Modeling and Tracking Clocks with Time-Varying Skews Using Information Criteria

Synchronizing clocks is important to guarantee reliable and efficient network functionality for precise coordination among different nodes or components. Clock synchronization becomes more challenging for networks with low-cost devices such as sensors, because these nodes are equipped with unreliable clocks which are vulnerable to noises and environmental changes. With the drastic increase in the use of sensors, modeling and tracking clocks are becoming critical. In this paper, we develop a stochastic clock model using information criteria which follows the principle of Occam's razor. The model is optimized in terms of the number of parameters and validated by real measurements on low-powered micro-controllers. Based on our model, we show that Kalman filter tracks the clock uncertainty well with a fairly short training. We further evaluate our clock modeling and tracking methods by considering missing observations.