Smart Power Plugs for Efficient Online Classification and Tracking of Appliance Behavior

Smart power plugs are notable cyber physical systems that can track and control appliance behavior. Traditional smart plugs provide limited functions, often only capable of recording consumption data, without advanced automation features. In future smart homes, sophisticated functions are expected to be provided by smart plugs, such as online learning, classification, and diagnosis of appliance behavior. Instead of relying on centralized cloud servers for providing advanced functions, the paradigms of edge and fog computing are increasingly populated by the applications of Internet-of-things, which aim to provide timely intelligent processing using small local memory space in a standalone manner. In this paper, we develop standalone smart plugs that are capable of providing efficient online classification and tracking functions on the continuous power signals of appliances. We built and implemented our smart plug systems using low-cost Arduino platform with a small amount of memory space. In general, our standalone online classification and tracking systems can be applied to a variety of smart sensors for wearable, biomedical, and environmental monitoring applications.