Study Of Localization In Mobile Sensor Networks Based On Distance Refinement

Wireless sensor networks (WSNs) are multi-hop and self-organizing networks thatconsist of the collection of distributed nodes, each of which has the ability of sensing,computation and communication. WSNs can mainly be categorized into two differentsubsets depending on the mobility of the nodes: static sensor networks and mobile sensornetworks. In WSNs, sensors can monitor and collect the information of the targets in realtime and send the data to the end users. WSNs have been widely used in the applications ofmilitary, national economy and public management. Localization is one of the primaryissues for wireless sensor networks and most of localization algorithms are based ondistance estimation. This dissertation focuses on the problems of distance estimation andlocalization in wireless sensor networks with various topologies. It studies the localizationtechnology via theoretical analysis and simulated way in order to propose methods withhigh efficiency and precision. It also presents some new ideas for static networks andmobile networks respectively.First, the localization problem in wireless sensor networks is analyzed deeply. Thisdissertation summarizes the existing algorithms in static networks and mobile networks.For the static networks, the methods based on multidimensional scaling technology andVoronoi diagram are highlighted. It is proved that the distance estimation errors can lead tothe low efficiency and precision. For the mobile networks, the path planning of the mobileanchors and Monte Carlo method are discussed.In the static networks with holes, the irregularity of the network will cause distanceestimation errors. The virtual hole is introduced in order to refine the distances betweenpairs of nodes. Based on the multidimensional scaling technology, a centralized algorithmcalled multidimensional scaling via distance optimization (MDS-DO) is proposed forsolving the problems of MDS-MAP algorithm in irregular sensor networks. The resultsshow that the performance of MDS-DO is much better than that of the classical method fordifferent node quantities and connectivity. The localization precision can be dramaticallyimproved by the proposed MDS-DO approach. For the direct ranging error caused by measurement in static networks, a Voronoibased and geometric constraint assisted localization algorithm (VBGCA) is proposed toaddress above issues. In VBGCA, Cayley-Menger determinant is introduced to formulateconstraints of the errors. When there are multiple anchors, a linear constraint equationbased on four anchors is derived. Combined with the quadratic constraint, the feasibility ofthe distance constraint can be theoretically proved. Simulation and experimental resultsshow that VBGCA can effectively avoid the localization failures in classical Voronoi basedalgorithm and decrease the localization error.Furthermore, sensor networks with mobile anchors are studied and analyzed forsolving the problems of distance estimation and localization. Mobile anchors are not onlyused to provide more reference information, but also developed for distance refinement. Forthe anisotropic networks with holes, a distributed method is introduced to explore theboundary of the holes firstly. Then some nodes on the boundary of the holes are marked bythe mobile anchors to refine the distance information. For normal networks, a mathematicaloptimization model is proposed which introduces the application of geometric constraintsto refine these distances. Then a neighbor anchor constraint distributed localization scheme(NAC-DL) is proposed in which mobile anchors move to the vicinity of the sensor node tomake the distances to anchors more accurate. Simulation and experimental results show thatthe imprecise distances can be refined and the localization precision is improved. Finally, adynamic path planning of the mobile anchors is proposed. It approaches better performancein the length of the movement path and the number of the packets broadcast by the anchors.Numerical simulation results show that the novel path planning remarkably reduces theenergy consumption.