Home Skip to main content

Project Details

Title:Application of a Multi-Agent System with the Large-Scale Agent-Based Model for Freight Demand Modeling
Principal Investigators:EunSu Lee
University:North Dakota State University
Project #:MPC-458
RiP #:01528057
Keywords:agricultural products, demand, freight traffic, logistics, multi-agent systems, oil booms, state of the art, travel behavior


Statewide freight demand modeling is important in North Dakota to support agricultural logistics and energy development due to the recent oil boom and the long-term importance of the agricultural industry. We assume that a group of travelers, or agents, would provide a variety of driving patterns. Predicting travelers' behavior has been a cumbersome task in transportation planning because of the wide variation of behavior among travelers. With the advance of simulation and data mining, the agent-based model (ABM) has emerged as a solution. The agent-based modeling technique would provide a high level of detail for travel patterns in a region or state. The ABM includes three elements: agents, agent relationship, and agent's environment (Macal & North, 2011). The individual is known as an agent is an entity for decision-making. When an agent uses a vehicle, ABM is also called vehicle-based modeling. Agents interact with each other and in response to transportation infrastructures and policies. The agent-based freight demand modeling has been emerging as critical component in transportation planning to represent realistic travelling activities throughout the road networks and among facilities. ABM allows aggregations and disaggregation of agent characteristics, behaviors, and interactions under the freight demand context (Harper, et al., 2011). However, the agents should interact with environments for a long term such as in mid-term and long-range transportation planning. By simulating the agents, the freight movement in a large-scale network can be aggregated to provide critical information for statewide freight demand modeling without losing details. As a result, macro-level agent-based modeling benefits statewide freight demand modeling. In large-scale of road networks, the agents interact over space and time in response to information about transportation infrastructure and logistics facilities as well as policy. Thus, the multi-agent system is designed for the statewide macro level in response to the changes in these environments. Agent groups interact with other agent groups, and each agent within a group interacts with others within the group. Thus, the behavior (i.e. principle decision rules and response rules) of the agent group and each agent of a group can be simulated in transportation operations and planning. The multi-agent system includes decision-making rules such as destination choice, departure time, mode choice, and route choice, and sensitivity to travel impedance.

NDSU Dept 2880P.O. Box 6050Fargo, ND 58108-6050