FailureSense: Detecting Sensor Failure Using Electrical Appliances in the Home

Abstract

With the proliferation of inexpensive sensors, sensors are increasingly being used in smart homes. Recent experience on long term sensor deployment in residential homes has identified the potential risk of various types of sensor failure. Motivated by real examples, we develop new schemes to detect not only fail-stop failure, but obstructed-view and moved-location failures that are not the traditional fault detection foci. Our proposed solution, FailureSense, uses a novel idea of using electrical appliances to detect sensor failure at home. We learn the regular patterns of sensor firing with respect to appliance activation events and report a failure when we observe a significant deviation from the regularity. By using data from three real home deployments of over 71 days and 2818 recorded turn on and off events of 19 monitored appliances, we observe that FailureSense can detect obstructed-view, moved-location, and fail-stop failures with 82.84%, 90.53%, and 86.87% precision, respectively, with an average of 88.81% recall.