Autonomous Vehicles - IIT Bombay

Transcription

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionAutonomous VehiclesDipak ChaudhariSriram Kashyap M SIndian Institute of Technology, Bombay2008Dipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission ojects3Basic Design4Perception5Motion Planning6Mission Planning7Behaviour8ConclusionDipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionIntroductionUnmanned Vehicles:No driver on-board the vehicleTeleoperatedDriven by an operator viewing video feedbackToy remote control carAutonomousDriven by on-board computers using sensor feedback andautomatic controlsUsage:Dangerous tasksRepetitive tasksDirty tasksDipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission us under-water robot to explore water-filled sink holes inMexico. The image shows a 318 meter deep sink hole.Source: IEEE Spectrum, Sep-2007Dipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionExamplesMars Rover by NASASource: http://marsrover.nasa.gov/Dipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionExamplesStanley: The Stanford autonomous carSource:Thrun et al. “Stanley: The robot that won the DARPA Grand Challenge”Dipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission /www.ivtt.org/IVTTDipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionMotivationTravelling by car is currently one of the most deadly forms oftransportation, with over a million deaths annually worldwideAs nearly all car crashes (particularly fatal ones) are caused byhuman driver error, driverless cars would effectively eliminatenearly all hazards associated with driving as well as driverfatalities and injuriesDipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionContests and ProgrammesEUREKA Prometheus Project (1987-1995)ARGO Project, Italy (2001)DARPA Grand Challenge (2004-2007)European Land-Robot Trial (2006-2008)Dipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionEUREKA Prometheus ProjectVaMP and VITA-2 vehicles (1994)1000 km on a Paris multi-lane highway in heavy traffic at upto 130 km/hAutonomous convoy driving, vehicle tracking, lane changes,passing of other carsAutonomous Mercedes S-Class in 19951000 km on the German Autobahn at 175 km/hNot 100% autonomous. A human safety pilot was presentCar drove upto 158 km without interventionDipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionDARPA Grand ChallengeUS Department of Defense conducts the autonomous vehiclechallenge2004: Mojave Desert, United States, along a 150-mile track2005: 132 mile off-road course in Nevada2007: ’Urban Challenge’ at George Air Force BaseDipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionTypical Challenges to meetNavigate desert, flat and mountainous terrainHandle obstacles like bridges, underpasses, debris, potholesand other vehiclesObey traffic lawsSafe entry into traffic flow and passage through busyintersectionsFollowing and overtaking of moving vehiclesDrive an alternate route when the primary route is blockedCorrect parking lot behaviourMost important rule: No CollisionsDipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionDARPA 2005 TrackSource:Google Videos: The Car That Won The DARPA Grand Challenge: 2006”Dipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionDARPA 2007 TrackSource:DARPA Urban Challenge Participants Conference PresentationDipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionWhat should an autonomous vehicle do?Understand its immediate environment (Perception)Find its way around obstacles and in traffic (Motion planning)Know where it is and where it wants to go (Navigation)Take decisions based on current situation (Behaviour)Dipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionArchitecture: Junior (Stanford)Source:Thrun et al. “Junior: The Stanford Entry in the Urban Challenge”Dipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionArchitecture: Boss (CMU)Source: Urmson et al. “Autonomous Driving in Urban Environments: Boss and the Urban Challenge”Dipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionPerceptionLIDAR (Light Detection and Ranging)RADARVisionGPSInertial navigation systemDipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionSensors on Stanley, The Stanford CarSource:Thrun et al. “Stanley: The robot that won the DARPA Grand Challenge”Dipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionLIDARSource:Thrun et al. “Junior: The Stanford Entry in the Urban Challenge”Dipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionLIDAR for Obstacle DetectionLong range scanner has several lasers, each with a scanningringCompare radius of adjacent rings to identify height of objectsUse multiple short range LIDARs to cover blind spotsGenerate a point cloud based on LIDAR dataApply thresholds to this data to eliminate overhanging andlow objectsDipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionHandling OcclusionObjects may not be always visibleIntegrate range data over time, to keep track of objects thatmay be temporarily occludedDipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionHandling OcclusionObjects may not be always visibleIntegrate range data over time, to keep track of objects thatmay be temporarily occludedWhat about Moving objects?Dipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionHandling OcclusionObjects may not be always visibleIntegrate range data over time, to keep track of objects thatmay be temporarily occludedWhat about Moving objects?Integrate data only in those regions that are currently occludedDipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionObstacle Detection in actionSource:Thrun et al. “Junior: The Stanford Entry in the Urban Challenge”Dipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionObject TrackingIdentify and label distinct moving objectsObtain information about these objects, such as size, headingand velocityContinue to track these objects (even when they are occluded)Dipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionObject TrackingSource:Thrun et al. “Junior: The Stanford Entry in the Urban Challenge”Dipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionObject Tracking: DetailsIdentify areas of changeInitializes a set of particles as possible object hypothesesThese particles implement rectangular objects of differentdimensions, and at slightly different velocities and locationsA particle filter algorithm is then used to track such movingobjects over timeDipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionFurther Challenges in PerceptionWhat is a road?Self LocalizationBad/Noisy dataSensor failure (ex: GPS outage)Setting ’good’ thresholdsDipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionMotion PlanningMotion planning involves performing low level operations towardsachieveing some high level goalPath Variables:Steering (direction)SpeedPlanning:Vary these parameters and generate multiple local paths thatcan be followedAssign costs to paths based on time taken, distance fromobstacles, and other constraintsChoose the best path from the various possible pathsDipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionVarying directionDirection is varied by tracing possible paths from current positionto a set of (temporary) local goals. These goals are slightly spreadout so as to be able to navigate around obstacles. Paths of greaterlength, paths that are near obstacles incur higher cost.Source:Thrun et al. “Junior: The Stanford Entry in the Urban Challenge”Dipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionMission PlanningGlobal Path PlanningDARPA Urban Challenge: input filesRoute Network Definition File (RNDF)Mission Data File ( MDF )Dipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionRoad SegmentSource:DARPA Urban Challenge Participants Conference PresentationDipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionStop LinesSource:DARPA Urban Challenge Participants Conference PresentationDipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionZonesSource:DARPA Urban Challenge Participants Conference PresentationDipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionConnectivity GraphConnectivity GraphEdges are assigned costs based onExpected time to traverse the edgeDistance of the edgeComplexity of the corresponding area of the environmentValue functionPath from each way point to the current goalIncorporating newly observed informationDipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionBlockage DetectionStatic obstacle mapSpurious BlockagesEfficient, optimistic algorithm: Some blockages are notdetectedVirtual BlockageExtent of the blockage along affected lanesDipak Chaudhari Sriram Kashyap M SAutonomous VehiclesIndian Institute of Technology, Bombay

IntroductionProjectsBasic DesignPerceptionMotion PlanningMission PlanningBehaviourConclusionRevisiting BlockagesRevisiting of previously detected blockagesThe cost c increment added by a blockage is decayed exponentiallysc p2 a/h wherea is the time since the blockage was last observed,h is a half-life parameter,p is the starting cost penalty increment for blockagesCost ThresholdAvoiding too frequent visits to a blockage:Increment h for the blockage after each new visit would makethe traversal costs decay more slowly each time the obsta

Travelling by car is currently one of the most deadly forms of transportation, with over a million deaths annually worldwide As nearly all car crashes (particularly fatal ones) are caused by human driver error, driverless cars would effectively eliminate nearly all hazards associated with driving as well as driver fatalities and injuries