Decoupling Control for a Class of Dynamic Nonlinear Coupling Systems via ADRC based on DRNN

Abstract

In the present paper, a decoupling control scheme is designed for a class of dynamic nonlinear coupling systems with time-varying internal uncertainties, external disturbances and coupling effects between different subsystems which are seen as the total disturbance via the active disturbance rejection controller (ADRC) based on the diagonal recurrent neural networks (DRNN). The system is decoupled into many independent subsystems by introducing the virtual control variables. And ADRC is designed in order to deal with the total disturbance in real time, where the total disturbance can be observed/cancelled by the extended state observer (ESO) based feedback control law. Then the DRNN which is suitable for dynamic systems is utilized to identify the system model online and optimize the parameters in nonlinear state error feedback (NLSEF) of ADRC in real time, which apply the dynamic back-propagation algorithm to train its weights and the adaptive learning rate method to speed up the training time. To validate the effectiveness of the presented scheme, numerical simulation study has been performed respectively via traditional ADRC and composite ADRC–DRNN controller under the same operating conditions by MATLAB. It is proved that the proposed ADRC–DRNN controller has a better control performance.