Probabilistic Path Planning for Wheel-Legged Rover in Dense Environment Based on Extended MDP and Configuration Topology Analysis

Abstract

Wheel-legged planetary rovers possess superb locomotion capabilities. This article combines an offline predefined motion planning library with online path planning, integrating energy consumption and probabilistic aspects of the robotic system. The primary focus is on addressing the planning challenges in dense environments, where the distance between any adjacent obstacles is smaller than the width of the prototype. Therefore, it is necessary to consider the interaction between the prototype and the environment. First, the generalized function set theory and the configuration topology theory are utilized to mathematically describe the motions of multilimbed systems. Based on the representation, an offline planning library is established. Second, the Markov-decision-process-based path planning method is extended by incorporating the platform's geometry and locomotion capabilities. The concept of “limb-travel relevant nodes” is introduced. To address the numerous iteration problems, the informed value iteration algorithm is proposed. Third, a multilayered map is evaluated to further enhance computational efficiency. Finally, the proposed algorithm is implemented on the terrain adaptive wheel-legged rover. Experimental results demonstrate that the proposed algorithm is capable of finding the optimal path with high computational efficiency, and it exhibits excellent adaptability on nonuniform maps.