Adaptive Clock Skew Estimation with Interactive Multi-Model Kalman Filters for Sensor Networks

Zhe Yang,Jianping Pan,Lin Cai

DOI: https://doi.org/10.1109/ICC.2010.5502549

2010-01-01

Abstract:Clock synchronization is a fundamental issue in communication networks and distributed systems, and clock skew is the inherent cause for clock desynchronization. Clock skew estimation is essential to improve the efficiency and reduce the overhead of clock synchronization schemes, and it is especially beneficial for resource-constrained devices such as sensor nodes in dynamic environments. According to the measurement, clock skew is environment sensitive, and no existing clock skew estimation schemes can accurately capture such dynamic behaviors. In this paper, we investigate a general clock synchronization problem with variable clock skews and propose a new skew estimation model based on a hybrid approach to characterizing the dynamic of clock skews. To estimate the time-varying clock state vector, we employ the Interactive Multi-Model (IMM) Kalman filter, which can make soft decisions by combining results from different models. Extensive simulations have been conducted to demonstrate the effectiveness of the proposed adaptive clock skew estimation algorithm, which achieves a better performance with moderate computational complexity.