Path Planning for Mobile Robot's Continuous Action Space Based on Deep Reinforcement Learning

Abstract

In this paper, a path planning method based on deep reinforcement learning in the unknown environment of mobile robots is proposed, in order to meet the path planning of the kinematic model and constraint conditions of the mobile robot under continuous action space. The path planning method plans a path from the starting point to the target point to avoid obstacles, but these planned paths do not meet the kinematic model of the mobile robot and cannot be directly applied to the actual mobile robot control. The proposed approach based on depth reinforcement learning path planning meets the motion model and constraints of mobile robots. The optimal strategy is found in the continuous action space, and the optimal path is obtained through the evaluation criteria. This path is obtained by using the mobile robot motion model, so the movement configuration of the mobile robot can be solved directly. The experimental results show that a mobile robot motion model can be used to plan a collision free optimal path in the unknown environment, and this path is also the actual running track of the mobile robot.