Adaptive time window linear regression algorithm for accurate time synchronization in wireless sensor networks

Abstract

In this article we propose a new algorithm for time synchronization in wireless sensor networks. The algorithm is based on linear regression to achieve long-term synchronization between the clocks of different network motes. Since motes are built using low-cost hardware components, usually their internal local clocks are not very accurate. In addition, there are other effects that affect the clock precision, such as: environmental conditions, supply voltage, aging, manufacturing process. Because some of these causes are external and unpredictable, the clock drift between two motes can change in a random way. Due to these changes, the optimum time window used for performing the linear regression varies with time. The proposed time synchronization algorithm adjusts the resynchronization periods and the linear regression window size to these variations, minimizing the synchronization error. Our algorithm has been tested in real multihop network deployments and the results obtained show higher clock accuracy when compared to the related work.