A dynamic scheme for collaborative redundant manipulators aided with neural networks in a competitive manner

Abstract

Focusing on k-winners-take-all ($k$-WTA) strategy, this paper considers competitive kinematic program of multiple redundant manipulators using neighbor-to-neighbor communication topology and proposed a dynamic repetitive motion planning (DRMP). Aided by a distributed neural network solver, a cooperative control law of multiple redundant manipulators is formulated for dynamic task allocation with constraint equations, communication topology among manipulators, singularity avoidance of the winner manipulator and repetitive execution of the given tasks. Theoretical analyses prove the manipulation and feasibility of the proposed DRMP among multiple redundant manipulators. Computational simulations based on PUMA560 manipulators are conducted to verified the efficacy of the proposed DRMP and the underlying neural network.