<![CDATA[Localization is a challenging research issue in various sectors of activity, particularly in dynamic indoor environment with high perturbation. Many existing localization techniques in wireless sensor networks are not efficient because of many constraints such as the high cost, the memory and energy limitation and the environmental noise effects. Thus, the development of novel methods of localization has become a great concern for the wireless sensor networks. In this paper, we propose a tracking method that combines regression tree and Kalman smoother filtering. Previously, regression tree has been suggested for static positioning by means of received signal strength indicator measurements. In this work, we employ this strategy to solve the mapping relation between these measurements and the target position by means of an optimized propagation model. Moreover, the predicted position considered as the observed information is introduced to the Kalman smoother algorithm, to have more precise state of the moving target. The proposed algorithm has been assessed and compared to other existing methods using real measurements of the received power by the moving target in an indoor environment. The evaluation shows that our solution outperforms other methods regarding localization accuracy.]]>
