V-REP-based Virtual Platform on Multi-Pattern Neural Control of Robotic Manipulator

Shengzhao Huang,Min Wang,Yongtao Zou

DOI: https://doi.org/10.1109/ccdc.2019.8832664

2019-01-01

Abstract:The paper is concerned with a virtual experimental platform on multi-pattern intelligent control of robotic manipulator, which achieves the joint simulation of V-REP and Matlab. A novel pattern-based control scheme is developed based on deterministic learning mechanism. In the proposed scheme, the unknown dynamics of the manipulator with single reference trajectory is approximated by RBF neural network (NN) and stored in static NN with constant weights. Using such a method, for different reference trajectories, we construct a group of static NN learning controllers which embed constant NN, and then propose a novel switching control strategy. To make manipulator perform multiple control tasks autonomously, different recurrent reference trajectories are regarded as different patterns. RBF NNs are used to identify different patterns. and the corresponding identified results are stored in constant NN weight values. The stored values are employed to develop dynamical pattern estimators. The combination of pattern estimators and switching control strategy makes that the manipulator can rapidly identify different tasks and reuse the corresponding NN learning controller to achieve the desired control performance. To verify the effectiveness of the proposed scheme, V-REP and Matlab are jointed to build the virtual experimental platform. The paper also introduces how to set up the properties of manipulator joints inserted in the 2-link manipulator model in V-REP, the scripts types of embedded scripts, and the method of using remote API to realize the joint simulation. Finally, the operation environment and the result of joint simulation are exhibited to verify the validity of the proposed scheme and the applicability of joint simulation.