Distributed binary consensus in dynamic networks

Abstract

Motivated by the distributed binary interval consensus and the results on its convergence time we propose a distributed binary consensus algorithm which targets the shortfall of consensus algorithms when it comes to dynamic networks. We show that using our proposed algorithm nodes can join and leave the network at any time and the consensus result would always stay correct i.e. the consensus would always be based on the majority of the nodes which are currently present in the network. We then analyse our algorithm for the case of complete graphs and prove that the extra time it takes for nodes to implement our algorithm (to cope with the dynamic setting) does not depend on the size of the network and only depends on the voting margin. Our results are especially of interest in wireless sensor networks where nodes leave or join the network. © 2013 Springer International Publishing.