Hierarchical Distributed Consensus Filtering of a Sensor Network

This paper addresses the problem of distributed consensus filtering in a sensor network. An efficient distributed sensor fusion algorithm is crucial for large-scale sensor networks as it minimizes communication resource usage while delivering accurate estimates. To tackle the problem of limited node-to-node connections in a network, we propose a hierarchical and distributed filtering structure comprising two layers. In this regard, all sensing nodes are organized into separate clusters to facilitate the efficient implementation of distributed consensus filtering in bottom layer. Each cluster contains a node that serves as a representative for its cluster. This node transmits the filtered results of its cluster to the top layer, enabling an additional layer of distributed filtering to compute the final filtered value. This study reveals a trade-off between the number of communication links and tracking error, governed by the number of clusters and the underlying graph structures. To achieve the desired balance, we propose an optimization framework to optimally select the number of clusters. The applicability of the present idea is demonstrated by simulation of a sensor network.