Bio-inspired Decentralized Control and Actual Machine Verification

Abstract

In this paper, we present a decentralized controller for multi-agent systems that can autonomously set some of the control inputs by the agents to zero, depending on the relation between an output and the target. Specifically, a nonlinear oscillator is employed in our proposed controller of agents, inspired from the reflective interlimb coordination in quadruped locomotion. The control input is binary ON/OFF, and it switches depending on the oscillator state. Furthermore, our proposed controller does not require information of other agents, but it only uses the broadcasted control error. The effectiveness of our proposed controller is verified by the speed control experiment of a three-motor system.