UAV Obstacle Avoidance Method Based on Velocity Obstacle-Proximal Policy Optimization
Focusing on the problem of obstacle avoidance of UAV in complex obstacle environment,an ob-stacle avoidance strategy based on VO-PPO is proposed.Firstly,the state space is constructed according to the UAV's own information and the obstacle information described by the velocity obstacle method,and the reward and punishment function based on VO region including speed and distance is designed.Sec-ondly,an algorithm framework of Proximal Policy Optimization is designed under the Actor-Critic network structure,through the interaction of agent and environment,network parameters under the condition of maximum reward are trained,and obstacle avoidance of UAV in different environments is realized.Final-ly,compared with other algorithms in simulation experiments,the proposed algorithm is proved to have great generalization and effectiveness.
deep reinforcement learningUAVobstacle avoidingcomplex environment
万方数据
维普
