses tried to climb overthe debris in an attempt to rescue those trapped. To under-take a search operation safely, efficiently and effectively,the appropriate equipment and trained personnel must beused.

3 Robots At The WTC DisasterAccording to an Institute of Electrical and ElectronicEngineers (IEEE) report, researchers have beendeveloping reconnaissance robots for about five years[Vonderheid, 2001]. Research interest increased after the1995 Oklahoma City bombing and the Kobe earthquake inJapan where rescuers witnessed how difficult it was forhumans to manoeuvre safely though the debris.

The World Trade Centre collapse is the first known useof robots for USAR [Gibson, 2001]. Robots from theUniversity of South Florida, manufacturers Foster-Millerand iRobot, and Space and Naval Warfare SystemsCommand were used to search voids and surroundingbuildings, and to look for significant structural damage orevidence for the investigation such as the aeroplanes'voice recorder boxes [Vonderheid, 2001]. Robots used inthe WTC Search and Rescue Operation provided a meansto search small and hazardous spaces, a task usuallycarried out by human rescuers and specially trained dogs.

One of the problems with some of the robots used inthe WTC robot assisted response was the high ratiobetween operator and robot. The physical size of somerobots required several people to transport the robot andset it up. Robots that had little or no autonomousintelligence required additional operators to navigate andcontrol.

USAR robots are often at the mercy of the environmentthat they have to work in. The tracked mechanisms of therobots melted due to the temperature of the materials theywere trying to transverse at the WTC site [Casper, 2002].

Terrain is one of the most challenging problems aUSAR robot has to deal with. The power to weight ratiobecomes very important. Some robots had difficultynegotiating an ascent or decent greater than 45 degrees.Efficient and complex gearing systems are required tomaximise battery life.

4 Rescue Robot CompetitionsOne of the contributing reasons to the lack of robotics inUSAR is the infrequency of disasters. This reduces theability for developers to test and gain practical experience.Training exercises and scenarios can only go so far insimulating the real disaster environment [Casper, 2002].

Companies and institutions hold competitions in anattempt to encourage the development of robotics inUSAR. The Association for Unmanned Vehicle Systemsconducts an International Aerial Robotic competitionannually providing awards for best design, technical paperand presentation [Michelson, 2000].

Competitions are recognised as a way of standardising

robotics throughout the world. International competitionssuch as the RoboCup-Rescue and International AerialRobotics Competition (IARC) provide the opportunity forresearchers to compare and contrast various designs[Davids, 2002]. The desire to win is one of the reasonswhy competitions, such as RoboCup Soccer has over 3000participating researchers from 35 countries [Tadokoro,2000]. Competitions are a popular method to promotecompetitive international research co-operation.

4.1 International Aerial Robotics CompetitionThe 2000 IARC environment attempted to simulate the

disaster environment during and immediately after acatastrophe [Michelson, 2000]. Robots where intended tonavigate wreckage, fire, smoke and aerosols, and copewith acoustic shock waves, motion on the ground and inthe air. Points were awarded for successfully performingtasks that would normally have to be done by humanpersonnel such as hazard identification, locating andextracting casualties, equipment deployment andextinguishing fires.

Illustrated in Figure 7, German Real Time Systems &Robotics Group autonomously operated flying robot,MARVIN – Multi-purpose Aerial Robot Vehicle withIntelligent Navigation, won the 2000 IARC [Musial,2000].

Figure 7 2000 IARC winning robot, MARVINMARVIN was the only robot to fly autonomously,recognise target drums and persons and correctly transmitthe position. However, MARVIN did not fly near thedebris, it flew high over the disaster site and analysedimages to search for people in bright blue overalls lyingon a brown tarmac.

