Motion Path Planning of Agent Based on Proximal Policy Optimization Algorithm

Aiming at the problems of insufficient intelligence level, computational complexity, and lack of autonomous learning capability of traditional path planning algorithms, this paper proposes a solution for implementing the motion path planning of agent using proximal policy optimization algorithm based on deep reinforcement learning. The solution can provide a universal method for the path planning of agent moving in a two-dimensional plane, which splits the task into two subtasks: path generation and path tracking, where the agent generates a feasible path based on proximal policy optimization algorithm, and the path tracking task can be implemented by a traditional control algorithm. Finally, the proposed algorithm is simulated and verified in training and testing environments. The results show that the proposed algorithm can effectively complete the path generation task, and the proximal policy optimization algorithm has advantages over the A* algorithm in terms of intelligence level and environmental adaptability. Furthermore, the algorithm is generalizable to all movable objects.