My research focuses on building non-invasive sensing platforms for smart homes and using data from these sensors to perform unsupervised inference of resident activities, identities, and resource usage patterns. My advisor is Professor John Stankovic, and I also collaborate significantly with Professor Kamin Whitehouse. I have broad interests in cyber-physical systems, machine learning, wireless networking, and ubiquitous computing.
My most recent work is on disaggregating whole house power and water meter data to infer the resource usage of individual light and water fixtures in the home. In contrast to existing disaggregation approaches, our approach is unsupervised and requires minimal cost or effort investment in sensor installation.
To infer resident identities in the home, I have led the development of a cheap, accurate, and minimally invasive biometric identification system in homes using ultrasonic height sensors above the doorways in the home.
To infer resident activities, we have developed accurate, unsupervised algorithms that leverage cross-home activity models to infer detailed activities such as cooking, showering and toileting with very little information about the types and locations of sensors in the home. Our unsupervised approach also has implications for wireless privacy of future smart homes.
I have also worked on automatic, intelligent HVAC control in homes, underlying design guidelines for smart home sensor deployments, and in the distant past, on multicast protocols for mobile ad hoc networks.
Check the research projects and publications pages for more information!