Research On Localization Of Wireless Sensor Networks

Since the 21 st century,the technology of wireless sensor networks(WSN)is becoming more and more mature,in the era of information and intelligence,the application prospect of WSN is very broad.With the development and application of wireless sensor networks,as the key technology of wireless sensor networks,the node localization has gained the widespread concern from academia and industry.Related researchers have put forward many different types of localization algorithms that always have certain limitations such as high hardware costs,low localization accuracy and high complexity.Therefore,it is necessary that continuing research on the localization algorithm of wireless sensor networks.The various application environments of wireless sensor networks always require both high localization accuracy and low costs.The localization algorithm is divided into two categories according to whether it depends on the ranging device,that is,the range-based and the range-free localization algorithm.The existing range-free localization algorithms have accuracy limitation and range-based algorithms have the problem of costs that leads to difficult of practical application.The main content of this thesis is followed:(1)The definition,application background and node positioning technology of wireless sensor network are discussed in detail.The existing mainstream localization technology theory is studied,and several classical range-free localization algorithms are simulated and analyzed.(2)Combined with the research of machine learning theory,the theory of multiple localization algorithms based on support vector machine is studied in detail.Compared with the classical DV-hop positioning algorithm,the weakness of existing machine learning related localization algorithms are obtained,that provide an idea for improving localization algorithms.(3)An improved localization algorithm based on ensemble learning is proposed by using gradient boosting decision tree(GBDT),meanwhile,a related parameter strategy is proposed.The algorithm uses the relevant information of each beacon nodes to training prediction model,and each unknown node uses its hop vector as input of these models to obtain every distance between itself and beacon nodes.The simulation result show that the method has lower localization error and acceptable time complexity,which means improvement in localization algorithm of wireless sensor networks based on machine learning.