Auto Disturbance Rejection Control Based on A Neural Network Predictive State Observer

This paper addresses the auto disturbance rejection control (ADRC) for systems with time delay. ADRC is an effective control for most linear or nonlinear systems with noises and disturbances, because of its advantage on disturbance estimation and compensation. The extended state observer (ESO) is a critical component of ADRC. However, this observer is usually designed for systems without time delay. The time delay in a system could lead the ADRC unstable when the phase delay of the response cannot be ignored. This paper proposes a neural network based predictive state observer (NNPSO) to predict the response of systems with time delay. The simulation results validate the proposed method when systems with time delay are controlled.