Multiple Input Multiple Output Model Free Adaptive Iterative Learning Control of Two-dimensional Linear Motor Based on Neural Network

Abstract

The two-dimensional linear motor control system has the characteristics of nonlinear, multi-variable and strong coupling. In the actual operation process, it is affected by load disturbance, mechanical delay, friction resistance and other factors, and the dynamics model is difficult to accurately obtain, so its tracking control is extremely hard. Radial Basis Function (RBF) based neural network has the advantage of arbitrary approximation to nonlinear function and Model Free Adaptive Iterative Learning Control (MFAILC) does not depend on the characteristics of accurate mathematical model of the controlled system and the rule of sequential learning. A multi-input multi-output model-free adaptive iterative learning control(MIMO-MFAILC) strategy based on RBF neural network is proposed. RBF neural network is used to learn pseudo partial derivative (PPD) online in model-free adaptive iterative learning control. As feedforward compensation, the iterative learning control can overcome external interference, compensate system nonlinearity, and complement feedforward and feedback advantages, so as to realize the precision compensation of expected tracking and further improve position tracking accuracy. Finally, the accuracy and effectiveness of the proposed strategy are verified experimentally by combining RT-SIM simulation platform with 2d linear motor control system under no-load and load conditions.