基于深度强化学习的无人机目标跟踪研究

Study on UAV Target Tracking Based on Deep Reinforcement Learning

基于目标跟踪任务,阐述运用深度学习为我方无人机训练跟踪策略,寻找最适于无人机对目标跟踪的深度学习算法,建立了目标跟踪模型,并运用四种算法进行训练,比较了在线训练的指标以及离线执行时的结果,从而在双决斗深度Q学习和近端策略优化中取得了最好的训练效果。

Based on the target tracking task,this paper expounds the use of deep learning as our UAV training and tracking strategy,finds the most suitable deep learning algorithm for UAV target tracking,establishes the target tracking model,uses four algorithms for training,and compares the indicators of online training and the results of offline execution.The best training effect is achieved in double duel deep Q learning and near end strategy optimization.