Improved interval-based clock synchronization in sensor networks

Abstract

Interval-based synchronization provides the nodes of a distributed system with guaranteed bounds on a common time. This is a crucial piece of infrastructure in many distributed sensing and actuating systems. In this paper, we propose a modification to a known interval-based synchronization algorithm; our new algorithm obtains substantially better results in sensor-network scenarios by taking advantage of the typical drift diversity of the nodes' clocks. We propose a model for synchronization in ad-hoc, sporadic-communication scenarios. The model allows us to identify the worst and the best case in terms of achievable time uncertainty and to show the worst-case optimality of the discussed algorithms. Simulations show that in the average case, our modification significantly reduces the time uncertainty.