Gradient-based target localization in robotic sensor networks

Fast target localization without a map is a challenging problem in search and rescue applications. We propose and evaluate a novel gradient-based method which uses statistical techniques to estimate the position of a stationary target. Mobile nodes can then be directed toward the target using the shortest path. Moreover, localization can be achieved without any assistance from stationary sensor networks. Simulation results demonstrate nearly a 40% reduction in target acquisition time compared to a random walk model. In addition, our method can generate a position prediction map which closely matches the actual distribution in the field. Finally, experiments have been performed using MicaZ motes which further validate our techniques.