Research On Localization, Data Dissemination And Collection For Wireless Sensor Networks In Military Applications

Wireless sensor networks (also known as sensor networks, or WSNs) are the net-works that consist of multiple sensor nodes. These nodes are self-organized with wirelesscommunications. Sensor networks have a number of advantages, such as good obscurity,easy deployment, self-organization, low cost and invulnerability to bad weathers. Theywill be used to collect all kinds of battlefield situation information to achieve informationdominance, and play an important role in future military information system. However,the research on sensor networks has just started recently in our army. Many publicationsonsensornetworksfocusoncivilapplications. Ifthetechnologiesincivilapplicationsarecopiedtomilitaryones, theymaynotperformwellasanticipated. Thisthesisanalyzestherequirements and challenges of sensor networks in military applications, and focuses onsome associated technical problems, including localization-oriented network adjustment,data dissemination, non-real-time and real-time data collection. In order to reduce theenergy consumption, improve the network lifetime, adaptability and reliability, and con-solidate the real time property of data routing from sensors to users, this thesis researchthe aforementioned four problems, respectively.The thesis first proposes a localizability-aided network adjustment method namedLAL, after the execution of which all nodes can be localized. Localization is an importantservice in sensor network applications, however, no localization algorithm can guaranteethat all nodes in the network be localized. Facts in both working sensor networks and ex-tensive simulations have shown that part of the network may not be localizable in theory.To localize those nodes, the primary solution is to change the network configurations soas to make the whole network localizable. Previous adjustment methods areconsidered tobecoarse-grained, sincetheydonotdistinguishthelocalizablenodeswithnon-localizableones, which induce much unnecessary adjustment and accompanying overhead. In con-trast, LAL utilizes node localizability information, improves the efficiency of adjustmentandreducesdistancemeasurementoverheadandenergyconsumptioninlocalization. Thisalso helps to extend the network lifetime. After the adjustment of LAL, there may still ex-ist some unnecessary edges in the network. To approach optimal results, the thesis bringsforward a path-based adjustment method. With some additional hypothesis, the methodcan constrain the number of edges added into the network below 2|V_N|, where |V_N| is the number of non-localizable nodes in the network.Therequirementsofmilitaryapplicationsmayvaryasthebattlefieldsituationchanges.This motivates the need to disseminate configuration parameters, so as to control the op-erations of sensor nodes quickly and conveniently. The key of dissemination is how tokeep the consistence of the operations of all nodes. According to the variance of systemrequirements, more than one rounds of dissemination may be conducted in the network.In this case, users must assure that data items on different nodes must have the same ver-sion number. The thesis proposes a Bloom filters based dissemination protocol namedBDP for sensor networks. BDP uses Bloom filters as the compact data structure to storemetadata of all data items, and can find the version difference between two items withthe same key on different nodes, so as to accelerate the process of updating. Through theanalysis of node interactions and the detail of protocol implementation, the thesis pro-poses solutions for controlling the false positive rate, which can constrain the probabilitywithin a relative small value. So the protocol can guarantee the version consistence ofdata items throughout the network.In applications that does not have real-time requirement, or good deployment condi-tions are not available, users can deploy the sensor networks in unattended area. After aperiod of time, users can use mobile devices such as unmanned aerial vehicles to collectthe sensing data. As far as military applications are concerned, there are two challengeslies in the aforementioned scheme: first, sensor nodes are prone to damages, which re-sult in the lost of data; second, the data collection process is likely to cause unbalancedload distribution, and some nodes will soon exhaust their energy and stop working. Toovercome the former challenge, the thesis proposes a virtual node based distributed stor-age method, which uses Fountain code and random walk to distribute data throughout thenetwork. The method not only holds the advantage of network coding, but also make fulluse of all the sensor nodes for sensing and storage. This significantly improves the effi-ciency of nodes utilization. Even if most of the nodes are damaged, all raw sensing datacouldalsoberetrievedbydecodingtheremainingcodeddata. Forthelatterchallenge, thethesis proposes a data collection method that utilizes mobile sink with single-hop com-munications. The method first builds a dominating set in a distributed way. Then the basestation finds a relatively short path and passes all dominating nodes. These dominatingnodes act as checkpoints, at whose location the base station collects all the sensing data of neighbor nodes. In contrast to previous works, the method has very low complexityand good practicability. It not only significantly reduces the communication overhead ofmultihop routing, but also balances the load of all nodes, which avoids the existence of"hot spot"with high load.In real-time applications, nodes usually use duty cycling to extend the lifetime ofsensor networks. In duty cycling scheme, each node periodically switches its radio onandoff, soastogetatradeoffbetweenreal-timeperformanceandenergyconsumption. Inmost cases, nodes schedules are not synchronized. Most existing data collection methodsforasynchronousdutycyclingmainlyrelyontimesynchronizationmechanism,whichnotonly consumes precious program volume, but also brings in additional communicationoverhead. Moreover, the performances of those methods rely heavily on the accuracy andstability of time synchronization protocols. This thesis tries to remove the dependenceof time synchronization protocols, and proposes a method that uses multihop negotiationand collaboration to discover shortcuts constituted by a sequence of active nodes. Theseshortcuts are used to pass messages. The method reduces the latency of data collection,so as to improve the real-time property.