A Collision-Free Fast Planning Algorithm for Snake-Like Redundant Manipulators via Reinforcement Learning

This paper presents a planning algorithm via reinforcement learning to fast generate a collision-free path in the narrow and complex space for snake-like redundant manipulators. First, the snake-like redundant manipulator is introduced and its kinematic is analyzed. Second, the algorithm strategy is designed by combining the snake-inspired crawling motion and reinforcement learning. Finally, the simulations are conducted to test and verify the proposed algorithm. As a result, the snake-like redundant manipulator can explore the target position in the pipeline with a 93.33% success rate at an average planning time cost of 49.2s.