摘要
由于机械臂作业环境具有高度的非线性、时变性和不确定性,导致机械臂动态行为较为复杂,传统控制方法在进行控制过程中经常出现关节位置角度误差大、稳定性差的问题,提出基于改进深度确定性策略梯度(DDPG)的控制方法。首先,建立智能装卸机械设备运动学模型;其次,基于DDPG算法建立装卸动作控制模型;再次,利用演员(Actor)网络(策略网络)和评论者(Critic)网络(价值网络)来改进DDPG中的装卸动作控制过程;最后,优化奖励函数,对Actor网络和Critic网络进行训练,实现装卸机械设备控制。实验结果表明,智能装卸机械设备实际运动轨迹与期望轨迹非常重合和接近,关节位置误差始终低于5.0 cm,姿态角误差始终低于1.00°,能够对关节位置和姿态角进行有效控制,所提改进DDPG算法的控制效果较好。
Due to the highly nonlinear,time varying,and uncertain working environment of robotic arms,their dynamic behavior is complex.Traditional control methods often encounter problems such as large joint position angle errors and poor stability during the control process.Therefore,a control method based on the improved Deep Deterministic Policy Gradient(DDPG)is proposed.Firstly,a kinematic model of intelligent loading and unloading machinery equipment is established;secondly,a loading and unloading action control model based on the DDPG algorithm is established;thirdly,u the Actor network(strategy network)and Critic network(value network)are used to improve the loading and unloading action control process in DDPG;finally,the reward function is optimized to train the Actor network and Critic network.and achieve loading and unloading machinery equipment control.The experimental results show that the actual motion trajectory of intelligent loading and unloading machinery equipment is very close to the expected trajectory,and the joint position error is always less than 5.0 cm,and the attitude angle error is always less than 1.00°.This can effectively control the joint position and attitude angle.The proposed improved DDPG algorithm has a good control effect.
作者
龚宇平
李金瑾
卿柏元
潘学华
GONG Yuping;LI Jinjin;QING Baiyuan;PAN Xuehua(Measurement Center of Guangxi Power Grid Co.,Ltd.,Nanning 530024,China)
出处
《机械与电子》
2024年第12期43-48,共6页
Machinery & Electronics
基金
广西电网公司科技项目(044400KK52230001)。
关键词
改进深度确定性策略梯度算法
智能装卸机械设备
运动学模型
控制方法
improved deep deterministic strategy gradient algorithm
intelligent loading and unloading machinery equipment
kinematic model
control methods
