Skip to contentUnited States Department of Transportation - Federal Highway Administration FHWA Home
Research Home
Public Roads
Featuring developments in Federal highway policies, programs, and research and technology.
This magazine is an archived publication and may contain dated technical, contact, and link information.
Federal Highway Administration > Publications > Public Roads > Vol. 70 · No. 4 > A Model Endeavor

Jan/Feb 2007
Vol. 70 · No. 4

Publication Number: FHWA-HRT-2007-002

A Model Endeavor

by Vassili Alexiadis, James Colyar, and John Halkias

A public-private partnership is working to improve traffic microsimulation technology.

This photo shows one of seven digital cameras that were mounted atop a building overlooking I-80 near Emeryville, CA. The cameras recorded vehicle movements, which then were used to create a vehicle trajectory dataset, which in turn supported development of a behavior algorithm for driver behavior.
(Above) This photo shows one of seven digital cameras that were mounted atop a building overlooking I-80 near Emeryville, CA. The cameras recorded vehicle movements, which then were used to create a vehicle trajectory dataset, which in turn supported development of a behavior algorithm for driver behavior.

Traffic analysts today are faced with evaluating diverse and complex solutions to address congestion in transportation systems. Instead of "simply" deciding how many lanes to design for a new freeway or how long the turn bays should be at a traffic signal, practitioners now are analyzing advanced traffic signal and ramp metering systems, for example, and complex weaving and geometric configurations, intelligent transportation system strategies, multimodal corridor management plans, and congestion pricing strategies.

Traffic microsimulation analysis tools can help evaluate these complex solutions by modeling real-world transportation networks on a systemwide scale that is difficult with more traditional methods. Dramatic improvements in computer processing speeds and capabilities in the past decade have enabled traffic microsimulation software to model increasingly complex and larger scale transportation systems. As a result, microsimulation is quickly becoming popular among traffic analysts and is playing an important role in transportation investment decisions.

The Role of FHWA

The Federal Highway Administration (FHWA) is a leader in developing traffic microsimulation models, dating back to the development of NETwork SIMulation (NETSIM) in the 1970s, FREeway SIMulation (FRESIM) in the 1980s, and the merging of NETSIM and FRESIM into a single CORridor SIMulation (CORSIM) model, all of which was integrated into the Traffic Software Integrated System (TSIS) package in the 1990s. In the early 1990s, TSIS/CORSIM was the only viable traffic microsimulation model available to practitioners. By the late 1990s, however, a number of commercial vendors began offering their own versions of traffic microsimulation packages to meet the growing demand. Today, the popularity of microsimulation packages continues to increase, and there is now a viable market for commercial traffic simulation vendors.

In the early 2000s, FHWA reevaluated its future role in the traffic simulation market. A survey of traffic practitioners and existing traffic simulation packages revealed that while most of the software packages, although robust and providing a range of analysis options, still have some intrinsic limitations that can affect the performance and accuracy of the model results. These limitations in the functionality of current microsimulation systems have generated questions in the transportation community. For example, simulation users view many microsimulation software packages as "black boxes" in that users are not sure how model outputs are calculated and, as a result, are not confident in the accuracy and validity of the model results.

As a result of the market assessment, FHWA decided to take a different role in the traffic simulation market. Rather than compete with the commercial simulation vendors by continuing to develop TSIS/CORSIM, FHWA would act in a "market facilitator" role by focusing public resources on fostering an environment of public-private coordination through research products that will benefit the entire traffic simulation community: practitioners, vendors, and researchers.

Defining "Algorithm"

Under the NGSIM program, FHWA is sponsoring fundamental traffic simulation research through development of core driver behavior algorithms to help improve the capacity and accuracy of commercial microsimulation software. A core driver behavior algorithm is a mathematical model that attempts to replicate the decisions that drivers continuously make in response to other vehicles. NGSIM focuses on the immediate decisions about when to change lanes, how closely to follow the car in front, and what size of gap to accept when changing lanes or entering an intersection.

Enter the NGSIM Program

With the goal of improving the quality and use of traffic microsimulation tools to facilitate transportation decisionmaking, FHWA's Traffic Analysis Tools Program began the Next Generation Simulation (NGSIM) program in 2002. NGSIM is a unique public-private partnership between FHWA, transportation consulting companies, university researchers, and foreign and domestic commercial microsimulation software developers.

The objective of the program is to develop a core of driver behavior algorithms that represent the fundamental logic in traffic microsimulation models, with supporting documentation and validation datasets. NGSIM products will be well documented, openly distributed, and free to the transportation community through the NGSIM Web site (www.ngsim.fhwa.dot.gov).

"The NGSIM program represents a model public-private partnership that has yielded demonstrable benefits for both sectors," says Nagui Rouphail, chairman of the NGSIM stakeholder traffic modelers group and director of the Institute for Transportation Research and Education at North Carolina State University. He adds, "Here the [U.S.] Government acts as the catalyst for developing sound science and the data to back it up, while the private sector commits to participate in the development process as well as incorporating the research findings into its commercial software. This process ensures wider dissemination of the research results and even wider acceptance of the underlying science."

