Four-wheeled Ackermann Robot Path Planning in 2D Maze Based on Q-Learning-Bezier Method
Zichen Zhang · 2023
Path planning methods have been increasingly utilized in autonomous driving. However, traditional path planning methods, such as Q learning, is not suitable for wheeled vehicles. In this paper, a Q-learning-Bezier method (QLB) is proposed to address the path-planning problem of four-wheeled Ackermann robot by combining Q learning with Bezier curve to produce smooth trajectories that can be passed by Ackermann robot. Meanwhile, two methods with different numbers of optional actions, QLB4a and QLB8a are compared and constructed separately. Experiments show that both proposed methods can generate paths that can reach the end point in a random maze environment. Simulation results show that the Ackermann chassis robot can complete the whole path successfully, which reflects the effectiveness of the algorithm in this paper.