Adaptive robust control of a class of motor servo system with dead zone based on neural network and extended state observer

Abstract

In the actual motor servo system, dead zone, saturation, and hysteresis are some of the most common nonlinear characteristics. Among them, the dead zone is the most serious to the system performance. This article focuses on the nonlinearity of the dead zone, and uses back-propagation neural network to smoothly and continuously compensate the dead zone. Considering that larger disturbances will slow down the convergence of the neural network and become easier to diverge, it is necessary to use extended state observer to share part of the disturbance observations. Based on this, a robust adaptive controller is designed for a class of motor system with torque control to achieve high-precision motion control. Using feedforward cancelation technology, extended state observer and back-propagation neural network are combined with robust adaptive controller to achieve high-performance control of motor system. By using Lyapunov theorem, the adaptive laws of parameters and weights of neural networks are derived. The global robustness of the control strategy is guaranteed by the proper feedback robust law. In addition, the controller guarantees the tracking performance under various uncertainties in theory, which is of great significance to the high-precision control of the motion system. The high performance of the control strategy is verified by simulation and experiment.