Autoregressive Integrated Model for Time Synchronization in Wireless Sensor Networks

Time synchronization is challenging in wireless sensor networks due to the use of low-precision oscillators and the limited computational capacity of resources limited sensor nodes. While several schemes exist, the performance analysis of a majority of them is based on simulations and fail to capture key features of real world deployments. This paper explores the use of autoregressive integrated moving average models to provide a general clock model for sensor nodes with low precision oscillators and limited computational power. Based on measurements with off-the-shelf sensor devices Z1, an autoregressive integrated model for time synchronization is proposed. We derive a synchronization scheme (ARI-Sync) based on this model and compare it against the well known Flooding Time Synchronization Protocol (FTSP) observing significantly improved accuracy, roughly doubling the resynchronization period of Z1 nodes for a typical wireless sensor network application.