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

Programs & Labs

ITS Sensor Header

Fall 2000

Researchers team up to develop new techniques

Photo of Eil Kwon

Dr. Eil Kwon

Photo of Nikolaos Papanikolopoulos

Associate Professor Nikolaos Papanikolopoulos

Freeway entrance and exit ramps are often the chief contributors to difficult-to-solve traffic bottlenecks. Even the most basic entrance/exit 'weaving sections' involve complex vehicle interactions that are difficult for engineers to track as vehicles change lanes, cross paths with other vehicles, and merge with through traffic to enter or exit an expressway. Without the ability to track this movement, engineers cannot collect or accurately compute traffic data—such as driver behavior and real-time capacity estimates—and consequently, cannot accurately determine how to improve traffic flow in these areas.

Historically, little research has been done to address those particular needs. However, a recent project at the ITS Institute has made great strides in learning more about weaving areas. In a project funded by the Minnesota Department of Transportation (Mn/DOT), Dr. Eil Kwon, advanced traffic systems program director of the ITS Institute at the Center for Transportation Studies and adjunct professor in the Department of Civil Engineering, has developed an improved procedure for estimating time-variant capacity in weaving sections.

Kwon's research required the ability to accurately track vehicles and measure speeds and locations of both merging and diverging vehicles as they change lanes in a weaving section. At the time the research began, existing visual tracking systems were not able to provide the necessary data, so Kwon contacted Associate Professor Nikolaos Papanikolopoulos of the Department of Computer Science and Engineering. With separate funding from the ITS Institute, Papanikolopoulos worked with Kwon to specify system requirements and develop a portable, PC-based video-detection system to meet Kwon's needs.

Papanikolopoulos' system uses a video camera to capture images of vehicles moving through the selected weaving sections. Once the video data are collected, the software uses recursive filtering to remove invalid measurements, and advanced segmentation techniques to extract the vehicle images and remove unnecessary background objects, enabling accurate computation of velocity and trajectory for as long as the vehicle is present in the image. The system has a user-friendly interface that allows the videotaped images, along with a few geometric parameters such as lane width, to be fed into a PC. The system will then output the weaving data in the form of a spreadsheet. To check the accuracy of the computation, engineers manually counted and tracked vehicles as they changed lanes within a specified frame, and then compared that information to that of the computational algorithm.

Once the software was complete, Kwon began his study by selecting six weaving sites throughout Minneapolis-St. Paul, Minn. He then developed a database using data collected from loop detectors at those sites. To collect video data from these areas, Kwon worked with Mn/DOT engineers who assembled a mobile video recording system consisting of a special trailer with a 44-foot mast on which the video camera was mounted. From this elevated vantage point, the video camera was able to capture clear images of vehicles moving through weaving sections.

The initial results of Kwon's research are promising. Kwon documented differences in driver lane-changing behavior for congested and free-flow conditions–behavior that suggests that the length of the merge lane is not a primary reason for limited capacity in weaving areas. Rather, Kwon was able to quantify theories that the speed of merging flow is very close to that of exiting flow, and that the downstream conditions, such as the geometry of the exit ramp, directly affect the speed of the merging vehicles.

Based on the results from weaving behavior analysis, Kwon developed an adaptive procedure that can estimate the maximum possible weaving volume of a simple ramp-weave section through time with reasonable accuracy. This type of accurate information will allow traffic managers to better control traffic flow in weaving areas, and in turn, improve congestion problems throughout the freeway network.

Kwon and Papanikolopoulos are now moving toward the next phase of the weaving area capacity study scheduled to begin in November 2000.

"In phase two, we¡ll study weaving areas that have multiple lane-changing patterns," explains Kwon. "The video detection tool developed by Nikolaos' group will be used extensively to collect vehicle trajectory data in those areas, and we'll use the estimating model developed in phase one to analyze these more complex types of weaving areas in the Twin Cities area."

Besides the benefit it provides Kwon in his work, Papanikolopoulos' vehicle-tracking system can also be used to monitor vehicle and pedestrian interactions at complicated urban intersections and possibly to predict situations that may evolve into accidents. By working together, Kwon and Papanikolopoulos have developed engineering tools that can improve the design of weaving sections and may potentaially reduce accidents and improve traffic flow. Since traffic delays and accidents cost the economy real dollars and cents, their work has far-reaching financial implications.