Event-triggered state fusion estimation for wireless sensor networks with feedback

In this paper, the event-triggered state fusion estimation problem is considered for wireless sensor networks with limited communication resources. We propose two event-triggered fusion estimation algorithms under sequential and parallel fusion structures, respectively, where each sensor sends its observations to the fusion center only when its event-triggering condition is satisfied. In the sequential fusion estimation algorithm, the global estimate is updated with the received sensor information sequentially. While in the parallel algorithm, local estimates are first generated and then fused to obtain the global estimate. Moreover, feedback is adopted from the fusion center to sensor scheduler modules in order to make the triggering condition adaptive. The sequential fusion estimation algorithm has better local estimation performance while the parallel algorithm outperforms in computational efficiency, robustness and fault detection. Simulation results show that by adopting event-triggered strategy, the proposed algorithms can dramatically reduce the data transmission of the system with the cost of just a slightly deterioration of the estimation performance, compared with the traditionally time-triggered scheme.