Development of an Agent-Based Traffic Simulation Framework Using a Generalized Cell Transmission Model for Regional-Scale Dynamic Assignment
This study presents iTLE-TSIM, an agent-based traffic simulation framework that integrates activity schedules with a generalized Cell Transmission Model to achieve finer spatial resolution and dynamic congestion representation for regional-scale applications while maintaining computational efficiency.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Traffic planners have long faced a difficult choice between seeing the forest and seeing the trees. To understand how a city moves, they traditionally relied on models that treated traffic as a smooth, continuous flow, much like water moving through a pipe. These models are efficient for large regions but blur the details of individual drivers, their specific daily routines, and how a single jam might ripple backward through a neighborhood. On the other end of the spectrum are microscopic simulations that track every car and driver as a distinct person with a unique schedule. While these offer incredible detail, they demand so much computing power that they become impractical for simulating an entire region with millions of people. The challenge has been to find a middle ground that keeps the human element of travel without getting bogged down by the sheer weight of the data.
Researchers at Dalhousie University have developed a new tool called iTLE-TSIM to bridge this gap. It is a traffic simulator designed to work directly with activity-based models, which track people's daily lives rather than just their trips from point A to point B. Instead of forcing complex human schedules into simplified summaries, this new system lets individual agents—representing real people with real jobs, errands, and home lives—enter the road network exactly when they are ready to leave. The system then moves these agents through the city using a method that breaks every road into small, manageable segments. This allows the simulation to show exactly where a line of cars forms, how long it takes to clear, and how a blockage at one intersection can cause a backup that stretches back for miles, all while running fast enough to handle a whole region.
The core of this new framework is a way of thinking about roads that is different from the standard approach used in most commercial software. Traditional models often treat a road link as a single unit, assuming that traffic enters at one end and exits at the other after a set time. This can hide the reality of how congestion actually builds up inside a road. The researchers instead divided every road into a chain of small cells. Imagine a road as a series of parking spots, where each spot can hold only a certain number of vehicles. As cars move forward, they must wait for the spot ahead of them to open up. If the spot is full, the car stays put, creating a queue that grows backward, one cell at a time. This simple rule allows the computer to naturally reproduce the physics of a traffic jam, including the way a backup can spill over from one road onto another, without needing to calculate complex wave equations.
To make this work on a regional scale involving hundreds of thousands of people, the team built the simulator to handle the unique behavior of individual drivers. In the real world, people do not just drive from home to work; they drive to the grocery store, then to a child's soccer practice, and finally home, adjusting their plans if they get stuck in traffic. The new system captures these full daily chains. When a driver is released into the simulation, they choose a route based on current conditions. If they encounter a sudden delay, the system allows them to change their path mid-journey, just as a real driver might turn onto a side street to avoid a backup. The model also includes safeguards to prevent the computer from getting stuck in a loop where cars block each other in a circle, a problem that can happen in digital simulations but rarely occurs in the real world because human drivers eventually find a way out.
The researchers tested this framework using the city of Halifax, Nova Scotia, as a prototype. They fed the system data representing 457,000 individual agents and their 1.6 million daily activities, including car trips, delivery trucks, and scheduled buses. The simulation ran for a full day, tracking every vehicle's movement second by second. The results showed that the system could successfully reproduce the rise and fall of traffic throughout the day. It captured the quiet early morning hours, the gradual build-up of congestion as the morning rush began, and the dense, gridlocked conditions that formed around the city center by 8:00 a.m. The visual output showed traffic spreading along major corridors and bridges, with queues forming and dissipating in a way that matched real-world observations.
When the team compared the results of this new simulator against both actual traffic counts collected in Halifax and the outputs of the existing commercial software used by the city, the agreement was strong. The new model matched the spatial patterns of traffic seen in the older system, correctly identifying which roads were busiest and where congestion clustered. In fact, the new approach offered a clearer picture of local streets. Because the older models often assigned traffic to roads based on broad geographic zones, they sometimes missed the nuances of neighborhood traffic. The new system, by placing each driver at their exact starting coordinate, showed a more realistic distribution of vehicles on local roads. The statistical comparison showed that the new model's predictions aligned closely with real-world data, particularly during the morning and evening rush hours.
This work represents a significant step forward in how we understand and plan for urban mobility. By combining the efficiency of large-scale modeling with the realism of individual behavior, the researchers have created a tool that can help city planners test policies and infrastructure changes with greater precision. The system can show not just how many cars are on the road, but how those cars interact, how delays propagate through a network, and how individual choices ripple through the entire city. While the current version focuses on passenger cars and delivery trucks, the framework is designed to eventually include public transit passengers and more complex intersection rules. For now, it stands as a practical, computationally efficient way to simulate the complex, human-driven reality of regional traffic, offering a clearer view of the city in motion.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.