Sensor and actuator networks are often installed in buildings for energy-related applications such as lighting and climate control. Such systems require metadata about the deployed hardware (e.g. which room each is in, what the function of each room is) in order to operate effectively. In this paper we present methods to automatically determine such metadata, in particular the room connectivity graph (i.e., which rooms share a doorway/interior window). Crucially, our method works with just one sensor unit per room, does not require special placement of any of the sensors, and can therefore work on data from existing widely-deployed applications (such as burglar alarms). We apply this method to a 30-day data set from single per-room sensor units deployed in two residential homes in the United Kingdom. Room connectivity is determined based on: spillover of artificial light

between rooms; occupancy detections due to movement between rooms; and a fusion of the two. The fusion of both techniques is shown to work better than either technique alone, with a 93% true positive rate and 0.5% false positive rate (aggregate across both houses), and a convergence time of under a week.

PDF file

In  Proceedings of BuildSys 2012

Publisher  ACM
Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee.

Details