Radar Sensing Systems for Smarter Indoor Human Environments
Kalyanaraman, Avinash, Computer Science - School of Engineering and Applied Science, University of Virginia
Whitehouse, Cameron, EN-Comp Science Dept, University of Virginia
Today, on average, humans spend over 90% of their time indoors. Given the considerable amount of time spent indoors, sensing context in these human-centric indoor environments can enable several applications in healthcare, energy, elderly monitoring, personal welfare, etc. Despite the potential slew of applications, people are resistant to having cameras and/or microphones in many indoor human environments such as homes, office cubicles or automobiles, due to privacy concerns. Furthermore, this context sensing must be performed with no onus on the human to wear/carry any device, owing to the well documented forget to wear, forget to charge problem.
Given these sensing constraints, this dissertation explores the usage of wireless radar signals to sense context in indoor human environments. However, quite unlike their primary usage in outdoor spaces, radars in indoor environments are subject to (a) very strong indoor multipath reflections, (b) power consumption constraints, (c) space constraints, (d) transmit power regulations, and (e) an incomplete observation of the entire sensing region (i.e. partial field-of-view).
We address these challenges by building a suite of hardware and software solutions that builds on past radar literature to sense context in two indoor human environments — homes and cars, by leveraging the structure of the environment. In particular, the components of FormaTrack and Doorpler perform room-level localization of home occupants using radar sensors mounted atop room transition spots such as doorways via the Doppler Effect. To prevent sensing errors caused by these devices from becoming tracking errors, we build TransTrack – a tracking algorithm that uses sensor data from (subsequent) doorway events. Finally, we build CaraoKey, a system that repurposes the radar setup that pre-exists in automobiles for keyless entry, as a sensing modality to infer the state of a car. It does so in a manner that is robust to location changes and does not warrant any form of transceiver synchronization. The dissertation concludes by pointing out other context sensing applications that can be enabled by this radar infrastructure. Such features will become increasingly important as driverless cars and shuttles become the norm, resulting in an increased importance to passengers’ sense of in-vehicle security and well-being.
PHD (Doctor of Philosophy)
Wireless Sensing, Radar Sensing, Smart homes, Smart automobiles
All rights reserved (no additional license for public reuse)