DRNN-PID趋近律的在线丢包补偿伺服电机NCS滑模控制

考虑存在马尔科夫时延和随机丢包的多包传输直流伺服电机网络控制系统(networded control system,NCS),文章设计了一种基于组合核函数最小二乘支持向量机(least squares support vector machine,LS-SVM)丢包在线预估的对角回归神经网络(Diagonal Recurrent Neural Network,DRNN)PID趋近律滑模控制器。首先,利用无时延等价变换与滑动窗口组合核函数LS-SVM丢包在线预估建立了系统时延补偿模型;其次,通过DRNN非线性映射实现对PID趋近律参数的在线调整,并分析了此种滑模控制器的稳定性;最后利用Truetime对进行仿真,结果表明组合核函数LS-SVM预估策略可以提升丢包补偿的精度,提出的DRNN参数自适应调整PID趋近律滑模控制能够在较快响应速度的条件下减小系统抖振,提升了控制效果。

NCS Sliding Mode Control of Online Packet Loss Compensation Servo Motor Based on DRNN-PID Reaching Law

Considering multiple packet transmission DC servo motor network control system with Markov delay and random packet loss, a kind of diagonal regression neural network PID reaching law sliding mode control based on combined kernel least squares support vector machine on-line prediction of packet loss is designed. Firstly, a delay compensation model of the system is established by using LS-SVM on-line packet loss prediction based on the combination of equivalent transformation without delay and sliding window. Then, the parameters of PID reaching law are adjusted online by DRNN nonlinear mapping, and the stability of the sliding mode controller is analyzed. Finally, the simulation results with Truetime show that the combined kernel LS-SVM prediction strategy can improve the accuracy of packet loss compensation. The DRNN parameter adaptive adjustment PID reaching law sliding mode control can reduce the system chattering and improve the control effect under the condition of faster response speed.