Node location estimation scheme in wireless sensor networks based on support vector regression

In view of the issue that the accuracy of the node location of wireless sensor networks (WSN) is low by adopting maximum likelihood estimation (MLE) in estimating the measurement information value with big noise, a new node location estimation scheme based on support vector regression (SVR-NLE) is proposed. Through learning the relation between the real value of trilateral and node coordinate, this method utilizes the generalization capability of SVR (support vector regression) to achieve better location on the same noise level. The experiments choose LS-SVR (least squares SVR) and epsiv - SVR ( epsiv -insensitive SVR) to estimate the location of 100 randomly distributed unknown nodes. The result indicates that this new method can improve 15-20% location accuracy than MLE.