Advanced communication protocols for distributed estimation

The problem of distributed estimation is one of very active topics in the modern control theory and signal processing literature. Interest in this problem is motivated by a growing number of applications where a decision about the observed process must be made simultaneously by spatially distributed sensors, each taking partial measurements of the process. In the past several years, a number of results have been presented in the literature which develop the H∞ control and estimation theory for distributed systems subject to uncertain perturbations. In particular, methodologies of distributed sampled-data H∞ filtering have been considered to address the challenges of realistic sensor networks, among them coupling between sensor nodes through the information communicated between neighbouring sensor nodes and the sampled nature of that coupling, which is dictated by the digital communication technology.

In this work we addressed some of the challenges specific to Round-Robin type communication protocols. The type of communication we consider is where the nodes broadcast their information at every scheduled time instant to all nodes in their vicinity, but they listen to only one node within their neighbourhood at a time, according to the Round-Robin rule. We show that instead of continuously exchanging information (the type of networks considered in those references), the node observers can achieve the relative H∞ consensus objective by exchanging information at certain sampling times, by polling one neighbour at a time. We demonstrate that the Round-Robin design can be applied to derive a network of non-switching observers. Of course, each observer periodically switches between input channels, but the observer gains remain unchanged. This is an important feature of our methodology to ensure its scalability. This research has been supported by the Australian Research Council under Discovery Projects funding scheme. The results of this work have been published in [V. Ugrinovskii and E. Fridman, A Round-Robin Protocol for Distributed Estimation with H∞ Consensus, Systems & Control Letters, 69, 103–110, 2014.]