Distributed Gossip Algorithms For Quantized Consensus

Wireless sensor network(WSN) has attracted more and more researchers inrecent years.It is composed of a large number of low-cost microsensors nodes andformed a multi-hop self-organizing system through wireless communication mode.Itcan real-time collaborate,collect,transmit and process perceived object informationin areas and send data information to the relevant nodes.WSN has low cost, lowpower consumption, high fault tolerance, strong real-time performance,etc.Therefore,WSN has great applicaion value,which can be used in militaryapplication,medical health,building security monitoring and other aspects ofapplication.Gossip algorithms are simple and efficient, similar to the spread of viraltransmission algorithms.They are kinds of communication mode through choosing anearby node for exchanging information communication according to theirrequirements. Instead of the end-to-end routing, Gossip algorithms have advantagesof decentralization, strong expansibility and so on. Gossip algorithms are moresuitable for distributed wireless sensor networks because congestion and link failurecan be efficiently resisted.In this paper, we develop distributed gossip algorithms for quantized consensus.In the distributed systems such as wireless sensor networks, distributed consensusproblem based on Gossip algorithms has been widely used. There are many researchresults of pairwise gossip algorithms, which have slow rate of convergence and onlyavailability in symmetrical link. But in recent years, the proposed broadcast Gossipalgorithms can make full use of the radio characteristics of wireless channel. At thesame time, They are suitable for asymmetric-link networks thereby improving theconvergence speed greatly. However, in iterative computation process of Gossipalgorithms,there will usually be a decimal even irrational number. Therefore,quantization for broadcast Gossiop algorithms has significant use value.We first introduced the work process, characteristics and applications of Gossipalgorithms. Respectively for classical pairwire gossip and broadcast gossipalgorithms,the detailed working process, advantages and disadvantages ofcomparison are described.Then we briefly introduce several kinds of commonlyused quantitative methods,and mainly described the broadcast Gossip algorithms ofdithered quantization.There are detailed theoretical derivations and anaylysis fromthe mathematical expectation and second order matrix for them.The result shows that quantitative operation will not change the original convergence of Gossipalgorithms.So broadcast Gossip algorithms of dithered quantization have importantvalue for practical applications.