CORVALLIS, Ore. – Engineers at Oregon State University and NASA have developed a new approach to air traffic control that might be able to reduce congestion in crowded airways by 50-60 percent, once the technology has been perfected and implemented.
The concept, the system engineers say, is to help existing pilots and air traffic controllers make local decisions that will better address the larger national concerns at any given moment, such as an airport closed by snow, thunderstorm delays in the Midwest, or major mechanical problems.
If the technology works as well as it has so far in successful computerized tests, it might be a significant aid to a problem estimated in one recent year to cost $3 billion, due to 437,000 flights delayed for a total of 322,000 hours - not to mention millions of frustrated travelers.
A new study on this approach to air control was presented at the Sixth International Joint Conference on Autonomous Agents and Multi-Agent Systems, held recently in Honolulu, where it received the “best paper” award.
“Air traffic congestion around the world is already very bad and it’s getting worse,” said Kagan Tumer, an OSU associate professor of mechanical engineering and expert on the control of large, multi-agent systems. “More than 40,000 flights a day move through U.S. airspace. Air traffic controllers have an extremely demanding job to do, and our approach is a new way to assist them, with what are actually some fairly minimal changes to the existing system.
“We think the approach could work,” Tumer said. “And if it does, this could have a huge potential impact on more convenient and efficient air travel in this country, which has been projected to double, if not triple, in the next 20 years.”
The crux of the problem, officials say, is that most air traffic controllers only have control, and even awareness of their own limited airspace. There are limits to the number of aircraft any one person can direct, enormous amounts of data to sort through, and it’s often unrealistic to think that an air controller in Philadelphia will be aware of all the latest issues that could be causing problems in Detroit, Des Moines, and Denver at the same time.
“We must emphasize that we’re not taking humans out of the loop or turning air control over to computers,” Tumer said. “What this approach does is help air traffic controllers re-direct or change the traffic flow intelligently, preventing an overflow in the first place or re-routing traffic in ways that help solve the overall concerns of the national system, not just a local airspace.”
Centralized routing strategies that are now in use often are slow to respond to developing weather or airport conditions, the scientists said in their report, and they allow minor local delays to cascade into large regional congestions.
“In the current system, air traffic controllers have some awareness of things like major airport closures, but they have to make multiple decisions about how to move and re-route traffic just based on their experience,” Tumer said. “That’s better than nothing, but you have 16,000 air traffic controllers at more than 5,000 public airports making different decisions with very little information to help guide them. The end result is decisions that are not always optimal for the larger, national needs.”
To address this issue, researchers created a multi-agent “algorithm” for traffic flow management, where an agent is associated with each “fix,” or location in two-dimensional space. Because aircraft flight plans consist of a sequence of fixes, this approach allows the agents to have direct impact on the flow of air traffic.
For example, by setting the distance required between aircraft going through a fix, an agent may direct aircraft that are still long distances away to slow down, speed up or otherwise change flight patterns to improve traffic flow, given all the constraints on the system at that particular moment. The key is to ensure that the agent at each fix makes decisions that benefit the full system and not just a local region.
“Conceptually, this is a way to align the incentives of an individual component with those of the overall system,” Tumer said. “What’s best for one component is not always best for the larger system.”
If one driver on a highway, for instance, decides to leave work early to beat rush hour, the plan may work. But if everyone in the city has the same idea, traffic still turns into a mess – about what you see in a big city every Friday afternoon before a holiday. What’s really needed is greater organization so everyone can get out of town in a reasonable manner.
In tests of the new system with an air traffic flow simulator, congestion was reduced up to 67 percent over a current industry approach. Mathematical and computerized approaches such as this are probably the only realistic way that future congestion can be addressed, researchers say – the U.S. is not going to be able to build its way out of the congestion with more airports and hugely expensive infrastructure.
Further studies are still needed with real congestion data, Tumer said, and extensive testing would need to be done with regulatory agencies before a system such as this could be implemented. Research has been funded by the NASA air safety program, and the recent report was co-authored by Adrian Agogino at the NASA Ames Research Center.
Conceptually, the researchers said, such approaches may also have value in dealing with other large system issues, including urban traffic management.