4.2 RoboCup-RescueIt was the 1995 Hanshin-Awaji Earthquake in Kobe City,Japan that killed 6,432 people, and crushed the houses ofone-fifth of the city’s 1.5 million inhabitants resulting in ainfrastructure damage bill of $100US billion that sparkedJapanese researchers Hiroaki Kitano and SatoshiTadokoro to develop RoboCup-Rescue [Tadokoro, 2000].The purpose of RoboCup-Rescue is to promote researchand development in search and rescue robotics toovercome problems with traditional methods ofemergency response.

RoboCup-Rescue began with a disaster simulation

28

environment constructed on network computers thatcreates a virtual environment to conduct search and rescueactivities. In 2001, the RoboCup-Rescue competition wasextended to contain a physical rescue robot competitionknown as RoboCup-Rescue Robot League.

2002 RoboCup-Rescue Robot League

The goal of the 2002 RoboCup-Rescue Leaguecompetition was to accurately map the area and detectcasualties. The competition consisted of three fields, ayellow zone, an orange zone and a red zone based ontestbeds defined by the National Institute of Standards andTechnology (NIST). The yellow zone simulated an officeenvironment, the orange a mock-up of a Japanese-styleroom and the red zone a typical debris field faced byrescuers after a major earthquake. Zones containcasualties based on statistical data from FederalEmergency Management Agency (FEMA) andinternational USAR databases.

The RoboCup-Rescue competition attempts to makethe environment as realistic as possible. Robot operatorsare required to face the opposite direction to simulate thetrue rescue environment where the operator/s cannot seethe target area.

5 Hovering Robot AdvantagesA restrictive feature of current robot technology used insearch operations is that they require a tracked or walkingmechanism in order to manoeuvre. This method ofmovement can cause rubble to move further endangeringtrapped people, is susceptible to snagging, and is largelydependent on the surface topology.

Hovering robots can be used to gather intelligence, getwhere people can't and to establish a connection that canhelp them survive. Robots can explore urbanenvironments contaminated by radiation or chemical orbiological agents. Traditional tracked robots can be toobig and heavy to manoeuvre in the terrain. Tracked robotshave difficultly climbing rubble that is hot and heaped insteep piles.

A hovering robot may overcome some of thelimitations of current response methods given itscapability to reach destinations that may be obstructed bydowned power lines, flooding, fire and damaged roads.The ability to fly offers a distinct advantage. A hoveringrobot can be despatched to initial site assessment activitieswhile emergency crews concentrate on overcomingimpediments blocking the access of more traditionalrescue vehicles and equipment. The speed andmanoeuvrability gains of a hovering solution will enableemergency organisations to more effectively prioritise andschedule rescue and recovery tasks [Millennium Jet Inc.,2002].

One of the advantages a search and rescue robot hasover its human counterpart is the relative speed at whichthey can enter a disaster site and begin to collectinformation. There is no need for shoring up a structurecommonly required to permit human rescue - robots canbe sent in immediately because they are expendable.

A hovering robot provides a means to get lights,cameras, sensors and microphones into rubble pockets or

confined spaces where casualties might be. They can beused to help assess the number and location of casualtiesand help identify dangerous situations such as gas leaks,live wires, overhanging walls, unsafe structures oranything else which might endanger rescue personnel orsurvivors. Robots are unaffected by stench, fatigue or thepsychological effects of rescue situations. Robots are ableto perform tasks humans or animals cannot do, or cannotdo safely [Casper, 2002].

A feature that is found in most USAR robots is the useof on board microphones and speakers to communicatewith rescuers and casualties. Establishing communicationwith a casualty is one of the first steps in administeringfirst aid.

There are alternatives to a hovering robot. USAR robotplatforms vary considerably in terms of size, type, andmobility. There are many different types includingtethered, marsupial, shape-shifting, and serpentine.Researchers at the Sensor Based Planning Laboratoryfrom Carnegie Mellon University, Pittsburgh, areinvestigating the possibility of 'snake like' or serpentinemovement mechanisms [Choset et al., 2001]. Thesemechanisms have several advantages including moredegrees of freedom, a smaller cross sectional area, theycan work in confined areas, and are coupled with hyper-redundant manipulators.

