Path planning of mobile robot based on deep reinforcement learning with transfer learning strategy

Abstract

Under complex environments, mobile robots can decision-making, autonomous learning, intelligent obstacle avoidance, and complete the task from start point to endpoint. This paper designed the mobile robot, excluding planners and unknown maps, which can successfully reach the target by autonomously learning and navigating in the unknown environment. By applying deep reinforcement learning to the path planning of mobile robots, the robot can collect data and conduct training on its own, and improve it autonomously without manual supervision. Consequently, it can complete the path planning task. The application of transfer learning improves the adaptive efficiency of the mobile robot to the environment. Finally, the results are verified by comparative experiments in three simulation environments.