Intelligent control of mechatronic systems: From autonomous vacuum cleaners to self-driving cars, intelligent mechatronic systems are becoming an intrinsic part of our daily lives. Although conventional control schemes have been successfully employed to industrial robots, they might not represent the most proper choice to deal with mechatronic systems subject to large uncertainties and/or drastic changes in their operating conditions. Therefore, in order to let mechatronic systems undertake assignments in an environment of uncertainty and imperfect information, in our group we combine modern control methods with soft computing techniques, such as fuzzy logic and artificial neural networks.

Obstacle avoidance using Genetic Algorithms and Bézier curves

Intelligent control of a single-link flexible manipulator

Biologically inspired approaches for autonomous agents: The development of reliable autonomous agents depends on the adopted learning and planning algorithms. Thus, we propose the implementation of reinforcement learning algorithms to let autonomous robots learn by interacting with their own environment. By means of a cooperation with researchers from the Department of Physiology (UFRN), we propose computational approaches that are able to emulate the learning process of living organisms.

Sliding modes with Gaussian process regression for the control of uncertain nonlinear systems: A Gaussian process may be understood as an extension of the Gaussian probability distribution. On this basis, Gaussian process regression (GPR) can be adopted as non-parametric model to represent samples from a multivariate Gaussian distribution, in order to estimate both structured and unstructured uncertainties related to plant dynamics. Although GPR alone is not able to provide an efficient feedback controller, it could be conveniently associated with modern control schemes. Hence, in order to meet both robustness and tracking requirements, sliding mode control is combined with Gaussian process regression.