Direct adaptive inverse control of nonlinear plants using neural networks

Neural networks are universal approximators that can estimate and control plant dynamics under severe process nonlinearities. In this paper, we propose a direct adaptive inverse control (DAIC) for nonlinear plants based on multilayered feed forward neural network (MFNN) using back propagation through model (BPTM) algorithm. Neural Network (NN) controller is designed directly in the feed forward loop using the estimated model of nonlinear plant, which is obtained using forward modeling. NN adaptive control is designed such the NN controller responds online to the changes in system dynamics. NN adaptive controller adapts to the error signal between desired output and plant output; back propagated through estimated nonlinear plant model, using BPTM algorithm. Our proposed system is applicable to stable or stabilized, minimum/non-minimum phase, linear and nonlinear plants. Simulation results show the effectiveness of the proposed algorithm.