The coupling of sensors with mobile phones, which are ubiquitously available and location aware, opens the door to the creation of applications for pervasive sensing and detailed spatial modeling of environmental phenomena. In order to ensure widespread participation of mobile users, these applications must have limited per-device resource requirements, must place no expectations on individual user behaviors, and must be sensitive to user privacy concerns.
In this work, we introduce Environmental Tomography, a novel approach to environmental sensing and spatial data modeling that meets the challenges of the mobile network. Environmental Tomography consists of two phases, a data collection phase in which the network of mobile devices computes aggregate values of sensor readings along roads and sidewalks, and a reconstruction phase in which the aggregates are used to generate an estimate of the distribution of the sensed phenomenon. The data collection process is robust to the dynamics of the mobile network and protects the location privacy of the participating device users. The reconstruction phase can be posed as a convex optimization problem with an objective function that takes the physics of the underlying phenomenon into account to produce accurate estimates from the limited available data. We verify the validity of our approach through extensive simulations using physically accurate models of environmental phenomena.