An estimation scheme based on pairwise broadcast to synchronize wireless sensor networks

A simple and elegant method to synchronize a wireless sensor network is pairwise broadcast, where each pair of nodes exchange timing packets while their neighbors listen opportunistically and exploit the packets for synchronization. The traditional method of pairwise broadcast, however, neglects to use all available packets in synchronization, resulting in inaccuracy as measured by the mean square error (MSE). This paper uses all available timing packets, at a small overhead, to estimate the clock offsets and the related parameters required for network synchronization. The method of estimation is a classic maximum likelihood estimation together with an ingenious approach to optimization. The method yields closed-form expressions of the estimators, which are crucial for sensor nodes that are simple and limited in battery life. We compare the proposed estimators with those of the traditional pairwise broadcast. The proposed estimators reduce the MSEs of the traditional pairwise broadcast by four times. To achieve a target MSE, the proposed estimators halve the number of packets required by the traditional pairwise broadcast, hence synchronizing a network at half the time of the traditional approach. These findings confirm practical utility of the proposed synchronization method, which suits the simple and energy-conscious sensor nodes.