机械臂外力的时延强跟踪Kalman估计与自学习控制
Time Delay Strong Tracking Kalman Estimation and Self-Learning Control of Manipulator Contact Force
淡乾川 1崔凤坤2
作者信息
- 1. 重庆科创职业学院智能制造学院,重庆 402160
- 2. 山东交通学院,山东 济南 250357
- 折叠
摘要
为了提高机械臂与环境交互过程中的接触力估计精度和关节控制精度,提出了基于时延强跟踪Kalman滤波的接触力估计方法和基于熵-自学习控制网络的关节角位置控制方法.建立了机械臂系统的关节驱动模型和动力学模型;基于时延估计理论,建立了4自由度机械臂系统的时延模型;在Kalman滤波中引入了渐消因子,使估计残差强行正交,从而提出了基于时延强跟踪Kalman的接触力估计方法.以接触力误差和关节位置误差为输入,以关节力矩为输出构建了自学习控制网络,并提出使用熵聚类确定网络结构和参数,从而设计了熵-自学习控制网络.经仿真验证,基于时延强跟踪Kalman的估计误差区间分布小于标准Kalman和文献[12]柔顺控制,且分布密度在0误差处极高;熵-自学习控制网络对期望轨迹的绝对跟踪误差和收敛时间远小于自学习控制网络和变阻抗阻尼控制.仿真结果验证了本文接触力估计方法和角位置控制方法的优越性.
Abstract
In order to improve the accuracy of contact force estimation and joint control in the interaction between manipulator and environment,a contact force estimation method based on time-delay strong tracking Kalman filter and a joint angle position control method based on entropy self-learning control network are proposed.The joint driving model and dynamic model of the manipulator system are established.Based on the theory of time delay estimation,the time delay model of 4-DOF manipulator system is established.The fading factor is introduced into the Kalman filter to force the estimation residual to be orthogonal,soa contact force estimation method based on time-delay strong tracking Kalman is proposed.Taking the contact force error and joint position error as the input and joint torque as the output,the self-learning control network is constructed,and the entropy cluster-ing is proposed to determine the network structure and parameters,so as to design the entropy self-learning control network.The simulation results show that the estimation error interval distribution based on time-delay strong tracking Kalman is smaller than that of standard Kalman and reference[12],and the distribution density is very high at the error of 0.The absolute tracking error and convergence time of entropy self-learning control network for the desired trajectory are much smaller than that of self-learning control network and variable impedance damping control.The simulation results verify the superiority of the contact force estimation method and angular position control method in this paper.
关键词
接触力估计/关节控制/强跟踪Kalman滤波/熵-自学习控制/机械臂系统Key words
Contact Force Estimation/Joint Control/Strong Tracking Kalman Filtering/Entropy Self-Learning Control/Manipulator System引用本文复制引用
基金项目
重庆市教委科技项目(KJQN202005401)
出版年
2024
NETL
NSTL
维普
万方数据