Path planning of power line inspection based on DDPG for obstacle avoidance with UAV

Obstacle avoidance is an important application in the path planning for Unmanned Aerial Vehicle power line inspections. Only by avoiding obstacles during the inspection process can the UAV complete the inspection task well. After analysis we propose to use reinforcement learning algorithms to solve the obstacle avoidance problem in UAV inspection. We use a depth-deterministic policy gradient algorithm to implement obstacle avoidance for UAV inspection and combine it with an artificial potential field method to speed up training. We validated the DDPG algorithm in a simulated environment with simulated obstacles, and the UAV was able to avoid the obstacles well and reach the target point.