Model-free learning control of nonlinear discrete-time systems

Abstract

A model-free learning controller for a general class of nonlinear discrete-time state-space systems is introduced. The learning component of the proposed controller can use an arbitrary function approximator such as a Polynomial, Radial Basis, or Neural Network to directly learn the inverse of the input-state mapping of the plant while forcing its state to track a prescribed desired trajectory. Unlike most of the existing direct adaptive or learning schemes, the nonlinear plant is not assumed to be feedback linearizable. The developed controller is subsequently applied to control the configuration of a nonholonomic differential drive robot. The simulation results of this application demonstrate a significant improvement in the tracking performance of the robot once the control input is fully learned.