University of Minnesota
September 16, 2008
At a rural Minnesota intersection, a graduate student checks a LIDAR apparatus, a laser-based system to track sizes and positions of vehicles.
Photo: Jonathan Chapman
Researchers at ITS are out to improve all kinds of transportation
By Deane Morrison
A bus driver on a freeway wants to save time by using a "bus-only shoulder," but driving snow obscures the shoulder boundary. A rural driver crossing a busy highway misjudges the traffic and gets broadsided. The National Oceanic and Atmospheric Administration wants to monitor the oceans for lost, miles-long commercial fishing nets that can wreak havoc on coasts and marine life, but manned surveillance planes are too expensive.
Vexing—and sometimes fatal—transportation problems like these may seem intractable, but not to researchers at the University of Minnesota’s Intelligent Transportation Systems (ITS) Institute. Drawn from several University departments, they are designing and testing new technologies to make all kinds of transport safer and more efficient.
No wayward buses
Rush hour congestion is the bane of many an urban commuter. To reduce it, several major Twin Cities arteries have dedicated bus lanes on the right shoulder. The lanes are barely wider than a bus, however, and navigating them can be tricky.
"The biggest problem is low visibility," says Craig Shankwitz, director of the University’s Intelligent Vehicles Laboratory, which is associated with the ITS Institute. And in winter, "plows move snow onto the shoulders, making it very difficult for drivers to see the shoulder boundaries."
Not to mention the irresponsible driver who resents the buses’ advantage and veers partway into their lane to block them.
To help bus drivers, Shankwitz and his colleagues are testing a system originally developed by ITS researchers to guide snowplows in bad weather. A major element is a display in the driver’s forward field of view that shows how the bus is moving with respect to its lane.
The display begins with highly accurate data on the bus’s position and heading from an onboard GPS receiver. A computer uses this information to query an onboard database containing the GPS coordinates of area roads and their lane boundaries. Combining these data, the computer generates a virtual view of the road by which the driver can safely steer.
Another feature is a radar that looks for obstacles in the lane and shows them in the display. Also, the sides of the bus are equipped with LIDAR, a laser-based detection system that tracks vehicles and pedestrians on either side of the bus and displays them on a monitor on the bus’s instrument panel.
Further, if the bus strays to one side, the driver’s seat will vibrate on that side in warning. And the steering wheel will try—not too hard—to steer the bus back into its lane.
The system will be tested in 2009 on 10 buses covering two express routes between the southern Twin Cities metro area and downtown Minneapolis.
"[Our] goal is to improve schedule adherence and efficiency and to increase ridership," says Shankwitz. And, to lure even more riders, "we provide wireless Internet to passengers on these buses."
In Minnesota, about 70 percent of fatal car crashes happen in rural areas. Often, drivers crossing a busy highway get hit when they misjudge distances between oncoming cars. Now, a project headed by ITS Institute Director Max Donath, a professor of mechanical engineering, aims to help drivers determine when cars in the cross traffic are spaced adequately to allow crossing.
"We want to communicate to stopped drivers when to enter intersections, using new kinds of signs that contain information coupled to sensors along the road," Donath says. In August and September 2008, the signs get a real-world test at a rural Minnesota intersection between a county road and a trunk highway.
In those tests, radar will gauge the location and speed of vehicles on the highway, says Shankwitz, who is also on the project. LIDAR will track the size and position of vehicles on the county road. A computer will use these data to determine when it is unsafe for test vehicles on the county road to cross and will so indicate on the signs. The test vehicles will be a snowplow and a passenger car, both outfitted with head- and eye-trackers to tell the researchers what people look at as they decide when to cross the highway.
If all goes well, "people won’t be fixating on the signs, but using them as an assist," says Shankwitz.
And if driver behavior indicates an improvement in safe decision making, the researchers will ask the U.S. Department of Transportation to support a field test in which the general public will be exposed to the system and its long-term benefits can be determined.
Spinning straw into gold
Can an airplane with only a four-foot wingspan carry out sophisticated flight maneuvers safely?
Small planes have enormous potential for tasks like monitoring hard-to-reach areas and military or police surveillance. But as ITS researcher Demoz Gebre-Egziabher well knows, proving to the FAA that a very small craft is capable of doing such a job both safely and autonomously isn’t easy.
The problem is that to guide itself, a plane must have sensors to detect its position, speed, and orientation in space. But the highly precise sensors used on commercial jets are too expensive and heavy for tiny planes.
"Our research is on how to scale it down [without compromising safety] and reduce its cost," says Gebre-Egziabher, an associate professor of aerospace engineering and mechanics.
Therefore, instead of a few high-quality sensors, Gebre-Egziabher uses many inexpensive ones that individually are too inaccurate to satisfy FAA safety regulations but that together could do the job. The rub comes in trying to figure out which sensors can be trusted at any given time and how to write computer programs to deal with the situation.
For example, suppose a sensor monitoring the orientation of the airplane fails. The mechanism for detecting failure depends on other sensors that are also subject to failure and inaccuracy. And what to do if sensors disagree? How does one spin the straw of iffy sensor readings into the gold of high-quality data?
To satisfy FAA requirements, the probability of an undetected failure that could lead to a crash or collision must be kept below one in a million, says Gebre-Egziabher.
"We can’t do a million experiments," he explains. "We have to come up with statistical models of [how sensors behave]."
He and his students have used such modeling to build a small plane equipped with multiple sensors and a camera, which they are now testing. The Minnesota Department of Transportation and state patrol are interested in whether the plane can be used to inspect highway conditions and infrastructure. It can fly on its own, but stays in sight of a ground controller who is always ready to take control.
If they’re perfected, Gebre-Egziabher says small autonomous planes could be used for tasks like searching for lost people or survivors of disasters, or for dangerous operations like crop dusting or inspecting miles and miles of power lines. And maybe for finding those deadly drifting "ghost nets" that plague our oceans.