An adaptive protocol for pattern recognition and estimation of time synchronization in WSNs

Time synchronization has become an important and well-studied issue in wireless sensor network (WSN). Because of the physical characteristics of WSN such as resource limitation, dynamicity, and randomicity, building an efficient and accurate time synchronization model is a challenge. Most existing time synchronization model use first-order or second order model to describe the time variance. However, these models cannot describe the model alternation shows in our experiments. In this paper, we propose a new time synchronization model based on Multimodel Kalman Filter to depict the model alternation in time tracking. Furthermore, in order to achieve the Kalman filtering for time synchronization, we put forward an adaptive algorithm based on hypothesis testing. We demonstrate the performance gains of our protocol through experiments using different Kalman filter models based on experiment data.