传感器网络分布式事件触发多目标估计

本文主要研究无线传感器网络中目标数目已知且固定的一类分布式多目标跟踪问题,提出了一种完全分布式的基于事件触发的测量和通信策略使得每个节点在不需要全局信息的情况下实现估计误差和能量消耗之间的平衡.监测区域存在多个移动目标,传感器能否测量到单个目标由事件触发测量机制和节点的测量半径来综合决定.基于节点和邻居的信息采用k-means聚类算法来解决数据关联问题,同时提出了基于最小迹原则的一致性卡尔曼滤波算法.从理论上证明了该事件触发策略不仅在性能指标上优于基于时间触发的算法,而且在网络中如果存在节点对多目标协同可观,系统估计误差在均方意义下是稳定的.最后给出了仿真例子验证了该算法的有效性和可行性.

Distributed event-triggered multi-target filtering in sensor networks

This paper mainly studies the energy efficient distributed multi-target estimation problem for a class of multitarget tracking problem with known and fixed number of tracked targets in wireless sensor networks. A fully distributed event-triggered measurement and communication strategy to enable each node to achieve better trade-offs between estimation error and energy consumption without global information is proposed. There are multiple mobile targets to be tracked in the monitoring area, and whether the sensor can measure a single target is determined by the event-triggered measurement mechanism and its measurement radius. k-means clustering algorithm based on each node’s neighbors’ information is applied to solve the data association problem and a consensus Kalman filter based on the minimum trace principle is also proposed. It is proved theoretically that the performance under the triggered strategy is better than that under the time-triggered algorithm. It is also proved that when there always exists a node which is collaboratively observable for all targets, the network estimation error is stable in the mean square sense. Finally, a simulation example is given to verify the effectiveness and feasibility of the algorithm.