Improved Sliding Mode Output Feedback Control with Adaptive State Observer Based on Neural Network

Electromechanical actuation systems are widely used in industrial production and military fields. However, the model uncertainty of electromechanical actuation systems will reduce the accuracy of model-based nonlinear controllers. At the same time, due to the limitation of installation space and cost, the system is often unable to install speed sensors. To solve these problems, an improved sliding mode output feedback control strategy based on neural network adaptive state observer is proposed. Aiming at the constant disturbance and parameter uncertainty in the system, the parameter adaptive rate is designed. Aiming at the time-varying disturbance in the system, the universal approximation property of Radial Basis Function (RBF) neural network is used to estimate, and the former is compensated by feedforward compensation technology. At the same time, the adaptive state observer based on neural network is used to observe the system speed value and design the control quantity, so as to realize the output feedback control. The nonlinear terms of the neural network and the observer observation residual are compensated by the observation error compensation term. By using Lyapunov stability theorem, it is proved that the designed controller can realize the asymptotic stability of the system. A large number of simulation results show that the designed controller can improve the control accuracy by an order of magnitude compared with the traditional proportional integral differential (PID) control, and the control accuracy, observation accuracy and anti-interference performance are significantly improved compared with the sliding mode controller based on other observers.