Discrete Active Disturbance Rejection Iterative Learning Control Based on Dynamic Linearization

A discrete active disturbance rejection iterative learning control method based on dynamic linearization is proposed for a class of discrete-time, nonlinear and non-affine system that run repeatedly within a finite time. The controlled system is dynamically linearized into an affine form related to the control input within the iteration domain. The control gain is initialized through the pseudo partial derivative of dynamic linearization model when needed and then fixed. The estimated errors of parameter, system uncertainty and external disturbance are compacted into a nonlinear term as the total disturbance of the system. Via iterative sliding mode scheme, the iterative extended state observer is designed to estimate the total disturbance and a discrete active disturbance rejection iterative learning control law is proposed. The convergence of the iterative extended state observer and tracking errors of the system are analyzed. The proposed method is a new intuitive and concise data-driven control method which does not need the system model information. The effectiveness of the proposed method is verified by simulations.