Virtual clustering: a communication cost reduction strategy for distributed consensus-based estimation in cooperative networks

In this paper, we consider the problem of distributed consensus-based estimation in cooperative networks, e.g., wireless sensor networks (WSNs). To solve this problem and achieve an accurate consensus-based estimate solution, many iterative distributed algorithms require the information exchange among nodes at each iteration suffering from huge communication overhead. In our previous work, a new strategy of virtual clustering was discussed with the purpose of reducing communication overhead for distributed consensus-based estimation. By classifying the data using virtual clusters, data with reduced size will be transmitted during the distributed processing. Here, we further propose two methods to reduce the size of transmitted data for arbitrary network topologies. One method is based on finding the shortest path in a network and the other relies on linear independence of constraint qualification (LICQ). The study shows that both methods can successfully reduce communication overhead. Moreover, the second method outperforms the first one and provides the optimal communication cost for the distributed consensus-based estimation.