| The main problem studied in this dissertation is the tracking of targets by particle filtering in wireless sensor networks, where the sensors measure received signal strength (RSS). The number of targets is assumed unknown and it may vary with time. The initial locations of the targets are also unknown as are the time-varying reference powers of the targets. The task of tracking multiple targets under the made assumptions is a very challenging task for several reasons. First, the initial localization can be complicated. Second the received signal is a superposition of possibly several sources (targets) whose number is unknown. Third, the reference powers of the targets can also be unknown. Fourth, model selection has to be implemented at every time instant in order to choose the number of targets present in the network.;In the proposed tracking, we apply the least squares (LS) method for quick initialization of the procedure. This includes initial localization of the targets and estimation of their reference powers. Basically, we solve a nonlinear LS problem by using an iterative method. We also implement a criterion for deciding the number of targets in the sensor field. Once the number of targets and their initial parameters are determined, we apply particle filtering for performing the tracking. The algorithm allows for decrease or increase of the number of tracked targets and quick initial localization of the newly detected targets. |
