Distributed Finite Time Termination of Consensus in the Presence of Delays

TL;DR

This paper presents a distributed algorithm enabling networked agents to detect approximate consensus within a finite number of iterations despite communication delays, reducing resource wastage and allowing real-time system tasks.

Contribution

It introduces a novel finite-time termination algorithm for consensus under delays, including a practical implementation for averaging initial values in a distributed manner.

Findings

Algorithm achieves finite-time approximate consensus despite delays.

Experimental validation on Raspberry-Pi network demonstrates effectiveness.

Reduces computational complexity for hardware deployment.

Abstract

Linear consensus iterations guarantee asymptotic convergence, thereby, limiting their applicability in applications where consensus value needs to be used in real time to perform a system level task. It also leads to wastage of power and communication resources. In this article, an algorithm is proposed which enables each node to detect in a distributed manner and in finite number of iterations, when every agent in the network is within a user specified threshold of the consensus value (approximate consensus) and hence terminate further communications and computations associated with consensus iterations. This article develops a distributed algorithm for achieving this approximate consensus in presence of random time-varying bounded communication delays. Moreover, the article instantiates the algorithm developed to distributively determine the average of the initial values held by agents in finite number of iterations. Specifically, this algorithm relies on distributively determining the maximum and minimum of values held by the agents. The approach presented here offers several advantages, including reduced computational complexity, and hence, is suited for hardware implementation. An experimental test bed of Raspberry-Pi agents that communicate wirelessly over neighborhoods is employed as a platform to demonstrate the effectiveness of the developed algorithm.