Sampling-Based Real-Time Motion Planning under State Uncertainty for Autonomous Micro-Aerial Vehicles in GPS-Denied Environments

摘要: This paper presents a real-time motion planning approach for autonomous vehicles with complex dynamics and state uncertainty. The is motivated by the problem navigating in GPS-denied dynamic environments, which involves non-linear and/or non-holonomic vehicle dynamics, incomplete estimates, constraints imposed uncertain cluttered environments. To address above problem, we propose an extension of closed-loop rapid belief trees, random trees (CL-RBT), incorporates predictions position estimation uncertainty, using factored form covariance provided Kalman filter-based estimator. proposed planner operates incrementally constructing tree dynamically feasible trajectories prediction, while selecting candidate paths low uncertainty efficient update propagation. algorithm can operate real-time, continuously providing controller execution, enabling to account Simulation results demonstrate that generate reduce handling environment constraints.