A Sampling-Based Approach for Achieving Desired Patterns of Probabilistic Coverage with Distributed Sensor Networks.

A new method is derived for finding the best positions in which to locate the sensors in a distributed sensor network in order to achieve a desired variation, or pattern, in spatial coverage over a specified domain. Such patterning is important in situations when there are not enough sensors to completely cover a region adequately. By providing coverage based on a desired pattern, this approach allows a user/designer to specify which sub-regions of the domain are more important to cover, and to what level that is desired. The method that is developed is novel in that it is an analytic approach, as opposed to existing numerical optimization approaches, and thus provides solutions rapidly and can also be applied to provide online repositioning for existing sensor networks to respond to changes in the environment. The method is based on deriving an expression for the probabilistic density of sensor locations that best matches the desired coverage under given spatially varying environmental conditions; and then samples from that sensor density to determine specific sensor locations. The performance of the method is demonstrated on numerical examples in both one-dimensional and two-dimensional settings. Comparisons are made between solutions found from this approach and solutions obtained by a numerical optimization technique.