Consensus-based Distributed Estimators In Sensor Networks

Consensus-based distributed state estimation over sensor networks has extensive applications such as target tracking, traffic estimation and situation monitoring. It can drastically reduce the utilization of communication resources, has no requirement on the network topology of sensor networks, and is more flexible for ad-hoc sensor networks when compared with centralized state estimation.This thesis investigates the problem of consensus-based distributed state estimation in sensor networks with some constraints such as event-triggered, unreliable communication links and randomly switching topology. The main contents of this thesis are summarized as follows:1. An event-triggered approach to distributed state estimation for nonlinear discrete time-delay systems over sensor networks is studied. First, an event detector is configured at each sensor node to decide when the node should send its sampled measurement output to its neighbors across the communication network. Then, sufficient conditions for the existence of an admissible distributed state estimator are established based on the theories of both time-delay systems and Lyapunov stability. A effective algorithm of the design of both the distributed state estimator design and the parameters in the event condition could be achieved by solving a set of linear matrix inequalities. Finally, two numerical examples are provided to verify the designed event-based distributed state estimator.2. The design problem of consensus-based distributed sampled-data state estimators for a class of continuous-time delayed nonlinear systems in sensor networks with unreliable communication links is concerned. A switched system model is presented to describe the distributed estimators considering intermittent communication links. The exponential stability of the estimation error dynamical network is analyzed in the average dwell time framework. The effectiveness of the designed distributed estimators is verified by numerical examples.3. The problem of consensus-based fault detection filters design for Markovian jump systems with mode-dependent delays is studied. It is assumed that the fault detection filters may communicate state information via a sensor network with randomly switching topology governed by a Markov process which is different from the one dominating the nodes systems. It is also supposed that the two Markov processes are independent each other. Based on a mode-dependent Lyapunov functional, the existence conditions of the desired mode-dependent fault detection filters are derived such that the overall fault detection dynamics is exponentially mean-square stable and, at the same time, the errors between the residual signals and the fault signal are made as small as possible in the H? sense. A numerical example and a vertical take-off and landing helicopter system are provided to verify the effectiveness of the proposed algorithm.