Optimization and design methodology of fuzzy controller for industrial robotic tasks

Abstract

This paper presents an approach for complex task control involving robot/environment interaction. An effective hybrid force/position based-approach for multi-inputs/multi-outputs (MIMO) robot control is proposed. The approach is based on fuzzy logic controller (FLC) design and optimization methodology operating in two stages: at the first stage, the FLC architecture is defined and the input variables space is partitioned; and the FLC is trained offline on the basis of data acquired during free motion of the robot, in order to map FLC outputs to real behaviour. A method based on Solis' and Wetts' algorithm is then applied for fuzzy parameter optimization so that the constraints in terms of interpretability of the predefined rules are respected. Finally, an online learning of the generated FLC is implemented into the proposed control structure. The approach has been implemented into an experimental setup, including a 2D Cartesian robot linked to a C5 parallel robot, performing contour following under force constraints. The analysis and evaluation of the obtained results show the suitability and efficiency of the proposed approach