Hybrid-Level Joint-Drift-Free Scheme of Redundant Robot Manipulators Synthesized by a Varying-Parameter Recurrent Neural Network

Abstract

For a redundant robot manipulator, the external perturbation and digital computation errors are inevitable when executing a tracking task. In this paper, a hybrid-level joint-drift-free (HL-JDF) scheme is proposed and synthesized by a novel varying-parameter recurrent neural network [called varying-parameter convergent-differential neural network (VP-CDNN)] with inherent perturbation tolerance. The HL-JDF scheme is combined with torque and velocity level optimization schemes and aims to solve the joint-drift problem with noise considered in the tracking task of redundant robot manipulators. The tracking task of a redundant robot manipulator is first formulated as a time-varying convex quadratic programming (QP) problem. Second, the QP problem is converted into a matrix equation. Finally, the proposed VP-CDNN is applied to solve the matrix equation. What is more, theoretical robustness analysis of the proposed VP-CDNN is presented. In addition, computer simulations and physical experiments of a redundant robot tracking task synthesized by the proposed VP-CDNN and the conventional fixed-parameter convergent-differential neural network (FP-CDNN) are conducted for illustrations and comparisons. Both the theoretical analysis and experiment results prove the effectiveness of the proposed HL-JDF scheme and the higher accuracy and better ability to resist the disturbance of the proposed VP-CDNN compared with the traditional FP-CDNN.