University of Minnesota Driven to Discover
U of MNUniversity of Minnesota
Center for Transportation Studies

Investing for Robustness and Reliability in Transportation Networks

Presentation by Lei Zhang, Deptartment of Civil Engineering

November 23, 2004

While the discipline of civil engineering may be better known for researching bridge designs and pavement materials, researchers at the University of Minnesota are also pursuing work on "big picture" issues such as how large-scale road networks can be built and managed more effectively. Understanding the complex dynamics of transportation system development may hold the key to meeting tomorrow's transportation challenges.

On November 23, 2004, Lei Zhang presented his recent work with advisor David Levinson on developing their Simulator of Network Growth (SONG), a model that incorporates elements of travel demand and trip distribution as well as revenue, investment, and road use pricing.

Zhang and Levinson's work is based on the idea that transportation networks grow and change over time in response to many factors, including population and travel demand, financial policies, and land use. These networks may also exhibit different degrees of hierarchical organization, in which certain links are more important or more heavily used than others.

The researchers examined several different kinds of transportation networks, including roads, railways, and canals, noting the pattern of growth and decline in each case. Their SONG model aims to capture the relationship between road supply and travel demand as networks develop over time and through their spatial dimensions.

The current SONG model contains several components to simulate different factors affecting network growth, including:

  • Trip generation
  • Trip distribution
  • Traffic assignment
  • Revenue
  • Investment

Zhang and Levinson applied the SONG model to the Twin Cities highway network circa 1978 to see how well it would predict the characteristics of highway growth in the twenty years that followed. After comparing the model's predictions with the actual history of network growth, the researchers found that SONG had successfully predicted the construction of several important segments, including I-394, I-494, I-35W, Hwy. 10, and Hwy. 36.

However, the model forecasted greater expansion on roads that already had high capacities, and less expansion of arterial roadways than observed in reality. These differences led the researchers to suspect that the model may overestimate the costs of arterial roadway expansion, or underestimate costs of freeway expansion.

In another experiment, running the SONG with a hypothetical network in which all links are uniform showed how hierarchy could emerge naturally as a result of several characteristics, including natural barriers (such as rivers) or activity centers (like downtown areas).

Practical applications of Zhang and Levinson's work is in understanding how hierarchical structuring affects the reliability and security of transportation networks, and determining what kind of investment policies tend to create networks reliable and secure. Reliability, in this context, refers to a network's ability to function after a random link failure; security, on the other hand, refers to the ability to function after deliberate attack intended to damage the network.

As a first step toward understanding these issues, the researchers used a deliberately unrealistic network consisting of a simple grid of links and nodes. They then modeled the effects of two distinct investment policies—one focused on bottleneck removal, the second based on a form of benefit-cost analysis that weighed the cost of construction and maintenance against the benefits of reduced congestion.

They found that investment based on benefit-cost analysis tended to produce a less hierarchical network than that which developed under bottleneck removal policies. This "flatter" network was both more reliable and much more secure than the hierarchical type of network that was favored by bottleneck removal investment policies.

In the future, Zhang said, the researchers plan to continue to develop and refine the SONG system in several areas. Among these are the development of an agent-based demand model and combining SONG with a land use model such as UrbanSim. Future applications could include assessing alternative pricing structures, investment policies, and ownership structures for transportation networks, modeling substitutional effects between different modes such as automobile and light rail, modeling the addition of new links and nodes, and using SONG as a teaching tool in the classroom.