基于自回归小波神经网络的机械臂自适应滑模控制

Adaptive Sliding Mode Control of Manipulator Based on Self-Recurrent Wavelet Neural Network

针对机械臂存在模型不确定和未知扰动的问题,提出一种动力学模型参数分块逼近的神经网络非奇异终端滑模(nonsingular terminal sliding mode, NTSM)控制方法.为加快系统跟踪误差的收敛速度,避免传统终端滑模存在的奇异性问题,采用一种非奇异终端滑模面.利用多组自回归小波神经网络(self-recurrent wavelet neural network, SRWNN)分块逼近系统未知的动力学模型参数,并采用自适应更新律调整权重.通过积分控制项补偿SRWNN的逼近误差,并使用Lyapunov稳定性理论证明了系统稳定性.使用MATLAB进行仿真分析,分块SRWNN滑模控制与滑模控制、整体SRWNN滑模控制相比,关节角度跟踪误差的平均稳态误差分别降低了31.9%、76.5%,表明此方法是一种可靠、有效的轨迹跟踪控制方法.

A neural network non-singular terminal sliding mode(NTSM)control method with block approximation of dynamics model parameters is proposed to address the problem of model uncertainty and unknown perturbations in the robotic arm.First,a non-singular terminal sliding mode surface is used in order to accelerate the conver-gence of the system tracking error and avoid the problem of singularity in the traditional terminal sliding mode.Sec-ond,multiple self-recurrent wavelet neural network(SRWNN)chunks are utilized to approximate the unknown dy-namics model parameters of the system,and the weights are adjusted by using the adaptive update law.Meanwhile,the approximation error of the SRWNN is compensated by designing a robust control term,and the system stability is proved using Lyapunov stability theory.Finally,the simulation analysis using MATLAB shows that the average steady state error of the joint angle tracking error is reduced by 31.9%and 76.5%for the chunked SRWNN sliding mode control compared with the sliding mode control and the overall SRWNN sliding mode control,respectively,which demonstrates that this method is a reliable and effective trajectory tracking control method.