Tenth International Conference on The Constructed Environment

Abstract

The paper presents a Bayesian method for determining occupancy in a home using motion and presence detectors. Data were gathered from several locations and then analyzed to determine occupancy. Data were collected from a set of PIR motion detectors and then compared against a GPS and Bluetooth-based presence detector, which provided actual truth to verify the results of the PIR motion events. These PIR motion events were then used to build Bayesian network conditional tables. Once the conditional tables were constructed, the Bayesian network results could be compiled and then compared with the actual occupancy information, which was gathered from the presence detectors. The results indicated that the Bayesian modeling demonstrated an improvement in occupancy detection using the historical probability of the PIR motion sensors over the independent and grouped methods that were tested. This resulted in improved occupancy detection. Future research in the application of historical treading should be designed to explore detection improvement methods in detail.