meeting report ntu csie adviser : prof. jane hsu speaker : wen-chieh fang 2005/09/21
TRANSCRIPT
Meeting reportMeeting report
NTU CSIENTU CSIEAdviserAdviser :: Prof. Jane HsuProf. Jane HsuSpeaker Speaker :: Wen-Chieh Fang Wen-Chieh Fang 2005/09/212005/09/21
AgendaAgenda
Simulation ScenarioSimulation Scenario Behavior-based Reactive ParadigmBehavior-based Reactive Paradigm Problem description Problem description Robot model Robot model Random walkRandom walk CoordinationCoordination Area boundingArea bounding Emergency detectingEmergency detecting DemoDemo Summary and future work Summary and future work
Simulation scenarioSimulation scenario
Surveillance taskSurveillance task :: TThree robots on their arhree robots on their area to survey emergenea to survey emergencycy (( red color objecred color objectt ))
Robots are Robots are behavior-bbehavior-basedased (( random-walk, random-walk, coordination, area bocoordination, area bounding,emergency detunding,emergency detectingecting ))
Behavior-based Reactive Behavior-based Reactive ParadigmParadigm
Vertical decomposition of tasksVertical decomposition of tasks
PLANSENSE ACT
Problem descriptionProblem description
ProjectProject :: ITRI planITRI plan ProblemProblem :: Multi-robot coordinationMulti-robot coordination AssumptionAssumption ::
Robots are Robots are homogeneoushomogeneous Robots can Robots can communicatecommunicate without lost information without lost information Robots have no Robots have no model/knowledgemodel/knowledge of other robots of other robots Robots have no ability of Robots have no ability of self-localizationself-localization
Robot modelRobot model
Koala robotKoala robot :::: HigHighly modular, all terraihly modular, all terrain indoor robot n indoor robot
SensorsSensors :: 1616 Infra-r Infra-red proximity and ambed proximity and ambient light sensorsient light sensors
AccessoriesAccessories :: a a cacameramera, an Infra-red gr, an Infra-red ground proximity sensoound proximity sensorr Adopted from: http://www.cyberbotics.com
Random walkRandom walk
AlgorithmAlgorithm : : BraitenbBraitenberg vehicleerg vehicle approach approach
FeatureFeature :: Relying on the Relying on the chancchanc
eses to find targets to find targets Different from completDifferent from complet
e approach to canonice approach to canonical al clean-floorclean-floor task task (( mainly for stationarmainly for stationary targetsy targets ))
Adopted from [Acar et. al. 2003]
Random walk
Back & forth
CoordinationCoordination
MessagingMessaging ““inhibitioninhibition” message” message
StateState none none → “→ “commitmentcommitment” state” state
AssumptionAssumption At most At most twotwo robots tackle the same robots tackle the same
emergencyemergency
Area boundingArea bounding
Each robot is equippEach robot is equipped with an ed with an distance sdistance sensorensor below the base below the base of robot.of robot.
There are There are landmarkslandmarks on the bounding of son the bounding of surveillance areaurveillance area
Emergency detectingEmergency detecting
DeviceDevice :: CameraCamera AssumptionAssumption :: we uswe us
e stationary e stationary red colorred color objectsobjects in place of e in place of emergencies such as imergencies such as intruders or fire.ntruders or fire.
DemoDemo
Summary & future workSummary & future work
Make devices such as camera, emitter and receiver acMake devices such as camera, emitter and receiver accord with the cord with the physical modelsphysical models in real world (for example: in real world (for example: device size, measurement constraints or limits). device size, measurement constraints or limits).
In the In the random walkrandom walk algorithm, extend the behaviors that algorithm, extend the behaviors that maximize the distancemaximize the distance between adjacent robots so that between adjacent robots so that the robots have more chances to find the emergencies.the robots have more chances to find the emergencies.
Fault tolerance or uncertainty problemsFault tolerance or uncertainty problems such as commu such as communication lost, sensing noise etc. nication lost, sensing noise etc.
Search Search multiple moving targetsmultiple moving targets.. Mission supportMission support :: remainders can change their surveillremainders can change their surveill
ance area to compensate the lacking of committing robance area to compensate the lacking of committing robots.ots.
ReferenceReference
[Acar et. al. 2003][Acar et. al. 2003]
Ercan U. Acar, Howie Choset, Yangang Zhang, and MErcan U. Acar, Howie Choset, Yangang Zhang, and Mark Schervish, “Path Planning for Robotic Demining: Rark Schervish, “Path Planning for Robotic Demining: Robust Sensor-based Coverage of Unstructured Environobust Sensor-based Coverage of Unstructured Environments and Probabilistic Methods “, The International Joments and Probabilistic Methods “, The International Jounal of Robotics Research, Vol. 22, No. 7-8, July-Auguunal of Robotics Research, Vol. 22, No. 7-8, July-August 2003, pp. 441-466. st 2003, pp. 441-466.