Distributed Estimation In Industrial Wireless Networks Based On Sensor Coordination

With the development of wireless technology and wireless network technology,the industrial wireless network (IWN) technology becomes the key in industrial infor-mationalization, which could ofer competitive advantages, such as greater fexibility,improved scalability, lower cost and ease of maintenance. The IWNs could enablethe industrial end-users to achieve “ubiquitous sensing” of the whole assembly line,guarantee the normal operation of the feld devices, improve the product quality andincrease the productivity and energy efciency. However, the harsh industrial environ-ments, coexistence of several wireless services, overlapping of their frequency bandsand strong electromagnetic interference severely infuence the quality of wireless com-munication such that the information accuracy and wireless reliability can not be guar-anteed.In this dissertation, we focus on the information reliability issues of IWNs. Thenotion of reliability mainly has three facets: information accuracy of the sensor nodes,reliable data transmission, coverage and deployment of the network. In this disserta-tion, we pay our attention to the information accuracy at the sensor node level, andcoverage and deployment problem at the network level. Specifcally, we investigatethe distributed estimation problem to improve the information accuracy via sensor co-operation at the sensor nodes, and deployment problem to guarantee the estimationperformance at the network level.At the sensor node level, since there are usually many kinds of sensor nodes co-existing in the industrial felds, moreover, the measurements of sensor nodes are cor-rupted by environment noise and there are disagreements of data among sensor nodesat diferent places, we study the problem of distributed estimation of deterministic pa-rameters in IWNs with two types of sensor nodes, namely, the primary nodes and the relay ones. A novel consensus-based distributed estimation algorithm for two types ofsensor nodes is proposed. Then we analyze its properties by using results borrowedfrom Markov chain and algebraic graph theory. The results show that the estimatesof the relay nodes are determined by those of the primary sensor nodes. Moreover,its mean-square stability is established by casting it into a stochastic framework. Thesimulation results show that the performance of the proposed algorithm is highly re-lated with the network topology, which means that we should pay much attention tothe network topology design.Wealsostudytheproblemsofdistributedstateestimationatthesensornodelevel.Combined with the classical Kalman flter and distributed consensus algorithm, wepropose a novel distributed optimal consensus Kalman flter for IWNs with two typesof sensor nodes. This flter inherits the merits of Kalman flter, which can attenuateprocess and measurement noise, and those of consensus algorithms, which can enableall the sensor nodes to achieve agreement on their estimates. In order to minimize theestimate errors, we adopt the sequential scheme to analyze its unbiasedness and opti-mality, and presentthe designguideline ofthe optimalflters, whichprovidesafeasibleapproach to the flter design in the present of two types of sensor nodes. Moreover, itsmean-square stability is established by using It stochastic diferential theory.Atthenetworklevel,weinvestigatethesensordeploymentproblemfordistributedestimation in IWNs. Firstly, we discuss the relation of distributed estimation and net-work topology. And we then propose a sensor deployment scheme with coverage,connectivity and estimation requirement guarantees. In order to decrease the structureirrationality and link redundance, we further present a topology optimization schemecoupled with the tuning of estimation parameters. Simulation results show that theproposed deployment algorithm and topology optimization scheme can guarantee theperformance of the corresponding distributed estimation algorithms.