Californian company, Millennium Jet Inc, aredeveloping a concept aircraft called the MULE, MobileUnmanned Lift Enabler. This aircraft is designed toefficiently and autonomously, act as a re-supply vehicleand perform surveillance and reconnaissance. Powered byducted fans it is similar in concept to the hovering robotsolution proposed by the authors of this paper.

The advantages of a hovering vehicle for the initialreconnaissance stage of urban search operations includethe speed at which such a device would be able to surveythe region quickly and thoroughly. Millennium Jet Inc.argue that this is important because traditionalcommunication and transportation systems often failduring and after disasters [Millennium Jet Inc, 2002].

6 ConclusionA thorough understanding of the urban disasterenvironment and an appreciation for traditional search andrescue techniques are crucial to determining the success ofa hovering robot solution. Much can be learnt from theurban disasters discussed. It provides an opportunity tohypothetically apply the proposed hovering robot solutionto the 1989 Newcastle earthquake and 1997 Thredbolandslide and compare and contrast its success with robotassisted response at the WTC disaster.

Technology acceptance is crucial to the success ofrobots in USAR. Until further research is conducted andemergency services begin to accept the technology, theprice for USAR robots will remain high. The cost ofcurrent USAR robots is between $10,000US and$40,000US and can climb much higher depending on thenumber of sensors and other features they have. They arevery expensive when compared to more traditional USARtools and equipment [Gibson, 2001].

The research area is very young and there is much to

29

learn. One of the contributing reasons for the lack ofrobotics in USAR is the infrequency of disasters. Thisreduces the ability for developers to test and gain practicalexperience. Training exercises and scenarios can only goso far in simulating the real disaster environment [Casper,2002].

References[Bureau of Transport Economics, 2001] Bureau of

Transport Economics, Economic Costs of NaturalDisasters in Australia, Bureau of Transport Economics,Canberra, 2001.

[Cape, 2000] T. Cape, 'The Problems and Needs of aCommunity Immediately After Disaster Strikes',National Disaster Relief Committee, Australian RedCross Society, 1980, in J. Oliver (ed.), Response toDisaster, James Cook University of North Queensland,1980, p.111.

[Casper, 2002] Jennifer Casper, Human-RobotInteractions during the Robot-Assisted Urban Searchand Rescue Response at the World Trade Centre,Masters thesis, University of Florida, 2002.

[Choset et al., 2001] Howie Choset, Doug Hull, JonLuntz, Elie Shammas, Search and Rescue, SensorBased Planning Lab, Carnegie Mellon University,Pittsburgh, 2001http://voronoi.sbp.ri.cmu.edu/projects/prj_search_rescue.html. Available: 19 June 2002

[CRASAR, 2001] Center for Robot Assisted Search andRescue, ‘Photos From World Trade Center,http://www.csee.usf.edu/robotics/crasar/pics/allpics/Available: 17 May 2002.

[Crane, 1980] G. Crane, 'Warning Systems: Possibilitiesand Problems', Bureau of Meteorology, Brisbane, 1980,p.47.

[Davids, 2002] A. Davids, ‘Urban Search and RescueRobots: From Tragedy to Technology’, IEEE IntelligentSystems, Institute of Electrical and ElectronicEngineers, March/April 2002,http://www.computer.org/intelligent/ex2002/pdf/x2081.pdf. Available: 20 June 2002

[Department of Humanitarian Affairs, 1994] Departmentof Humanitarian Affairs, Strategy and Action Plan forMitigating Water Disasters in Vietnam, United NationsDevelopment Programme, New York and Geneva,1994. in K.Smith, ‘Environmental Hazards’, AssessingRisk and Reducing Disaster’, Routledge, London, 1996,p. 39.

