Robot inverse kinematics: a modular neural network approach

Abstract

In this paper, a new method, based on modular neural networks, for the inverse kinematics of robotic manipulators is proposed. Neural modules are assigned to each link in order to realize its own inverse kinematics. The inverse neural modules are concatenated in a global scheme for the updating of the inverse kinematics of the manipulator. Three learning strategies are proposed for the inverse modular scheme. Simulation results for a 3 DOF manipulator and for a 4 DOF SCARA robot are presented. The training scheme, based on a simple training set is discussed.