Exploring the Efficacy of Sparse, General-Purpose Sensor Constellations for Wide-Area Activity Sensing

Abstract

Future smart homes, offices, stores and many other environments will increasingly be monitored by distributed sensors, supporting rich, context-sensitive applications. There are two opposing instrumentation approaches. On one end is full sensor saturation, where every object of interest is tagged with a sensor. On the other end, we can imagine a hypothetical, omniscient sensor capable of detecting events throughout an entire building from one location. Neither approach is currently practical, and thus we explore the middle ground between these two extremes: a sparse constellation of sensors working together to provide the benefits of full saturation, but without the social, aesthetic, maintenance and financial drawbacks. More specifically, we target a density of one sensor per room (and less), which means the average home could achieve full coverage with perhaps ten sensors. We quantify and characterize the performance of sparse sensor constellations through deployments across three environments and 67 unique activities. Our results illuminate accuracy implications across key spatial configurations important for enabling more practical, wide-area activity sensing.