Agent-Based Model Architecture for Mesoscopic Traffic Simulations

Agent-based modeling (ABM) techniques have become another viable option for solving highway transportation problems. Research advancements in recent years have guided the possible integration between the ABM techniques and traffic simulations in addressing various transportation demand problems especially related to predicting road user behaviors. However, previous ABM simulation frameworks were implemented with a very high amount of modeling details (abstraction) to simulate microscopic vehicle movements and study real-time road user behaviors and responses. The application of the agent-based integrated traffic simulations at this level of abstraction may not be efficient for producing a more aggregate result required in the analysis of long-term transportation projects. In addition, very few ABM simulation frameworks have taken into account complex interactions between the supplied levels of service from a highway network and the aggregate demand created by road user choices. This research, therefore, proposes a model architecture and a theoretical framework for an agent-based traffic simulation at a mesoscopic level. In the framework, traffic simulation is created from an agent-based system where road user agents are interacting with one another and the highway network through network levels of service. The framework's model architecture is defined by two components including a highway network and user agents. The definition of this modeling architecture is targeted at utilizing the strength of the ABM techniques to address long-term transportation problems by providing an appropriate aggregate level of traffic simulations, while maintaining the ability to represent road user behaviors.