The NGSIM team is composed of traffic simulation and modeling experts managed by a private company. The team is supported by senior advisers from respected transportation institutions across the Nation and includes three stakeholder groups: a traffic modelers group that represents researchers and others who develop driver behavior models, a software developers group of private vendors responsible for developing and maintaining commercial traffic simulation software, and a model users group that represents the practitioners who use traffic simulation models for decisionmaking.

Says Martin Fellendorf, an NGSIM stakeholder in the developers group and also the director of the Institute of Highway Engineering and Transport Planning at the Graz University of Technology in Austria: "So far the NGSIM project has made good progress in bringing together various stakeholders, such as domestic research groups, core users, and foreign software developers. NGSIM is a neutral platform in which competitors can exchange ideas and upcoming developments to the benefit of the users."

Currently, the NGSIM stakeholder team is developing, estimating, coding, and testing the core simulation algorithms. They are using real-world datasets, with their corresponding data descriptions, to estimate and validate the algorithms. The traffic simulation community also is using these datasets to help verify, validate, and calibrate newly developed and existing behavioral models. "I expect these datasets to be a benchmark in comparing simulated data with real-world data," says Fellendorf.

The ultimate goal of the NGSIM program is to improve the quality, trustworthiness, and use of traffic microsimulation models, which will lead to improved decisions by public agencies on transportation investments.

This computer rendering from a microsimulation modeling program depicts a transportation corridor with two-way traffic, including cars, a bus, a truck, and an exit ramp; a commuter train running parallel to the highway; and an overpass.
This computer rendering from a microsimulation modeling program depicts a transportation corridor with two-way traffic, including cars, a bus, a truck, and an exit ramp; a commuter train running parallel to the highway; and an overpass.

Getting Started

At the outset, the NGSIM team, together with a variety of experts and stakeholders, conducted a market assessment of traffic microsimulation. The team looked at existing microsimulation models, identified the limitations of the models, and prioritized the NGSIM research objective as: improving the core of driver behavior algorithms in microsimulation software.

To accomplish this research objective, the NGSIM team formulated plans for collecting data, developing algorithms, and validating the algorithms. These plans ensured that the research would be conducted through a consistent, rigorous process. The team also developed the project infrastructure for the free and open sharing of data by developing data formats and a Web site for online dissemination of NGSIM products.

The team proceeded to collect and process detailed data on vehicle trajectories on arterials and freeways. The data then were used to develop driver behavior algorithms and then to validate those algorithms with various commercial microsimulation models.

NGSIM Program Structure
NGSIM Program Structure

First Algorithm

Looking at the entire development cycle of an NGSIM algorithm provides an informative cross section of the program. One of the highest ranked algorithm needs identified was in freeway lane selection (FLS).

Existing simulation software models are based on the assumption that drivers evaluate the current and adjacent lanes and choose to change (or not to change) direction based on the relative conditions (for example, average speeds and traffic densities) of these lanes only. But the choice of target lane may require a sequence of lane changes from the current lane. Because existing models can explain only one lane change at a time, they are not equipped to deal with such relatively complex scenarios.

A good example is roads with high occupancy vehicle (HOV) lanes, which may be significantly more attractive than other lanes to a subset of eligible drivers. Drivers may perform several lane changes to reach the HOV lane, but again, current models would not account for the HOV lane if it was not adjacent to the current lane.

Once the NGSIM team identified the FLS algorithm as a priority, the team began collecting data, drawing on a section of I-80 near Emeryville, CA. Estimating and validating behavior algorithms requires detailed information on the movements of individual vehicles, as well as other vehicles in the vicinity. Vehicle trajectory data, which include observations of the positions of all vehicles in a section of road, are useful in this context. Almost all vehicle positions are tracked at 0.1-second intervals, resulting in detailed knowledge of lane positions and relationships to other vehicles. The team also collected loop detector data, aerial images, computer-aided design diagrams, and other supporting data. As a result, this and other NGSIM vehicle trajectory datasets are the most detailed and accurate field data collected for traffic microsimulation research and development to date.

One of the university members of the NGSIM team led development of the FLS algorithm. Overall, the process consisted of estimating the algorithm from the I-80 vehicle trajectory dataset, and then validating the algorithm using aggregate data, such as that available from vehicle loop detectors, from the I-80 dataset.

The FLS algorithm consists of a generalized lane-changing model that for the first time explicitly incorporates choice of the ultimate target lane. The lane-changing process consists of two steps: choice of target lane and deciding the minimum acceptable gap in traffic for making the switch. The target lane is the lane the driver perceives as most desirable, considering a wide range of factors and goals.

NGSIM Algorithm Development Process
NGSIM Algorithm Development Process

The algorithm went through an extensive stakeholder review process from its conceptualization to its development and validation. It then underwent rigorous validation in three widely used commercial microsimulation software systems—PTV's VISSIM, TSS-Transport Simulation System's AIMSUN NG, and Quadstone Paramics, developers of which all served on the NGSIM developers' stakeholder group. After that, the team measured the algorithm against real-world aggregate datasets. The developers were involved in the entire process of creating the algorithm, and their willingness to incorporate and validate the algorithm in their own software is evidence of the need for the algorithm and of the success of the partnership developed through the NGSIM program.

