Using tomography for ubiquitous sensing

Abstract

By embedding sensors in mobile devices, it is possible to exploit the ubiquitous presence of these devices to construct applications for large-scale sensing and monitoring of environmental phenomena. To this end, we present Environmental Tomography, a novel approach in which mobile devices participate in the collection of aggregate sensor readings along roads or sidewalks, and these aggregates are used to reconstruct an estimate of the contaminant distribution throughout a region. We demonstrate how our data collection process preserves user location privacy and is robust to sensor and location reading errors. We also show how the estimation process can be formulated as a convex optimization problem that incorporates the physical dynamics of the phenomenon of interest. We study the performance of Environmental Tomography using various road network layouts and realistic models of pollution. Results indicate that estimates generated from path aggregates are of comparable accuracy to estimates generated from significantly greater numbers of individual sensor readings.