A recurrent neural network for minimum infinity-norm kinematiccontrol of redundant manipulators with an improved problem formulationand reduced architecture complexity

摘要

This paper presents an improved neural computation where schemenfor kinematic control of redundant manipulators based on infinity-normnjoint velocity minimization. Compared with a previous neural networknapproach to minimum infinity-non kinematic control, the present approachnis less complex in terms of cost of architecture. The recurrent neuralnnetwork explicitly minimizes the maximum component of the joint velocitynvector while tracking a desired end-effector trajectory. Thenend-effector velocity vector for a given task is fed into the neuralnnetwork from its input and the minimum infinity-norm joint velocitynvector is generated at its output instantaneously. Analytical resultsnare given to substantiate the asymptotic stability of the recurrentnneural network. The simulation results of a four-degree-of-freedomnplanar robot arm and a seven-degree-of-freedom industrial robot arenpresented to show the proposed neural network can effectively computenthe minimum infinity-norm solution to redundant manipulators