Conclusions

"This particular algorithm proved to be superior with respect to clarity, scientific purity, and comprehensibility," says Fellendorf. "It is much better documented than any of the algorithms embedded so far in any of the commercial simulation packages."

One of the commercial participants in the program, the VISSIM developer, concluded that the NGSIM FLS algorithm is a valuable addition to his company's software and to the state of the art in microscopic traffic flow simulation generally. This conclusion was based on the fact that modeling specialized lanes, especially HOV lanes, is easier and more straightforward in the FLS algorithm compared with the current VISSIM model. Also, the concept of the FLS algorithm is extendible, that is, driving situations not covered in the current version can be added easily into the algorithm's framework. The developer plans to incorporate the NGSIM FLS algorithm in the next full version of VISSIM.

The AIMSUN NG developer determined that the FLS algorithm provides slightly better or comparable results to the existing AIMSUN NG model. He concluded that the main difference lies in the systematic use of the model parameters in the FLS algorithm as opposed to the trial-and-error process in AIMSUN NG. His firm plans to incorporate all the NGSIM algorithms into its commercial software as soon as they become available. The Paramics developer concluded that the FLS algorithm works in a wide range of scenarios and plans on incorporating the FLS algorithm, and additional NGSIM algorithms, into Paramics once the algorithms become available.

Additional NGSIM Algorithms

The NGSIM team currently is developing three additional driver behavior algorithms, following the same process it did for the FLS work. The new algorithms include a cooperative/forced freeway merging algorithm that explicitly considers cooperation and competition as drivers maneuver through congested freeway merging and weaving areas, an arterial lane selection algorithm that considers both preemptive lane-positioning behaviors and more aggressive overtaking behaviors on congested arterial corridors, and an oversaturated freeway flow algorithm that focuses on car-following and lane-changing behaviors during congested, stop-and-go conditions.

To support development of these algorithms, the NGSIM team has collected two additional freeway datasets (another on I-80 in Emeryville and one on U.S. 101 in Los Angeles, CA) and an arterial dataset (on Lankershim Boulevard in Los Angeles), and is planning to collect a second arterial dataset in Atlanta, GA.

This flowchart illustrates the structure of the freeway lane selection algorithm. First the driver chooses a target lane from any of the five on this sample freeway. Then the driver accepts or rejects the gap in adjacent lanes in order to maneuver to the target lane. The final choice is whether to change lanes.
This flowchart illustrates the structure of the freeway lane selection algorithm. First the driver chooses a target lane from any of the five on this sample freeway. Then the driver accepts or rejects the gap in adjacent lanes in order to maneuver to the target lane. The final choice is whether to change lanes.

Enabling Better Decisions

"NGSIM by itself will not bring viable products, but if its core algorithms prove to be superior—as the freeway lane selection [FLS] algorithm already has—then private industry will endorse them and embed them within their own software environments, which will be continuously maintained as long as a solid user base asks for them," Fellendorf predicts.

As a result of incorporating the FLS and other NGSIM algorithms into the widely used commercial microsimulation models, transportation practitioners will be able to use simulation software systems more confidently, knowing the core algorithms were rigorously developed by experts and peers using high-quality, real-world datasets. Improving the core algorithms of these models ultimately will lead to more reliable and valid transportation decisions, crucial in today's age of shrinking transportation budgets and need for efficient transportation investments.

Says Rouphail: "The FHWA NGSIM program has, for the first time in a long time, funded a national effort aimed at the collection and synthesis of several high-quality empirical datasets on driver behavior on freeways and surface streets that will undoubtedly advance the state of the art in microscopic traffic modeling and simulation for decades to come."


Vassili Alexiadis is a vice president of Cambridge Systematics, Inc. and principal investigator for NGSIM. He holds a doctorate in transportation from the University of Massachusetts Amherst, a master's degree in transportation engineering from the State University of New York, and a bachelor's degree in civil engineering from Greece's Aristotle University of Thessaloniki.

James Colyar is a highway research engineer in the FHWA Office of Operations Research and Development. He holds a bachelor's degree in civil engineering from the University of Idaho, a master's degree in civil engineering from North Carolina State University, and a master's degree in transportation policy from George Mason University.

John Halkias is the systems management team leader with FHWA's Office of Transportation Management in the Office of Operations. He holds a bachelor's degree in civil engineering from the University of Connecticut and master's and doctoral degrees from West Virginia University.

For more information on the NGSIM program, visit the NGSIM Web site at http://ngsim.fhwa.dot.gov or contact Vassili Alexiadis at 510-873-8700 or valexiadis@camsys.com, James Colyar at 202-493-3282 or James.Colyar@fhwa.dot.gov, or John Halkias at 202-366-2183 or John.Halkias@fhwa.dot.gov.

ResearchFHWA
FHWA
United States Department of Transportation - Federal Highway Administration