Abstract
WiFi connectivity is ubiquitous nowadays, specially in the new era of Internet of Things (IoT), where the majority of physical devices, home appliances, and vehicles have some kind of network connectivity. On the other hand, recent developments in wireless technologies have transformed the role of wireless signals from a pure communication medium to an enabling tool for non-intrusive sensing. Radio signals propagate along multiple paths and reflect from objects before arriving at a receiver, so they carry information from the environment. In this thesis, we exploit the traditionally challenging multipath propagation and convert it into an opportunity for human sensing, device localization, and object tracking by mapping each wireless reflection to relevant physical and behavioral measurements. Beyond leveraging the pervasive wireless infrastructure, the major breakthrough enabled by this thesis is our innovative approach of unilateral sensing, in which a single WiFi device unilaterally senses the physical world without requiring coordination or data sharing with any other devices. This, in turn, converts every WiFi-enabled device into an individual sensor that learns about the environment, leading to a scalable sensing platform.
This dissertation delivers four fundamental contributions. First, it presents a novel localization approach called Multipath Triangulation, which combines the geometric properties of wireless multipath signals to triangulate WiFi devices and reflection surfaces. Next, the multipath triangulation is exploited to produce the first decimeter-level unaided localization system that requires only a singleWiFi receiver to unilaterally locate any otherWiFi devices in the room. Beyond localizing WiFi devices, we further extend multipath triangulation to develop the first WiFi-based object tracking system that can localize the passive wireless reflections from a battery-free tag in the presence of complex multipath propagations. Finally, we demonstrate that multipath reflections provide peripheral WiFi vision for sensing the presence of people in a room, even if they are stationary, without requiring them to carry any devices or wear a tag.
To deliver these contributions, we employ the underlying physical properties of wireless multipath propagation and map the frequency, temporal and spatial characteristics of these signals to the physical environment. We implement new systems and algorithms that are compatible with commodity WiFi devices, which are also evaluated in regular indoor environments. A broad range of applications benefits from this sensing information including health and elderly monitoring, home automation and security, or search and rescue missions. We believe that these approach becomes a necessity in the near future as IoT devices become even more ubiquitous and context-aware services such as home well-being monitoring, robot assistants, and autonomous driving turn into daily life routines.