[Emergency Management Australia, 1996] EmergencyManagement Australia Mt Macedon paper Number 7,Urban Search and Rescue Workshop, September 9 – 12,1996 cited ‘Main Task for and Definition of UrbanSearch and Rescue (USAR)’, Perceptual RoboticsLaboratory, University of South Florida, Tampa.http://www.csee.usf.edu/robotics/USAR/RobotRescue2002/definition.htm. Available: 3 September 2002.

[Gibson, 2001] L. Gibson, ‘USF robots sought survivorsin World Trade Center rubble’, St. Petersburg Times,September 28, 2001.http://www.sptimes.com/New/092801/Hillsborough/USF_robots_sought_sur.shtml. Available: 3 November

2001[Hand, 2000] Derrick Hand. Report of the inquest into the

deaths arising from the Thredbo landslide. New SouthWales Attorney General’s Department, State of NewSouth Wales, 2000, p.161.

[National State & Territory Emergency Service, 1998]National State & Territory Emergency Service, ‘Stormand Water Damage Operations’, Storm DamageResource Kit, ACSES, 1998, p. 6.

[National State & Territory Emergency Service, 1999]National State & Territory Emergency Service, 'Part 1Preparation Introduction and Preparing for GeneralRescue', General Rescue Resource Kit, AustralianCouncil of State Emergency Services, 1999, p. 9.

[National State & Territory Emergency Service, 1999b]National State & Territory Emergency Service, ‘Part 1Preparation’, General Rescue Resource Kit, ACSES,1999, p.10.

[Nelson, 2002] G. Nelson, pers. comm., 15 May 2002.[Nijhoff, 1994] IFRCRCS, World Disasters Report 1994,

Martinus Nijhoff, Dordrecht, 1994. in K.Smith,‘Environmental Hazards’, Assessing Risk and ReducingDisaster’, Routledge, London, 1996, p. 41.

[Michelson, 2000] R. Michelson, 'LaunchPoint for AerialRobotics', Aerial Robotics Competition, Association forUnmanned Vehicle Systems, Georgia Tech ResearchInstitute, Aerospace Transportation & Advanced SystemLaboratory, 2002,http://avdil.gtri.gatech.edu/AUVS/IARCLaunchPoint.html. Available: 20 June 2002.

[Millennium Jet Inc. 2001] Millennium Jet Inc., 'Mule™Mobile Unmanned Lift Enabler', A Practical VTOLPlatform Concept for the 21st Century', 2002,http://www.solotrek.com/mjet/disaster.html Available:19 June 2002.

[Musial, 2000] Marek Musial, ‘MARVIN AnAutonomously Operating Flying Robot’, Real TimeSystems and Robotic Group, Department of ComputerScience, TU Berlin, Germany, http://www.pdv.cs.tu-berlin.de/MARVIN/. Available: 26 August 2002.

[Murphy, 2001] R. Murphy, J. Casper, M. Micre, USARFrequently Asked Questions, CSEE, 2001,http://www.csee.usf.edu/robotics/USAR/faq.htmlAvailable: 1 November 2001.

[Price, 1980] J. Price, 'Living with Disaster', Departmentof Psychiatry, University of Queensland, 1980, in J.Oliver (ed.), 'Response to Disaster', James CookUniversity of North Queensland, 1980, p. 96.

[Smith, 1996] K.Smith, ‘Environmental Hazards’,Assessing Risk and Reducing Disaster’, Routledge,London, 1996, p. 25.

[Tadokoro, 2000] Satoshi Tadokoro, ‘RoboCup-RescueSimulation Project Overview’, RoboCup-RescueCommittee, RoboCup Federation, http://www.r.cs.kobe-u.ac.jp/robocup-rescue/simoverview.html. Available: 29August 2002.

[Vonderheid, 2001] E. Vonderheid, 'Robots Used in WTCSearch and Rescue Operation', The Institute, IEEESpectrum Online, December 2001, Vol. 25, No. 12,http://www.spectrum.ieee.org/INST/dec01/frobots.html.Available: 19 June 2002.

30