Publications:

• Newsletters and Journals
• Annual Reports
• Research Reports
• Proceedings & Presentations
• Technical Assistance Resources
• Videos
• Subscribe
• Order

Current Issue | Previous Issues | Subscribe May 2006 - Vol. 4 No. 5

Policy & Planning

Modeling the future of Minnesota's transportation network

Today’s state and county decisions shape what Minnesota’s transportation network may look like many years from now, according to David Levinson, associate professor of civil engineering at the University of Minnesota. At a recent research seminar, Levinson invited the audience to consider whether this future calls for changes in today’s policies.

Building on his 2003 research project, If They Come, Will You Build It?, Levinson set out to predict the growth of transportation networks, specifically what the network of 2020 or 2035 will look like. That project received funding from the Minnesota Department of Transportation as part of the agency’s efforts to understand the evolution of the state’s transportation system. In his seminar, Levinson described how he developed a growth model by categorizing the decision-making processes of Mn/DOT and counties as formal (structured, with point ranking and committees) or informal (priority-based). People not directly involved in the process don’t understand how decisions are made to distribute funds fiercely sought by different stakeholders, Levinson said. Projects that are seen as most capable of improving safety or relieving congestion are most likely to get a green light, he explained.

Levinson’s model predicts that the rate of lane-miles added by Mn/DOT each year from 2005 to 2010 will decrease, and after 2010, expansion will double. Rates will vary among the counties, with those in the metro showing a constant rate of growth.

Though the model does not yet consider factors like wetland preservation, right-of-way acquisition, and legacy links (projects planned decades ago and given priority), Levinson said by comparing data from different organizations, it is possible to get a general idea of what the future holds: increases in trips, households, and construction budgets. The end goal, slated for completion in 2006, is to model future growth based on differing decision-making policies.

Levinson’s previous report, If They Come, Will You Build It? can be downloaded from the Mn/DOT Web site and from the CTS Web site.

Intelligent Transportation Systems

See that car? Vision-based vehicle tracking continues to evolve

To a person standing beside a busy intersection, the movements of cars, trucks, and buses are easy to follow. Traffic researchers have long relied on the work of human observers to gather accurate, detailed data on traffic patterns at intersections, freeway ramps, and other areas of complex traffic flow. But ongoing research by computer scientists at the University of Minnesota is bringing fully automated, vision-based tracking closer to deployment.

Computer science and engineering professor Nikos Papanikolopoulos is leading the team of researchers and students in the Artificial Intelligence, Robotics, and Vision Laboratory (AIRVL), which he directs. For the most recent research report on the work, Papanikolopoulos is joined by co-authors Harini Veeraraghavan, Stefan Atev, Osama Masoud, and Grant Miller.

The new report presents a set of algorithms related to vision-based vehicle tracking at traffic intersections, building on past work carried out at the AIRVL. These algorithms address several of the principal challenges to accurate collection of visual data in real-world environments, including spatial resolution, variable light conditions, and camera placement.

Because the visual processing capabilities of computers are extremely crude compared to the human visual system, vision-based tracking depends on the ability to process “cues” about the scene—such as the distribution of colors, or variations that appear on top of a stable background. Although it is possible to track targets using only one kind of visual cue, the ability to integrate multiple cues can yield greater accuracy. The AIRVL researchers have taken this approach, integrating color distribution with the motion of pixel “blobs” against the scene background. Raw tracking results are fed to a switching Kalman filter for integration.

Experimental evaluation of the system was carried out in the field using different types of intersections, including a “T,” a one-way street crossing a two-way street, and the intersection of two two-way streets. It proved capable of tracking a variety of movements by multiple vehicles using video from a pole-mounted camera near the intersections.

Development of a Tracking-Based Monitoring and Data Collection System (Mn/DOT 2005-40) is available on the Mn/DOT Web site.

Remote sensing takes wing

From a few hundred feet above the ground, the video cameras have a superb view of the network of roads below. Assistant Professor Demoz Gebre-Egziabher of the Department of Aerospace Engineering and Mechanics peers at the screen of a laptop computer displaying real-time video beamed down to the ground station via a wireless video link. When he wants to view another area, instead of radioing instructions to a pilot, Gebre-Egziabher turns to a graduate student beside him who adjusts the flight controls. The video perspective shifts as the unmanned craft banks to follow the curve of a suburban highway.

The birds-eye view afforded by an aerial platform offers many advantages for gathering data on traffic movements and general surveillance of transportation infrastructure, including a wide field of view and the capability of moving rapidly between different monitoring sites or following a single vehicle as it traverses a network. But the expense and risk associated with keeping a piloted plane in flight for hours at a time—not to mention the highly trained personnel and maintenance required—limit the effectiveness of conventional aircraft as monitoring platforms.

