A path planning algorithm based on RRT and SARSA (λ) in unknown and complex conditions
ZOU Qijie, Zhang Yue, Shihui Liu · 2020
The path planning problem of wheeled mobile robot is different from the traditional method in the environment of outer space which is completely unknown. If robots are needed to explore unknown environments while adapting to terrain changes, it is a more challenging problem. In this paper, a Rapidly-exploring Random Tree(RRT) path planning algorithm based on reinforcement learning SARSA(λ) optimization(RL-RRT) is proposed to solve the local path planning problem in unknown and complex conditions. RL-RRT method improves the performance of RRT algorithm by adding optimization for selection of extension points, introducing the idea of biased goal, using task return function, target distance function and angle restriction, ensuring the randomness of RRT, reducing the number of invalid nodes and optimizing the planning performance. A simulation platform is built under Robot Operating System(ROS) and MATLAB to test the role of multi-objective optimization in path planning. The simulation results show that the RL-RRT method enables the wheeled mobile robot to reach the target node smoothly and steadily in complex and unknown environments without collision with obstacles, which verifies the reliability and effectiveness of the method.