Research On Distributed Estimation Over Network Structure In Wireless Sensor Networks

In distributed estimation algorithm,sensor nodes need collectively estimate some parameters of interest from noisy measurements.For large-scale systems,the distributed estimation method is flexible and easy to operate.The distributed processing method can effectively speed up the data processing and improve the robustness of the algorithm.In the literature,the incremental,consensus,and diffusion strategies are among the most popular propositions.The diffusion-based algorithms outperform their consensus-based and incremental-based counterparts.The information of Wireless Sensor Network(WSN)obtained from spatial perspective has greater signal-to-noise ratio,and can improve the estimation accuracy and robustness by processing large information.The unique topological structure characteristics of wireless sensor network are very important for many applications.Wireless sensor network is the structure foundation of distributed estimation and distributed estimation algorithm is the core of the distributed estimation.Therefore,it is necessary to combine the network structure with the distributed estimation algorithm to overcome the problem of estimation and get better estimation performance.The diffusion least mean-square algorithm(DLMS)is a typical distributed estimation method.Because DLMS has the characteristics of simple structure,easy implementation,stable performance and strong robustness,it has been widely used in many fields.However,DLMS algorithm also has some shortcomings.Every node will receive and transmit data to its neighbor nodes,which is a considerable communication burden.In order to reduce the inter node communication burden of DLMS,this article explores the impact of motif structure in wireless sensor networks on the performance of the DLMS algorithm,including size-3 motifs and size-4 motifs.It is found that the communication network containing motifs with closed triangles will hinder the performance of diffusion LMS algorithm,therefore we try to avoid such motifs.Therefore,we proposed a broken-motifs diffusion LMS algorithm(BM-DLMS).BM-DLMS algorithm can efficiently reduce the communication load with less performance degradation.This study provides some guidelines on how to save energy and bandwidth in network.This article also incorporates the diffusion strategy into phase estimation,and combines alternating iterative minimization method,then develops a distributed phase estimation algorithm over sensor network.The proposed algorithm can produce accurate enough results and has low communications.In addition,we explore the impacts of network topology on the performance of a distributed phase estimation algorithm.The different network models of WSN,including regular network,small world network,random network and scale-free network are covered.It is found that the performance of the proposed algorithm is quite different from that of the different network models.This research is very important for designing a topology of sensor networks such that the estimation performance are optimized.