Research on Path Planning of Cloud Robot in Dynamic Environment Based on Improved DDPG Algorithm

Abstract

Aiming at the problems of low success rate and slow learning speed of DDPG algorithm in dynamic environment path planning, an improved DDPG algorithm is designed. In this paper, radam algorithm is used to replace the neural network optimization algorithm of ddpg algorithm, and priority experience replay is added to improve the success rate and convergence speed. Then introduces transfer learning enhancement training based on the improved algorithm. In order to solve the problem of limited storage space and insufficient local computing ability caused by the increased complexity of the improved algorithm in the path planning of traditional mobile robots in complex dynamic environment, a cloud robot path planning system is proposed. Among them, robots are responsible for transmitting environment and motion information to the local cloud, and the cloud is responsible for complex computing services. Through ROS robot operating system and Gazebo simulation software, the simulation environment is established, and the contrast experiment between improved DDPG algorithm and DDPG algorithm is carried out. The simulation experiment of dynamic environment path planning is carried out in the cloud robot path planning system. The results show that the improved DDPG algorithm improves the convergence speed by 18% and the success rate by 35% compared with the original DDPG algorithm. Compared with the original system, the cloud robot path planning system effectively saves 37.2% CPU utilization and 23.5% memory utilization. For cloud robot with continuous action space, it has a good effect in dynamic environment path planning.