Activity-Based Travel Demand Analysis using Michigan Connected Vehicles Test Bed Data

Lecture / Panel
For NYU Community


Professor Xuan (Sharon) Di Assistant Professor
Department of Civil Engineering & Engineering Mechanics
Columbia University


The activity-based approach is becoming a popular tool for travel demand modeling due to its power of accommodating individual travel data. The challenge of employing the activity-based approach, however, is to infer the complete sequence of activities and travel of the entire population from sample trajectories.

This talk aims to tackle such challenge by creating a methodological framework wherein similarity among household travel activity patterns is analyzed. To illustrate this methodology, 400 travelers’ GPS trajectories and survey collected from the world’s first connected vehicle testbed in Ann Arbor, Michigan, Safety Pilot project, is utilized. Each individual’s travel activity pattern is first extracted from the dataset. Then a similarity measure is defined to describe the extent to which two travelers’ activity patterns are alike. Based upon this measure, spectral clustering is conducted to divide travelers into multiple communities. A mathematical mapping is then established to explain similarity in travel activity pattern using travelers’ demographic information. This work will be generalized to reconstruct one’s activity sequence based on his/her demographic characteristics and infer one’s demographic group given his/ her demographic information.