Wireless sensor networks formed using unattended ground sensors promise to provide an effective low cost solution for observing a physical phenomenon. Due to the energy limitation of WSN , for such a network to be effective over an extended time, power conservation is an absolute necessity. On the other hand, due to the high density in the network topology, it may not be necessary to active all the sensors in each time instance. This work is concerned with the selection of a subset of sensors for a target tracking.
To do so, it is assumed that the locations of all sensors are, a priori, known and the predicted target state obtained from the tracking algorithm is used to approximate the target position. Therefore, the criterion for sensor selection is defined by considering the location of all the sensors relative to the target position. Accordingly, a cost function is proposed based on the spatial correlation obtained using the best estimation of the event source. Then, another cost function is derived using the geometrical dilution of precision (GDOP) metric for power measuring sensors. As a result, a sensor selection algorithm is proposed which adaptively determines the number of active sensors and finds the best active set topology for target tracking.
The sensor selection methods are evaluated in terms of event distortion, their RMS errors of the target position, energy consumption, percent of intersection between the active set selected by each approach to the best active set and the execution time. First, It is shown that the distortion achieved by the proposed spatial-based cost function and its corresponding minimum number of active sensors for a given distortion constraint was less than those obtained using the suboptimum distortion function. Moreover, the proposed distortion function is less sensitive to the sensor density and range parameter. Also, simulation results are revealed that the performance of the sensor selection algorithm evaluated by the MSE and also computational burden has been improved compared with other algorithms.