Controle ℋ∞ não linear adaptativo baseado em aprendizado profundo para robôs móveis com rodas

Abstract

The implementation of autonomous mobile robots raises important issues when you want a robust navigation system to transport loads. These types of vehicles are often subject to parametric uncertainties and external disturbances. Parameter uncertainties usually arise due to extra devices or loads that are attached to the robot since they influence the parameters of mass, inertia, center of mass, and other parameters initially raised to compose the vehicle mathematical model. And the external disturbances are related to the robot’s collision with static or dynamic obstacles or in overcoming obstacles on the ground by the robot’s wheels, where skidding and slippage of the wheels can occur. Based on this context, the present work proposes a robust and adaptive control architecture for the trajectory tracking problem of a mobile robot subject to external disturbances and parametric uncertainties. The proposed approach will comprise a Nonlinear Adaptive Control ℋ∞ based on Deep Learning. The nonlinear ℋ∞ controller will be responsible for attenuating external disturbances, and the adaptive part will be based on Deep Neural Network for learning the parametric uncertainties related to the robot’s mathematical model. Simulation results of a four-wheel mobile robot for the tracking problem are used to compare the control strategies proposed.