Vector Kalman Filter Using Multiple Parents for Time Synchronization in Multi-Hop Sensor Networks

In multihop wireless sensor networks, global clock error will accumulate as the hops grow. Sensors far away from the root are likely to suffer larger synchronization errors, which may deteriorate the accuracy of sensing data or even result in the failure of TDMA MAC protocols. A novel algorithm based on vector Kalman filter using multiple parents (KFMP) is proposed to address this problem. It´s pointed out that receiving a synchronization message from one parent is equivalent to obtaining an observation of the global clock, thus more accurate result could be achieved by combining observations from multiple parents. In KFMP, vector Kalman filter is adopted to combine these observations. Both hardware implementation and system simulation are conducted to evaluate the performance of the proposed algorithm. The results show that global clock error is significantly reduced compared with FTSP, especially for sensors at the edge of the network.