Principal Investigator

  • Current

CMSA, working with UAHuntsville’s Office for Freight, Transportation, and Logistics, is developing an agent-based event-driven model of Alabama’s transportation system that can be used to analyze sources and effects of traffic congestion.

Alabama’s transportation infrastructure provides critical arteries for the movement of freight in the southeastern United States. Nearly 300,000,000 tons of freight are transported annually on Alabama’s highways, while about 20,000,000 tons of freight pass through the Port of Mobile—tens of billions of dollars worth of cargo rely heavily upon the quality of Alabama’s transportation infrastructure. The Center for Modeling, Simulation and Analysis (CMSA) and the Office for Freight, Logistics & Transportation (OFLT), at the University of Alabama in Huntsville (UAH) have developed an agent-based model to study the impacts of Alabama’s transportation infrastructure upon freight mobility in the state.

The primary goal of the model is to assess the degree to which traffic problems in the state impede the flow of freight. Problems typically arise during daily rush hour congestion in the heavily populated areas of Mobile and Birmingham, though specific traffic incidents can be added to the model as well.

To play out realistic traffic scenarios, the modeled vehicles must behave similarly to real vehicles on the roadways. Each modeled vehicle is an autonomous agent with full control over its actions. A modeled vehicle plans its route, observes its environment, and reacts. For instance, if a car ahead slows down, the vehicle may change lanes, or slow down, according to the conditions observed. Each agent vehicle has its own internal properties that allow the user to change the vehicle’s physical characteristics, such as braking distance, but also the psychological properties of the driver.

The state roadways are represented as a directed graph, in which each road intersection is a graph node, and each stretch of road between two intersections (called a link) is a graph edge. Each vehicle determines its own route from origin to destination upon instantiation into the system. The route is determined using the A* algorithm, which finds the shortest time between the vehicles origin and destination. Vehicles carry their own routes along with them so that they are free to re-route on-the-fly in the presence of slow traffic.

The simulation is event-based, which is to say that time advances when a vehicle changes its state (position and speed) or encounters the need to evaluate its situation. The events are placed into a chronologically ordered event queue. The most basic event type is a link change event, whose time is calculated each time the state is updated. This event causes the car to change links along its route at the appropriate time. If cars could “pass through” each other, they would simply conduct their necessary lane changes to navigate through their routes. Of course, a great deal of vehicle interaction events arise as traffic thickens, and these events begin to govern the effects of congestion.

The system adopts a multi-threaded Model-Viewer-Controller architecture. The model runs in a separate thread from the viewer and controller. The viewer features an interactive graphical user interface (GUI) which shows the movement of the vehicles (pictured). Display updates occur at fixed intervals of logical time.

As with any model, the input data are critically important to the validity of the model’s results. OFLT, under the direction of Gregory Harris, and currently under funding from the Federal Transit Administration, has made ground-breaking strides in understanding regional and local freight traffic impacts from the coarse state-by-state data provided by the US Department of Transportation. The group uses “freight analysis zones,” which are groupings of counties by several demographic factors, including population density and economic production by industry. This breakdown allows the group to estimate with high confidence the number of freight vehicles that enter the state’s highway system in each zone. Those numbers become the stimuli for the CMSA/OFLT model.

The primary outputs of the model are the delays associated with congestion. This is measured simply by evaluating the difference in trip time for a vehicle under current conditions to a trip under free flow speed conditions. This core metric leads to a host of other salient metrics. For example, the value of a truck on the road is estimated to be about 70 dollar per hour. Our model is currently finding total statewide delays of order thousands of hours, which translates into hundreds of thousands of dollars each day that are burned up purely by freight congestion on the roadways.

Currently, work is underway to expand the road network to provide a more accurate representation of local congestion. The group will then focus on validation against actual traffic counts. The validated model will be used to assess economic impacts of planned and candidate roadway improvements, major road developments, and new industrial installations. The model can also serve as a tool for planning evacuations and other emergency management activities.

UAH’s CMSA and CMER/OFLT are rapidly developing a statewide traffic modeling capability ready to guide state planners in keeping Alabama at the forefront of economic growth.