Bi-criteria Acceleration Minimization of Redundant Robot Manipulator using LVI-based Primal-Dual Neural Network

Abstract

The infinity norm of joint acceleration minimization (also known as the acceleration-level minimum-effort solution) explicitly minimizes the largest component of joint accelerations in magnitude. It is useful in situations where focuses are on low individual magnitude, even distribution of workload, and analysis of motion diversity. However, the minimum-effort solution may encounter discontinuities because of the non-uniqueness of the solution. To remedy such a discontinuity problem, this paper involves two important matters. (1) A new acceleration-based bi-criteria scheme is proposed for preventing the INAM solution discontinuities and joint torques instability problem. It combines the minimum infinity-norm and minimum two-norm solutions via a weighting factor and formulates this scheme as a quadratic programming (QP) problem. (2) The LVI-based primal-dual neural network is presented to solve online such a weighting scheme, because of its simple piecewise-linear dynamics and higher computational efficiency. Simulation results based on PMUA560 robot manipulator illustrate advantages of such a neural weighting scheme proposed in this paper.