Unmanned aerial vehicle attitude balance control method based on deep reinforcement learning

• Main Authors: XU MINGHUAN, WU ZHANXIONG, CHEN XUANHENG, LIU HEHE, YU JIANGNAN
• Format: Patent
• Language: Chinese; English
• Published: 05.12.2023
• Subjects: CONTROLLING; PHYSICS; REGULATING; SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES

The invention discloses an unmanned aerial vehicle attitude balance control method based on deep reinforcement learning. The method comprises the following steps: S1, filling m pieces of initial empirical data collected in advance into an empirical pool; s2, initializing parameters of the deep reinforcement learning network; s3, pre-training the initialized deep reinforcement learning network to obtain a pre-training weight; S4, training the deep reinforcement learning network online in real time; and S5, the deep reinforcement learning network outputs a control quantity to control the unmanned aerial vehicle. The method is used for controlling and correcting the attitude of the unmanned aerial vehicle in a complex environment, and the safety and practicability of the unmanned aerial vehicle are improved. 本发明公开了一种基于深度强化学习的无人机姿态平衡控制方法，包括如下步骤：S1、将预先采集的m个初始经验数据并装填入经验池；S2、初始化深度强化学习网络的参数；S3、预训练初始化后的深度强化学习网络，获得预训练权重S4、在线实时训练深度强化学习网络；S5、深度强化学习网络输出控制量对无人机进行控制。该方法用于在复杂环境下对无人机姿态进行控制和矫正，提高无人机的安全性和实用性。