Controle adaptativo híbrido de mecanismos atuados por músculos de Mckibben

Abstract

A robotic manipulator drive by McKibben muscles or simply a pneumatic manipulator belongs to a class of systems that have a fast response, low damping and non-linear behavior. Thus, the use of linear controllers, such as the Proportional-Integral-Derivative (PID) is prohibitive, as the control system will present different performances for different operational conditions. Thus, this research presents a new topological approach for hybrid adaptive control, with the purpose of contributing to the development of controllers for this type of plant. Such hybridity corresponds to the union of adaptive controllers by reference model (MRAC - Model Reference Adaptive Controller) and PID controllers. The choice is based on the unnecessary phenomenological models or plant identification, which facilitates the use in real-time and the use of Lyapunov's theory as a design step, which eliminates the analysis of stability after the control system design. The proposed control law has three parts, feedforward, derivative and feedback. The proposed topological structure is dedicated to pneumatic manipulators, in which the feedforward portion is responsible for the rejection of disturbances, the derivative smooths oscillations and the feedback has the purpose of canceling the tracking error in a permanent regime. The PID controller operates on a transitory regime, at which MRAC's performance is low due to the convergence time of the adjustment parameters. As a result, the thesis showed less global error in position tracking, both in simulations and in experiments, when compared to results in the literature, satisfactorily rejecting disturbances and complying with the ISO/TS 15066 standard that governs human-robot physical interaction in collaborative and assistive robotics.