Distributed decentralized estimation

The problem of estimating states of a dynamical system appears in numerous engineering and scientific applications. It has been in the focus of research for many decades. Recently with the emergence of new sensor and communication technologies the old estimation problems received new attention. Decentralized and distributed estimation problems came back into light with the new technological abilities to sense and collect information locally, reliably transfer it to the desired locations and take local actions. In this thesis we consider a formulation of the decentralized distributed estimation problem and develop synthesis tools for the design of distributed estimators able to provide state information to the decentralized control systems.;We propose solutions to the problem of synthesis of distributed decentralized estimators for a formation of agents. We assume that the collected dynamics of the formation are modeled by a discrete linear time-invariant system. We consider the estimation structures where each agent of the formation carries an estimate of the entire formation state. Agents of the formation are allowed to transfer information between each other through unidirectional links specified by a communication topology. We consider several models for the communication links and communication protocols. The design challenges rise mainly from the properties of the considered communication systems.;We formulate the problem of optimal design of the gains of the distributed decentralized estimators and propose constructive solutions. The design procedures are based on a set of convex optimization formulations with linear matrix inequalities. The suboptimal gains calculated with the proposed methods stabilize the estimation error dynamics and minimize a performance index. As the performance measures we use the norms of the estimation error correlation matrices. In order to develop the proposed methods we use recent results from convex matrix analysis and the theory of Markov jump linear systems. Detailed examples illustrate the application of the proposed techniques.