1CS 188: Artificial IntelligenceFall 2011Advanced Applications:Robotics / Vision / LanguageDan Klein – UC BerkeleyMany slides from Sebastian Thrun, Pieter Abbeel, Jitendra Malik1Announcements This week: No sections this week Office hours modified (moved earlier), see Piazza Final contest cutoff is 8pm Wednesday Grades: W1-3, P1-4, Midterm in glookup, please check P5: full credit in grade computations22So Far: Foundational Methods3Now: Advanced Applications43Autonomous VehiclesAutonomous vehicle slides adapted from Sebastian Thrun[DEMO: Race, Short] 150 mile off-road robot race across the Mojave desert Natural and manmade hazards No driver, no remote control No dynamic passing 150 mile off-road robot race across the Mojave desert Natural and manmade hazards No driver, no remote control No dynamic passingGrand Challenge: Barstow, CA, to Primm, NV[DEMO: GC Bad, Good]4An Autonomous Car5 LasersCameraRadarE-stopGPSGPS compass6 ComputersIMUSteering motorControl ScreenActions: Steering ControlErrorSteering Angle(with respect to trajectory)5Sensors: Laser Readings[DEMO: LIDAR]123Readings: No Obstacles6∆ZReadings: ObstaclesRaw Measurements: 12.6% false positivesObstacle DetectionTrigger if |Zi−Zj| > 15cm for nearby zi, zj7xt+2xtxt+1zt+2ztzt+1Probabilistic Error ModelGPSIMUGPSIMUGPSIMUHMM Inference: 0.02% false positivesRaw Measurements: 12.6% false positivesHMMs for Detection8Environmental Tracking[DEMO: PEOPLE]Sensors: Camera9Object RecognitionQueryTemplateVision slides adapted from Jitendra MalikShape ContextCount the number of points inside each bin, e.g.:Count = 4Count = 10... Compact representation of distribution of points relative to each point1810Shape Context19Similar RegionsColor indicates similarity using local descriptors2011Match for Image Similarity21Vision for a Car[DEMO: LIDAR 1]12Self-Supervised Vision[DEMO: LIDAR 2]Complex Robot Control[demo – quad initial]13Robotic Control Tasks Perception / Tracking Where exactly am I? What’s around me? Low-Level Control How to move from position A to position B Safety vs efficiency High-Level Control What are my goals? What are the optimal high-level actions?Low-Level Planning Low-level: move from configuration A to configuration B14A Simple Robot Arm Configuration Space What are the natural coordinates for specifying the robot’s configuration? These are the configuration space coordinates Can’t necessarily control all degrees of freedom directly Work Space What are the natural coordinates for specifying the effector tip’s position? These are the work spacecoordinatesCoordinate Systems Workspace: The world’s (x, y) system Obstacles specified here Configuration space The robot’s state Planning happens here Obstacles can be projected to here15Obstacles in C-Space What / where are the obstacles? Remaining space is free spaceExample: A Less Simple Arm[DEMO]16Probabilistic Roadmaps Idea: sample random points as nodes in a visibility graph This gives probabilistic roadmaps Very successful in practice Lets you add points where you need them If insufficient points, incomplete or weird paths Demonstrate path across the “training terrain” Run apprenticeship learning to find a set of weights w Receive “testing terrain” (a height map) Find a policy for crossing the testing terrain.High-Level Control17High DOF Robots[DEMOS]Videos from Pieter Abbeel, Jean-Claude
View Full Document