A radial basis function network approach to approximate the inverse kinematics of a robotic system

This paper presents a novel solution using a radial basis function network (RBFN) to approximate the inverse kinematics of a robotic system where the geometric parameters of the manipulator are unknown. Simulation and experimental results are presented for a three-link manipulator to demonstrate the effectiveness of the proposed approach. To achieve this level of performance, centres of hidden-layer units are regularly distributed in the workspace, constrained training data is used where inputs are collected approximately around the centre positions in the workspace and the training phase is performed using either strict interpolation or the least mean square algorithm. These proposed ideas have significantly improved the network's performance.