One solution to these limitations is to find an alternative to the need for a pilot—the most sensitive and expensive component in the system. Gebre-Egziabher's research group is currently developing unmanned aerial vehicles (UAVs) as monitoring platforms specifically targeted at surveillance and inspection of the surface transportation system. Through advances in navigation and guidance systems, sensors, and flight operations techniques, UAVs may play an important role in future transportation data collection and security.

The potential advantages of UAVs as sensing platforms have fuelled the development of myriad designs, ranging from feather-light experimental “micro-UAVs” weighing only a few hundred grams to the imposing jet-propelled Global Hawk, operated by the U.S. Air Force, which has a wingspan of 116 feet and a takeoff weight of more than 25,000 pounds. More modest UAVs, similar to or based on hobbyist radio-controlled aircraft, have been used as airborne camera platforms, notably to assess coastal erosion and damage to structures in the aftermath of Hurricane Katrina in 2005.

While a remotely operated vehicle has advantages over a piloted aircraft, a partially or fully autonomous craft offers even more intriguing possibilities. Designing a degree of autonomy into the UAV’s guidance system would allow the vehicle to move from location to location, or around a patrol route, without requiring an operator on the ground to guide it through each turn.

Safe operation is among the most difficult issues facing UAV designers. Being pilotless does not exempt UAVs from the Federal Aviation Administration’s strict regulations governing all aircraft operations, including the requirement that aircraft operating in controlled airspace possess the capability to autonomously “sense and avoid” other aircraft. Currently, autonomous sensing and avoidance is beyond the capabilities of even the most advanced military UAVs. Without this ability, UAVs are restricted to operating at low altitudes, away from areas where they are likely to come near other aircraft, and within view of their operators on the ground.

Gebre-Egziabher describes the development of autonomous sense-and-avoid systems as the greatest challenge facing UAV researchers today. Such systems will make it possible for autonomous and semi-autonomous UAV sensor platforms to operate in a wide range of areas. Because highly accurate and dependable navigation is a necessity for future sense-and-avoid capability, much of the Minnesota researchers’ work is directed at developing operational techniques and navigation systems that provide verifiable levels of navigation safety consistent with FAA requirements.

Although Global Positioning Systems (GPS) are capable of high accuracy, the integrity of navigation devices based solely on GPS is subject to errors in the satellite-based system. Due to these inherent inaccuracies, a systen based purely on GPS cannot guarantee positional integrity at the small scales required for safe navigation.

Because UAVs will need other sensors to augment GPS for navigation, guidance, and control in order to operate in urban areas, the researchers are also facing the technical challenges of multi-sensor integration. The design of algorithms that combine data from different types of sensors is a challenge in itself. The goal is to take advantage of the strengths of each type of sensor while minimizing its weaknesses. As different sensor types produce data at different rates and with different levels of accuracy, successful “sensor fusion” depends on factoring out a large number of possible sources of error.

The Minnesota researchers’ work has important implications for all types of UAV applications. Unmanned vehicles that can operate autonomously or semi-autonomously could one day replace piloted aircraft in a variety of missions, reserving pilots for those tasks where local, human decision making is critical for success.

Transportation Infrastructure

Adhesive techniques offer promise for cracked metal bridge elements

Fatigue cracking is one of the most common—yet significant—forms of damage to metal bridges. The techniques used to control cracks depend on what part of the structure is affected and the type of stress that causes the cracking, among other factors. Yuying Hu, Carol Shield, and the late Robert Dexter of the University of Minnesota’s civil engineering department have recently studied a relatively simple retrofitting technique with the potential to extend the service life of bridges suffering from out-of-plane fatigue cracking.

The technique involves using an adhesive compound to attach an angled steel reinforcement. The application process does not interrupt the flow of traffic and the cost is low in comparison to other retrofitting techniques. These benefits make reinforcement using adhesives an attractive proposition for maintenance personnel.

The researchers conducted both a field test and a large-scale laboratory test of the adhesive-based technique, exploring many variables. Their report highlights the effects of exposure to different temperatures (both during application and during service), as well as typical exposure to outdoor elements such as moisture. The report also highlights practical considerations for application of the specified adhesive and suggestions for achieving good results.

Use of Adhesives to Retrofit Out-of-Plane Distortion-Induced Fatigue Cracks (Mn/DOT 2006-04) is available on the Mn/DOT Web site.

Upcoming Events

Here are selected events related to transportation research. Visit the CTS Web site, www.cts.umn.edu/events, for more comprehensive event information. You may also subscribe to e-mail event announcements using our subscription form.

May 24-25, 2006
17th Annual CTS Transportation Research Conference, Saint Paul RiverCentre. Contact Shirley Mueffelman, 612-624-4754, conferences2@cce.umn.edu. [More]

May 24, 2006
CTS Spring Luncheon with Thomas DeCoster, Saint Paul RiverCentre. Part of the 17th Annual CTS Transportation Research Conference. Contact Shirley Mueffelman, 612-624-4754, conferences2@cce.umn.edu. [More]