Bandwidth-efficient dynamic event region detection and reconstruction in wireless sensor networks

In this paper, the problem of dynamic event region detection and reconstruction using a wireless sensor network is studied. The spatiotemporal relationship of the evolving field is assumed and modeled by dynamic Markov random fields. Observations are collected from a network of sensors distributed in the field. To meet the stringent bandwidth and energy constraints in wireless sensor networks, only quantized observations from a subset of sensor nodes are allowed to be transmitted to the center. To reconstruct and track the field state map at a control center at each time step, a processing framework including sensor selection, local observation quantization and central processing is proposed. The performance of the proposed processing framework is analyzed through simulations and the effectiveness and efficiency is demonstrated by comparing with other methods.