| We study the problems of distributed estimation and detection, where a set of nodes wish to collectively reach a common objective by using their measurements, and through local interactions with neighboring nodes. In the distributed estimation problem, nodes interact in order to obtain an estimate of some quantity of interest from their measurements. In the distributed detection problem, nodes use these measurements to decide between multiple hypotheses. We address the question of how these nodes should cooperate in order solve the estimation and detection problems in a fully distributed manner. Our focus is on diffusion schemes, where nodes communicate with their neighbors only and no fusion center is necessary. Diffusion schemes have advantages of being robust to node and link failure, able to adjust to the network topology, and can save communication resources by communicating with neighboring nodes only. Through a set of local interactions, information is diffused across the network, and good performance can be achieved. We propose algorithms for distributed LMS, RLS and Kalman filtering, and analyze the performance of the proposed algorithms in terms of their convergence, transient and steady-state behavior. We extend the estimation framework to the case of distributed detection and discuss potential applications such as spectrum sensing in cognitive radios and bird flight formations. |
