Intelligent control of a 7-DOF manipulator based on model primitives

This paper presents a method to control industrial redundant manipulators based on a method similar to the mechanism ascribed to the human motor system that adaptively combines motor primitives to control human body motions. The idea is to identify a set of local model primitives based on Runge-Kutta-Gill neural networks (RKGNNs) and control the manipulator based on an adaptive selection and fusion method. The adaptive selection is carried out using a resonance signal that recalls only the most relevant memory primitives given a situation, and adaptive fusion is carried out by a method based on fuzzy memberships. This method imitates few recent findings on how human motor skills are developed by adaptive combination of motor primitives. Experiments on an industrial manipulator with seven degrees of freedom shows that the proposed method is a potentially viable approach to model based control of redundant robot manipulators.