Spontaneous synchronization in multihop embedded sensor networks: demonstration of a server-free approach

A common time reference is essential for distributed sensing, especially when the individual sensors are communicating over wireless, possibly through several intermediate nodes. Prior art in time synchronization over multihop wireless sensor networks has focused on centralized solutions that utilize one or more beacons or distributed peer-to-peer approaches, validated only in simulation environments. In this work, we demonstrate distributed synchronization using nearest neighbor communication and no other central point of control. The technique was inspired by natural synchronization in colonies of fireflies and was implemented in an embedded wireless network. The goal was to synchronize speakers and displays at various nodes of the network and quantify the synchronization error as a function of network diameter and communication overhead. One interesting finding was that the error does not scale linearly with the diameter of the network as reported previously in the literature, since it depends on the frequency skew distribution of the participating node oscillators.