A NSGA-II based sensor selection scheme for target tracking in wireless sensor networks

In this thesis, we study the sensor selection problem in target tracking for a wireless sensor network (WSN). The target emits energy and the sensors transmit their measurments from the target to the Fusion Center (FC). FC estimates the location of the target by using these measurements. Since a WSN may have limited resources, it is critical to gather measurements only from the most informative sensors rather than all the sensors in the WSN. Our aim is to find the sensor selection strategy at each time step of tracking by the joint minimization of objective functions representing the estimation error and total number of sensors transmitting to the FC, where we use a Non-dominated Sorting Genetic Algorithm - II (NSGA-II) to determine the solutions between the two conflicting objectives. Different from the existing results in the literature, our aim is to get the solutions of NSGA-II accurate and fast by setting right parameters. Firstly, rather than randomly initializing the initial population of NSGA-II at each time step of tracking, we use the solutions of the previous time step in the initial population of the current time step. Secondly, rather than executing NSGA-II for excessive generations to observe the near Pareto-optimal front, we define a stopping rule by using the Generational Distance metric. We further compare the solutions proposed Multi-objective optimization problem under different population sizes and crossover operators as well as under target trajectories with different process noise parameters, and different total number of sensors in the WSN.