A first look at the hidden cost of random node placement in wireless sensor networks

While wireless sensor networks are often envisioned to be composed of a collection of nodes scattered inadvertently in an environment, the implications of random node placement in terms of system resource requirements remain mostly unexplored. This work studies resource consumption in wireless sensor networks in the context of decentralized detection. It is argued that, for most detection problems, the number of bits necessary to capture most of the information contained in an observation is small and insignificant when compared to the communication overhead associated with data transfer. Message compression therefore plays a limited role in minimizing resource consumption at each node. The communication rate and the positioning of individual nodes are likely to have a much greater impact overall. In particular, it is shown that for typical dispersive environments, the random placement of wireless nodes may result in an average transmit power that is 6 to 24 times the power required to communicate on a regular lattice. This excessive power consumption may affect the coverage of a system, its life expectancy, the probability of error at the fusion center, and the number of nodes required for a system to operate properly. These preliminary results underscore the importance of planning in the deployment of sensor systems