Probabilistically Safe Avoidance of Dynamic Obstacles with Uncertain Motion Patterns

摘要: This paper presents a real-time path planning algorithm which can guarantee probabilistic feasibility for autonomous robots subject to process noise and an uncertain environment, including dynamic obstacles with motion patterns. The key contribution of the work is integration novel method modeling future trajectories. method, denoted as RR-GP, uses learned pattern model make long-term predictions their paths. done by combining flexibility Gaussian processes (GP) efficiency RRT-Reach, sampling-based reachability computation ensures feasibility. prediction then utilized within chance-constrained rapidly-exploring random trees (CC-RRT), chance constraints explicitly achieve constraint satisfaction while maintaining computational benefits samplingbased algorithms. With RR-GP embedded in CC-RRT framework, theoretical guarantees be demonstrated linear systems uncertainty, though extension nonlinear also considered. Simulation results show that resulting approach used realtime efficiently accurately execute safe