Skip to Main Content
One of the main issues with wireless sensor networks (WSNs) is communication efficiency to reduce power consumption. This is often achieved through data routing with aggregation. However, aggregation does not consider possible measurements correlation. This a priori information can be used for data fusion, in order to remove correlation of transmitted data, thus reducing amount of information to be transmitted. In this paper we explore different information processing strategy for state estimation of dynamic linear systems in the context of rooted wireless sensor networks. We propose three data fusion methods: standard measurement aggregation, measurement fusion, and fusion of partial state estimates. These strategies are specifically designed to include possible delayed or dropped packets. Finally, they are compared in term of power consumption efficiency, estimation accuracy, computational and memory complexity, showing tradeoffs between these metrics.