Wireless Sensor Network Localization Based On Range-free Technique

Wireless Sensor Network (WSN), as a new-generation technique for information acquiring and processing, can be wildly employed in many areas such as Intelligent Transportation System (TIS), national defense, environment supervision, medical treatment, out space exploration and so on. It is thought as one of the most important techniques in 21th century. Wireless Sensor Network Localization is one of key supporting techniques in WSN, and has widely attracted attention of domestic and foreign scholar.My research mainly focuses on three rang-free localization algorithms using hop-counter: Localization Relevance Vector Machine (LRMV), Localization Sparse Representation Classification (LSRC) and Multi-Corona Division (MCD).LRVM is improvement of previous Localization Support Vector Machine (LSVM) algorithm; LSRC is based on the emergence of Compressive Sensing (CS) theory and machine learning technique. MCD is more like circle division. The main detail is as following:(1)For solving problem of redundant localization information in LSVM, we proposed LRVM algorithm based on Relevance Vector Machine (RVM). RVM, a learning machine technique based on Bayesian framework, considers sparsity of localization information, and furthermore largely reduce the redundancy of information. RVM is non-parameter unlike Support Vector Machine (SVM), therefore LRVM is more adaptive in a rapidly changeing network than LSVM.(2) LSRC resembles LSVM,due to same concept of using machine learning to classify. However, before calculation the localization information, compressing it for reducing the redundancy is unique in LSRC.While in LSVM and LRVM after algorithm have been executed, redundant information is clear.(3) MCD is based on circle division.In previous algorithm such as Concentric Anchor-Beacons (CAB) using circle division, the anchor node usually equipped some special device such as direction antenna or multi-level power emitter. The MCD's anchors do not require these devices. And simulation experiment demonstrates MCD has higher localization accuracy in a large scale network than CAB and LSVM.