Mobile Sensing for Data-Driven Mobility Modeling

Abstract

The use of mobile sensed location data for realistic human track generation is privacy sensitive. People are unlikely to share their private mobile phone data if their tracks were to be simulated. However, the ability to realistically generate human mobility in computer simulations is critical for advances in many domains, including urban planning, emergency handling, and epidemiology studies. In this paper, we present a data-driven mobility model to generate human spatial and temporal movement patterns on a real map applied to an agent based setting. We address the privacy aspect by considering collective participant transitions between semantic locations, defined in a privacy preserving way. Our modeling approach considers three cases which decreasingly use real data to assess the value in generating realistic mobility, considering data of 89 participants over 6079 days. First, we consider a dynamic case which uses data on a half-hourly basis. Second, we consider a data-driven case without time of day dynamics. Finally, we consider a homogeneous case where the transitions between locations are uniform, random, and not data-driven. Overall, we find the dynamic data-driven case best generates the semantic transitions of previously unseen participant data.