Tuning of Nonlinear MPC Algorithm for Vehicle Obstacle Avoidance

Abstract

This work is concerned with tuning a nonlinear Model Predictive Control (MPC) algorithm. Typically, the weighting coefficients associated with the predicted control errors are constant for the consecutive sampling instants over the prediction horizon. This work discusses a tuning procedure whose objective is to find a set of coefficients which scale the influence of the consecutive predicted control errors. The coefficients are determined using an original simulation-based method. In order to demonstrate effectiveness of the method, an MPC algorithm for vehicle obstacle avoidance is developed. It is shown that the discussed tuning methods makes it possible to obtain much better control quality in comparison with the classical approach.