Skip to Main Content
A multiestimation-based adaptive controller is designed for robotic manipulators. The identification subsystem is composed of a bank of estimation algorithms running in parallel incorporating each one a relative adaptation dead-zone in order to guarantee adequate estimation properties under the influence of bounded perturbations and unmodelled dynamics. Moreover, a supervisory index is proposed with the aim of evaluating the identification performance of each identification algorithm. Then, a higher order level supervision algorithm decides in real time the estimator which will parameterize the adaptive controller according to the values of the above supervisory indexes. There exists a minimum residence time between switches in such a way that the closed-loop system stability is guaranteed. It is verified through simulation examples that multiestimation based schemes can improve the transient response of adaptive systems through appropriate switching between the various estimation algorithms running in parallel. The closed-loop system is robustly stable under the influence of uncertainties due to a poor modelling of the robotic manipulator while the switching process is subject to a minimum residence time. Some simulation examples show the usefulness of the proposed scheme.