Scarab Autonomous Traverse Carnegie Mellon December 2007 David Wettergreen
Carnegie Mellon | 13 December Mission Scenario Land in crater –Direct to floor, no crater wall descent –Minimal lander Communicate by polar orbiter relay Power from isotope source, no solar Navigate in darkness –Active sensing Operate with supervised autonomy Survey multiple locations –Characterize regolith composition and physical properties –Determine nature and abundance of hydrogen Survive 7 months –25 drill sites x (5 days/site, 3 days/traverse) = 200 days Mass kg
Carnegie Mellon | 13 December Rover Capability Kilometer-Scale Traverse –Terrain modeling for obstacle detection –Path planning for obstacle avoidance –Position estimation for path tracking Resource Regulation –Power –Thermal Health Monitoring –Fault Detection and Recovery –Contingent Plan Execution
Carnegie Mellon | 13 December Rover Architecture Health Monitor Rover Executive Vehicle Controller Mission Planner Far-field Evaluator Images Odometry Rover Interface Stop Navigator Curve & Speed State Observer State Instrument Controllers Near-field Detector Position Estimator State Telemetry Manager Inertial & Odometry ScansCommands Waypoints Telemetry Position State (All) FaultsPlans EvaluationEvalActions Specification Science Observer Instrument Manager Goal Manager Viewpoints Images Science Planner FeaturesGoals
Carnegie Mellon | 13 December Autonomous Traverse Total daily traverse exceeding 10km is achievable Demonstrated averages m per command cycle
Carnegie Mellon | 13 December Rover Architecture Health Monitor Rover Executive Vehicle Controller Mission Planner Far-field Evaluator Images Odometry Rover Interface Stop Navigator Curve & Speed State Observer State Instrument Controllers Near-field Detector Position Estimator State Telemetry Manager Inertial & Odometry ScansCommands Waypoints Telemetry Position State (All) FaultsPlans EvaluationEvalActions Specification Science Observer Instrument Manager Goal Manager Viewpoints Images Science Planner FeaturesGoals
Carnegie Mellon | 13 December Terrain Model Aggregate Geometric Model Traversability Analysis Persistence
Carnegie Mellon | 13 December Sensor Views
Laser Light Striping
Carnegie Mellon | 13 December Path Execution Navigator evaluates –near-term driving options while –guiding the rover to its long-term goal Many possible actions considered each sensing cycle Terrain model accumulates
Carnegie Mellon | 13 December Position Estimation Challenges –Skidding - wheel odometry inaccurate –Kinematics - vehicle model more complicated due to changing wheelbase and Approach –No wheel odometry –Optimal (Kalman) filtering –Inertial and optical sensing
Carnegie Mellon | 13 December Position Estimation Inertial Measurement Unit –3-axis rotation (Honeywell HG1700 ring laser gyro) –3-axis acceleration Global Positioning System –Omnistar differential (0.1m accuracy) for ground-truth position –Can be used to correct gyro drift Kalman Filter –Integrates inertial with other motion sensing
Carnegie Mellon | 13 December Position Estimation Optical velocity sensor Operates with ground lighting
Carnegie Mellon | 13 December Dark Traverse Demonstration Developments for Scarab –Integrated Laser Scanning –Integrated Inertial/Optical Position Estimation –Developed Vehicle Controller –Integrated Scarab Motion Planning Tonight –Goal is 1-kilometer of autonomous traverse in darkness Tomorrow –Outcome –Evaluation