Research On Message Dissemination Algorithm Of Underwater Wireless Sensor Network Based On Link Prediction

With the continuous exhaustion of terrestrial resources,many countries have been increasing the development and utilization of various water resources(such as oceans,lakes,etc).Through monitoring the underwater environments,the exploration of underwater resources and prediction of the potential underwater disasters can be realized.In order to explore the underwater environment better,underwater wireless sensor networks came into being.However,due to the large attenuation of electromagnetic waves,the typical underwater communication relies on the acoustic waves,and the acoustic waves have some characteristics,such as low propagation rate,low bandwidth,and irregular channels.These characteristics make the message dissemination of underwater wireless sensor networks be difficult.At present,many communication protocols have been proposed for underwater wireless sensor networks.For underwater sensor networks composed of mobile nodes,the opportunistic routing protocol is the most common protocol.In the opportunistic routing protocol,how to choose the best relay node determines the performance of message dissemination.Link prediction is a technology used to study complex networks.It is mainly applied to analyze the evolution law of complex network structures.Through the analysis of nodes and their relations in the network,the future links can be predicted.In this thesis,we apply the link prediction method to dissemiante the data messages in underwater wireless sensor networks,i.e.,the historical links are exploited to predict the future links and find the most suitable relay node.In the link prediction process,the similarity indices for link prediction are considered to be the typical indices.However,the opportunistic routing protocol complies with the "store-carry-forward" manner,that is,the message dissemination on relay nodes is sequential.Therefore,the traditional similarity indices such as common neighbors are not applicable in underwater wireless sensor networks.In this thesis,based on the similarity index of common neighbors,the spatio-temporal features of message dissemianted are investigated,and thus the spatio-temporal common neighbor index is proposed.Underwater wireless networks have a variety of network structures: some underwater wireless sensor networks only have mobile nodes,and some contain two types of nodes: anchor nodes and mobile nodes.The different types of nodes cause the different message dissemination strategies.This thesis design suitable message dissemination strategies for these two different network structures,respectively.With regard to the network structure that there are only mobile nodes in underwater wireless sensor networks,we modifty the Jaccard similarity index into the spatio-temporal Jaccard similarity index to calculate the possibility of the future encounters(links)between neighbor nodes and the destination node,so as to select the sutiable relay nodes.With regard to the network structure that there are mobile nodes and anchor nodes in underwater wireless sensor networks,we combine spatio-temporal Salton similarity index and spatio-temporal Jaccard similarity index to design different message dissemination strategies for the anchor nodes and mobile nodes,respectively.Especilaly,due to the large communication range and strong computing power of the anchor nodes,they can play the role of edge computingservers to improve the link prediction results.Besides,the term of link prediction deviation is measured,and the data message to be are dissemianted based on the link prediction results and the prediction deviations.The simulation experiment results show that the two proposed algorithms for different network structures can increase the message delivery ratio and reduce the dissemination delay while the number of disseminated copies of data messages can be limited.Applying these algorithms to underwater wireless sensor networks can provide available message dissemiantion schemes for various underwater applications.In the future,we will focus on the algorithm improvements in terms of energy consumption balance and reasonable node layout.