Researches Of Time Synchronization Methods For Wireless Sensor Networks

In wireless sensor networks (WSN), due to cost constraints, the accuracy of sensors’clocks is unfavorable. As time goes on, sensors’clocks will gradually drift from each other, and this issue would become more and more serious along with the growth of the sensor number of the network. While a common time frame over the sensor network is crucial for network in performing a number of its fundamental operations, thus time synchronization in wireless sensor networks has attracted more and more attentions from the researchers.Single-hop clock synchronization and global clock synchronization are profoundly studied in this paper, based on two different kinds of clock models, i.e., time-invariant clock model and time-variant clock model. As the basement of time synchronization researching, several time models are introduced, as well as the three mainstream sensor communication methods, which are the two-way message exchange, the one-way message dissemination and the receiver-receiver synchronization. All following researches for time synchronization in this paper is based on the two-way message exchange scheme.During the researching based on the time-invariant clock model for time synchronization, the Maximum Likelihood Estimation (MLE) is firstly researched, its performance for time synchronization is verified by simulation experiments. And to settle the computational issue of the MLE algorithm, a modified MLE algorithm, which has significant lower computational load, is introduced. The modified MLE algorithm is compared with the original MLE algorithm through simulations. Furthermore, the Recursive Least Square (RLS) algorithm are introduced into time synchronization, as well as another low-complexity RLS algorithm, i.e., the Exponentially weighted RLS algorithm with Dichotomous Coordinate Descent (ERLS-DCD). Simulations are implemented to contrast their performance for time synchronization situations.Then, the application of the adaptive algorithms for time synchronization are studied based on the time-variant clock model. Firstly, the adaptive filtering process formula and measurement formula are deduced. Then, the Kalman Adaptive filter and the H∞ filter are introduced and brought into the time synchronization applications. In further experiments, the synchronization performance of these two filters are discussed.Finally, the global time synchronization are studied to realize synchronization all over the network. The hierarchical tree structure for global synchronization is researched at first, the Kalman filter and the H∞ filter are applied into this structure and their synchronization performance are deeply tested by several experiments. Then, a distributed Consensus global synchronization algorithm is introduced and its performance for different synchronization situations is examined by simulations. Simulations show that this algorithm is sensitive to the message propagation delay, to overcome this drawback, a modified consensus algorithm is proposed, and simulations show that the proposed algorithm outperforms the original algorithm for situations in which propagation delay exists.