In order to improve the autonomous reconnaissance efficiency of unmanned aerial vehicle (UAV) in an uncertain environment, situation and observation information acquired by UAV are input into the replay buffer. Model-free training is performed on the data of the replay buffer by deep reinforcement learning (DRL) method, so as to generate the corresponding network model. The reward function is designed for UAV regional reconnaissance missions to further improve the generalization ability of the model. The simulation results show that the UAV autonomous reconnaissance route planning algorithm based on DRL has a high degree of sustainable coverage and its patrol path is unpredictable.
UAV Autonomous Reconnaissance Route Planning Based on Deep Reinforcement Learning
2019-10-01
2053713 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch