Analysis And Analysis Of Regression Modeling And Location Of Mobile Targets In Wireless Sensor Networks

Wireless sensor network (WSN) as a new kind of distributed measurement and control network technology has a very broad application prospects. The self-localization technology of Sensor nodes is one of the most important technologies during WSN which can support the location imformation of monitoring target with high Target tracking accuracy and routing efficiency. And it is also a key technical problem to be solved in the study of WSN application.Combined with the National Natural Science Foundation project, the progress and research of domestic and foreign target location technology of WSN is introduced in this paper. The method of WSN target location and Least Square Support vector Regression in the application and research of localization is discussed.The method of WSN target location prediction wake-up to reduce the loss rate is proposed. Main works of this paper are as follows:1.The WSN target location method based on LSSVR is in-depth researched in this system, and the characteristic and factors that influence of LSSVR three-dimensional nodes localization error are discussed.The structure model for target positioning error is created, several methods to reduce the error of LSSVR localization are proposed, and the spatial distribution characteristics of target positioning error of LSSVR is analyzed.The mechanism of nuclear the function positioning of the LSSVR target is discussed, the kernel function with simple form, less parameters and good prediction ability should be sllected.2. Local modeling is introduced, which includes the distribution of training samples, kernel function and the boundary coefficients selecting etc which make LSSVR local modeling method more practical.The relationship between different model parameters and LSSVR positioning error is pointed out.The positioning accuracy of LSSVR modeling method is improved by the use of particle swarm algorithm to optimize the parameters of the model.3. Node wake-up mechanism of LSSVR modeling positioning is researched, the numbers of measurement nodes Nd, awaken nodes Nw and the loss rate pm are deduced, several method of reducing loss rate by nodes wake-up are explored. The constructive propose several methods of improving the accuracy of target prediction are proposed constructively. In the low power state of measurement nodes, nodes wake-up mechanism of LSSVR modeling based on dynamic prediction is proposed, which can keep the target with low loss rate and improve the accuracy of target location.The results show that the application of the localization algorithm in the paper can realize the expected functions, the constructed LSSVR three-dimensional node positioning system is suitable for the nodes positioning applied research and has features such as high system positioning accuracy and good real-time performance.