Lead Presenter:  Yueshuai He | Assistant Professor | University of Louisville​​​​​​​

Presentation Description
Activity-based models (ABMs) are recognized the state-of-the-art approach for travel demand forecasting, offering explicit individual travel plans by integrating a series of interdependent travel behavioral choices. Compared to traditional trip-based models, ABMs provide enhanced capabilities for evaluating complex transportation policies and supporting strategic planning for agencies. However, their applications have been limited in the U.S., especially for small and medium-sized communities due to the substantial data requirements and computational complexity.

This study introduces ABM-Lite, a modeling framework designed to generate activity-based travel demand by leveraging existing origin-destination (OD) matrices from trip-based models in conjunction with open-source datasets. The proposed approach aims to lower the barriers to ABM adoption while maintaining sufficient analytical rigor. A case study is conducted using the Kentuckiana Regional Planning & Development Agency (KIPDA) regional travel demand model, which encompasses five counties across Kentucky (KY) and Indiana (IN).

The current travel demand model used by KIPDA, developed from a 2000 household travel survey, remains a trip-based system updated periodically to reflect demographic and socioeconomic changes. To generate the activity-based travel demand from the trip-based model, a synthetic population for the KIPDA area is first generated using census data from the American Community Survey, assigning key socioeconomic attributes to individuals. Long-term travel choices, such as vehicle ownership and driver licensing, are modeled using the data samples of KY and IN from the 2017 National Household Travel Survey (NHTS), providing foundational inputs for subsequent activity-based modeling. Activity patterns derived from the NHTS are used to construct individual activity chains, which are assigned to different demographic groups within the synthetic population. These activities are then allocated to traffic analysis zones (TAZs) based on spatial-temporal constraints and calibrated against the distributions from the existing OD matrix. The resulting disaggregated travel plans are simulated using the open-source Multi-Agent Transport Simulation (MATSim) platform, generating both aggregate traffic counts and individual travel trajectories for further analysis. The ABM-Lite framework is validated using standard performance metrics, including trip distance distributions, mode shares, and traffic counts on major corridors and screenlines.

By reducing data and computational demands, this framework offers a practical and scalable solution for transitioning from trip-based to activity-based modeling, addressing critical gaps in AMB applications for small and medium-sized communities. The application would enable these regions to adopt advanced modeling techniques without incurring prohibitive costs or resource requirements.

  Co-presenter(s): 

Eronmonsele Esekhaigbe
Kentuckiana Regional Planning and Development Agency (KIPDA)

View Presentation