Deep Reinforcement Learning Integrated RRT Algorithm for Path Planning

Rapidly-exploring random tree (RRT) algorithm, featured with strong exploration capability, is widely used in path planning tasks. However, it is difficult for RRT to find the optimal path due to its inherent characteristics of random sampling. In this paper, we propose an improved RRT algorithm integrated with deep reinforcement learning (IRRT-DRL), which can effectively search for feasible paths and exploit previous experience for optimization. It includes a unique reward function and a dynamic waypoint selection mechanism that automatically adjusts the interval between adjacent waypoints to help the agent bypass obstacles. Experimental results have verified the feasibility and superiority of the proposed approach.