Path planning for mobile robot based on improved reinforcement learning algorithm

With the application and development of mobile robot in various fields, a higher demand is required for the path planning ability of mobile robot. To improve the convergence rate of the standard Q-learning algorithm and the smoothness of path planned by the standard Q-learning algorithm, an improved Q-learning algorithm is proposed. First, the attractive potential field in artificial potential field (APF) is used to initialize the Q value. Then, the motion direction of the mobile robot is adjusted, the step size of actions is increased, and the direction factor is added in the state set to improve the accuracy of the route planning. Finally, the proposed algorithm is simulated and verified in the grid map. Simulation results show that, com- pared with the standard Q-learning algorithm, the planning time of improved Q-learning algorithm is reduced by 91%, and the smoothness of path planned by the improved Q-learning algorithm is increased by 79%.