Mobile Wireless Sensor Networks Localization Based on Support Vector Machines

It is necessary for some of the applications of wireless sensor networks (WSNs) to estimate the geographical position of nodes. For the localization of wireless sensor nodes in wireless sensor networks, we took advantage of the learning machine approach based on binary support vector machine (SVM). During offline learning process, the received signal strength from reference nodes was chosen as learning machine input. By dividing the sensor field into several regions the geographical position of reference nodes was used as class information in SVM, and during online localization process the decision function of SVM was used for estimating the blindfolded nodes. In our suggested approach, the mobility of nodes in the sensor field was taken into consideration, and the mobility of nodes for reducing the reference node and increasing the localization accuracy was used. The expected efficiency of our approach was offered using simulation.