Exploitation of Aggregate Mobility Sensing Data for the Synthesis of Disaggregate Multimodal Tours in Megacities

Abstract

The unprecedented volume of urban sensing data has allowed the tracking of individuals at remarkably high resolution. As an example, Telecommunication Service Providers (TSPs) cannot provide their service unless they continuously collect information regarding the location of their customers. In conjunction with appropriate post-processing methodologies, these traces can be augmented with additional dimensions such as the activity of the user or the transport mode used for the completion of journeys. However, justified privacy concerns have led to the enforcement of legal regulations aiming to hinder, if not entirely forbid, the use of such private information even for purely scientific purposes. One of the most widely applied methods for the communication of mobility information without raising anonymity concerns is the aggregation of trips in origin–destination (OD) matrices. Previous work has showcased the possibility to exploit multi-period and purpose-segmented ODs for the synthesis of realistic disaggregate tours. The current study extends this framework by incorporating the multimodality dimension into the framework. In particular, the study evaluates the potential of synthesizing multimodal, diurnal tours for the case where the available ODs are also segmented by the transport mode. In addition, the study proves the scalability of the method by evaluating its performance on a set of time period-, trip purpose-, and transport mode-segmented, large-scale ODs describing the mobility patterns for millions of citizens of the megacity of Tokyo, Japan. The resulting modeled tours utilized over 96% of the inputted trips and recreated the observed mobility traces with an accuracy exceeding 80%. The high accuracy of the framework establishes the potential to utilize privacy-safe, aggregate urban mobility data for the synthesis of highly informative and contextual disaggregate mobility information. Implications are significant since the creation of such granular mobility information from widely available data sources like aggregate ODs can prove particularly useful for deep explanatory analysis or for advanced transport modeling purposes (e.g., agent-based, microsimulation modeling).