Location Estimation of Sensor nodes in 3-d Wireless Sensor Networks Based on Multidimensional Support Vector Regression

Wireless sensor networks have been assumed as a valuable instrumentation for several location-dependent applications such as health care, geometrical routing, and military. Among many localization algorithms, range-free localization is more preferable because of its cost and implementation efficiency. Most of these algorithms are applied to two-dimensional wireless networks which do not truly model the real wireless networks. In this paper, a range-free localization method based on multidimensional support vector regression (MSVR) is applied to three-dimensional wireless sensor networks. To solve estimation problem, the location of sensor nodes is formulated using a nonlinear regression model. The accuracy of proposed method is evaluated for isotropic and anisotropic networks. The method has minimum localization error among range-free localization algorithms. Hence, simulation results clearly show that the performance of aforementioned method is better than other pioneer algorithms for such extended networks, and is insusceptible to the ratio of anchor nodes and network density. These results indicate the method effectiveness for three-dimensional networks for all sensor network deployments and its robustness to range error.