Controle preditivo com restrições aplicado ao rastreamento de trajetória de um robô escalador com rodas magnéticas

Abstract

In this work, a trajectory tracking control is presented for a climbing wheeled robot under magnetic adhesion, whose function is to carry equipment that perform the inspection of LGP (Liquefied Gas Petroleum) storage tanks. The robot has differential drive and, due to the magnetic adhesion, when performing closed curves, deforms the physical structure of the chassis. This is due to wheels have great difficulty in detaching themselves from the metal surface during the maneuver. The dissertation purpose is to develop a control system that, based on the modeling forces that act on chassis, is capable of solving the trajectory tracking problem without chassis deformation. This is a constrained multivariable control problem and the control technique used to development is the MPC (Model Predicitive Control), more specifically GPC (Generalized Predictive Control). A black box robot model is identified by least squares from simulated data. For developed controller validation, several simulations experiments are carried out by varying the parameters of the controller and comparing obtained results without and with the inclusion of constraints. Monovariable PI controllers are designed in a decoupling control scheme to serve as a reference for evaluate the performance of the proposed controller. All the results are validated through simulations using the ROS (Robot Operating System), V-REP (Virtual Robot Experimentation Platform) and MATLAB softwares.