Controle tolerante a falhas aplicado ao modelo de um robô móvel com restrições não holonômicas

Abstract

This work presents the theory and simulation of a model-based Fault Tolerant Control (FTC) strategy for a mobile robot with nonholonomic constraints. The control system consists of a controller whose gain is designed by Linear Matrix Inequalities (LMI) applied to track the robot's linear and angular velocity references. The model used for the robot is Linear Parameter-Varying (LPV) where an LPV observer is also designed by LMI for feedback control purpose. The presented FTC technique is capable of identifying and isolate either multiple faults, simultaneous or non-simultaneous, at the actuators or sensors. The fault identification process applies a Recursive Kalman Filter for each plant component, whether it is an actuator or sensor, in order to identify and isolate the fault. These faults are predicted in a state-space model and introduced in an additive or multiplicative way. LPV virtual actuators and sensors act in the fault correction, which spares the need of the real-time redesign of the controller. Simulation results showcase and validate the presented theory for the FTC strategy including discussion of the pros and cons of its application.