Controle preditivo não linear baseado em modelo aplicado a formações de robôs móveis terrestres com seleção dinâmica de líder

Abstract

This research presents a distributed control architecture for robot formation control, using a modified version of the Non-linear Model Predictive Control algorithm with a generic cost function, which allows for the combination of the behavior layer into the controller while keeping or improving upon the precision algorithm. A dynamic leader-selection structure iso also prpoposed. The theoretical concepts which served as a basis for the text and the reference works which constitute the state of the art are briefly presented. The proposed Non-linear Model Predictive Control modifications, which employs a virtual leader to improve closed loop performance, and the tree based selection structure, used as a state machine, are explained. The results obtained fora a three robot formation dealing with the trajectory tracking problem are shown, for both simulated and experimental environments.