Dynamic Neural Learning based Tricriteria Optimization Scheme for Dual-Redundant-Robot Manipulators Collaborative Motion Planning
Zhijun Zhang, Ren Ge, Jilong Xie, Yamei Luo · 2024
In order to solve the optimal solution for dual redundant robot manipulators for tracking cooperative end-effector paths, a novel dynamic neural learning based tricriteria optimization (DNLTO) scheme is proposed involving multiple robot manipulators collaborative working. Specifically, the DNLTO scheme consists of two parts, i.e., tricriteria optimization scheme and DNLTO solver. To simultaneously and collaboratively control both arms, the tricriteria optimization scheme are described as left/right robot manipulators quadratic programming (QP) issues and then composed into a single QP formulation. The tricriteria optimization scheme of DNLTO scheme integrates the minimum velocity norm, repetitive motion planning, and infinity-norm velocity minimization optimization scheme through use of two weighting factors, and is reformed as standard quadratic programming problem. Then, a penalty strategy based varying-parameter recurrent neural-network is applied to solve the QP formulation. Results from computer comparisons using dual robot manipulators can verify the effectiveness and precision of the DNLTO scheme. Compare with the traditional neural network, such as a primal dual neural network with simplified linear variational inequality, the proposed DNLTO scheme has less computation time and smaller end-effector positioning errors.