How valid is the regularly-spaced grid assumption in RSS source localization sensor networks?

Received signal strength (RSS) based wireless sensor networks are often employed for position tracking. Sensor deployment is generally modeled as either a regular grid or as a random, uniform distribution for each sensor. This paper provides a theoretical comparison between the geolocation performance of a regular grid and a randomized grid. We show that the Cramer-Rao bounds are comparable, and for typical network parameters, realizations of the bound of the random grid are within a factor of 2 of the regular grid, with at least 75% confidence.