Zoom Meeting: https://umass-amherst.zoom.us/j/93426387182
Abstract
Over the past decade, there has been extensive work on understanding human mobility at urban scales and on modeling such mobility by using a variety of data sources. This body of work has analyzed or modeled mobility patterns at a single spatial scale-often that of the underlying dataset--and has not considered the impact of mobility at different spatial scales on system design. Additionally, it assumes mobile devices to be independent, an assumption that no longer holds in an era of mobile Internet users who own a multitude of devices that exhibit correlated mobility patterns.
In this defense, we will first highlight work on characterization of human mobility at multiple spatial granularities and multi-device mobility using passively sensed WiFi data. Second, we use the insights gained from characterization work to argue that human mobility is inherently hierarchical in nature and present a new approach for mobility modeling that captures mobility at various spatial granularities. Third, we present two mobility-aware applications - (i) WiFiTrace, a network-centric approach for contact tracing of infectious diseases using passive WiFi sensing and back tracing and (ii) iSchedule - a machine learning-driven technique to automatically learn custom occupancy-based Heating, Ventilation, and Air Conditioning (HVAC) schedules for buildings across a large campus. An application based on predicting future building occupancies by observing past mobility and occupancy of each building.
Advisor: Prashant Shenoy