A Force-Feedback Assembly Method for Micro Parts Based on SRNN

Abstract

Force-feedback information is usful for micro-assembly system to enhence its contact sensing capability. On the basis of this view, a 3D force-feedback assembly method is proposed in this paper. It uses coordinate conversion to combine ideal pose data with pose error vector for assembly control. A kind of simple recurrent neural network (SRNN), whose weights is modified by using Levenberg-Marquardt (LM) algorithm, is applied to establish the mapping relationship between pose error vector and 6-DOF contact force/toque feedback form sensor. Experiments are carried out on backlash slider and base parts assembly to verify the performance of this method. It is proved that SRNN based on LM algorithm has good convergence ability and good fitting effects. Also,pose error can be accurately estimated and assembly searching times can be greatly reduced.