Motion Planning under Uncertainty for Autonomous Navigation of Mobile Robots and UAVs

This thesis presents a reliable and efficient motion planning approach based on state lattices for the autonomous navigation of mobile robots and UAVs. The proposal retrieves optimal paths in terms of safety and traversal time, and deals with the kinematic constraints and the motion and sensing uncertainty at planning time. The efficiency is improved by a novel graduated fidelity state lattice which adapts to the obstacles in the map and the maneuverability of the robot, and by a new multi-resolution heuristic which reduces the computational complexity. The motion planner also includes a novel method to reliably estimate the probability of collision of the paths considering the uncertainty in heading and the robot dimensions.