基于深度强化学习的近程无人机数据采集方法研究  

Research on Data Acquisition Method of Short-range UAV Based on Deep Reinforcement Learning

作　　者：温积群 杨迁 王绍荃 赵勇 WEN Ji-qun;YANG Qian;WANG Shao-quan;ZHAO Yong(Zhejiang Tusheng Power Transmission Engineering Co.,Ltd.,Wenzhou Technology Branch,Wenzhou 325000 China)

机构地区：[1]浙江图盛输变电工程有限公司温州科技分公司,浙江温州325000

出　　处：《自动化技术与应用》2023年第10期121-124,共4页Techniques of Automation and Applications

摘　　要：因近程无人机飞行反应时间较短,出现数据采集功能智能化程度较低的问题,对基于深度强化学习的近程无人机数据采集方法进行研究。采用深度强化学习算法处理采集时的高维环境状态,通过深度学习与强化学习的拟合能力和决策能力处理采集过程,结合独立目标网络和经验回放的深度强化学习算法设计的软件程序,通过上位机软件AHRS实施设计方法的实例验证。实验结果表明方法在启动而未飞行时,以及在匀速运动时的采集误差都很低,采集功能较好。Due to the short flight response time of short-range UAV and the low intelligence of data acquisition function,the data acquisition method of short-range UAV Based on deep reinforcement learning is studied.The deep reinforcement learning algorithm is used to deal with the high-dimensional environment state during acquisition.The acquisition process is processed through the fitting ability and decision-making ability of deep learning and reinforcement learning.The software program designed by the deep reinforcement learning algorithm combined with independent target network and experience playback is verified by the example of the design method implemented by the host computer software AHRS.The experimental results show that the acquisition error of the design method is very low when starting but not flying,and when moving at a uniform speed,and the acquisition function is good.

关 键 词：深度强化学习 独立目标网络 近程无人机 运动处理传感器 数据采集 

分 类 号：TP274.2[自动化与计算机技术—检测技术与自动化装置]