Merging ultra-low duty cycle networks

Energy is the scarcest resource in ad-hoc wireless networks, particularly in wireless sensor networks requiring a long lifetime. Intermittently switching the radio on and off is widely adopted as the most effective way to keep energy consumption low. This, however, prevents the very goal of communication, unless nodes switch their radios on at synchronized intervals, a rather nontrivial coordination task. In this paper we address the problem of synchronizing node radios to a single universal schedule in very large scale wireless ad-hoc networks. More specifically, we focus on how independently synchronized clusters of nodes can detect each other and merge to a common radio schedule. Our main contributions consist in identifying the fundamental subproblems that govern cluster merging, providing a detailed comparison of the respective policies and their combinations, and supporting them by extensive simulation. Energy consumption, convergence speed, and network scalability have been the driving factors in our evaluation. The proposed policies are extensively tested in networks of up to 4,096 nodes. Our work is based on the GMAC protocol, a gossip-based MAC protocol for wireless ad-hoc networks.