MODEL BASED PREDICTIVE CONTROL OF UNDERACTUATED NONLINEAR MECHATRONICAL SYSTEMS

Abstract

The paper deals with model predictive control of underactuated nonlinear mechatronical systems along known reference path. It generalizes the state space predictive control algorithm of linear time invariant (LTI) systems to linearized time variant (LTV) systems. An algorithm is presented, which calculates the LTV model online from the nonlinear model along the reference trajectory. The LTV is then used in the framework of the predictive control to find the optimal control in closed analytical form without using online optimum search in the moving horizon. After MATLAB-based simulation results of the algorithm, successful test experiments were performed for the predictive control of a real inverted pendulum system, both in the swinging up and upper stabilization phases.