Controle inteligente de um manipulador robótico com regressão por processo gaussiano

Abstract

The technological advancement promoted in the Kondratiev’s fourth Cycle gave rise to the internet of things and the industry 4.0, in a context in which the social dynamics, whether professional or domestic personal, changed by the insertion of artificial intelligence and robotics in daily activities, for example. Thus, the Mechanical Engineer must act as a change agent, acting in the technological vanguard in multidisciplinary areas such as robotics, which involves electronics, mechanics, programming and, often, biology, neuroscience and philosophy. This means that a robotic manipulator is not only an opportunity for research and development of its mechanical structure and its electronic components, but also a great opportunity to evaluate control algorithms, from more conventional techniques such as Proportional-Integral-Derivative (PID), to a learning algorithm. In practice, the dynamics of most of these systems present a high degree of nonlinearities and uncertainties, which makes their position and trajectory control a challenging situation, corroborating the effervescence of this research area. Thus, knowing that the controller design is shown as fundamental for these robots, this work represents a contribution to the Mechanical Engineering when proposing an intelligent controller for a robotic manipulator of two degrees of freedom. The control law uses Gaussian Process regression to compensate for uncertainties and unmodified dynamics in the feedback linearization technique.