This article also appears in
Subscribe now »

“We don’t necessarily want [air traffic controllers] to spend the bandwidth on processing 40 pieces of information. Instead, we can tell them the three top choices,” says Hamsa Balakrishnan, associate professor of aeronautics and astronautics at MIT and a member of MIT’s Institute for Data, Systems, and Society. (Bryce Vickmark)

Engineering smart systems for smart air travel, balancing human constraints

There are no shortage of predictions (or, “market outlooks,” as is the phrase) in the aerospace industry. And they are necessary as such predictive data feeds into decisions about how many planes to make, for whom, and for where.

Some experts predict that over the next 25 years the number of passengers flying through U.S. airport hubs will grow by almost 70%, to more than 900 million passengers per year. And while airframers all profess to being up to the task of making more and more airplanes to accommodate more and more passengers, by most accounts and predictions, the airspace to fly them all in is not getting any bigger. Airspace that doesn’t magically expand equates to delays. And, truth be told, airspace doesn’t magically expand.

As traffic continues to expand, any local delays, from a congested runway to a weather-related cancellation, could ripple through the aviation system and jam up a significant portion of it, making life increasingly difficult for a number of people, and not just for passengers and crew. Air traffic controllers’ jobs are also affected, becoming increasingly difficult.

“The system is large, and there’s a lot of connectivity,” said Hamsa Balakrishnan, Associate Professor of Aeronautics and Astronautics, MIT. “How do you move along today’s system to be more efficient, and at the same time think about technologies that are lightweight, that you can implement in the tower now?”

Such are the questions that Balakrishnan is seeking to answer. She is working with the FAA and major U.S. airports to upgrade air-traffic control tools in a way that can be easily integrated into the existing infrastructure. These tools are aimed at predicting and preventing air-traffic delays, both at individual airports and across the aviation system. They will also ultimately make controllers’ jobs easier.

“We don’t necessarily want [controllers] to spend the bandwidth on processing 40 pieces of information,” says Balakrishnan, who is a member of MIT’s Institute for Data, Systems, and Society. “Instead, we can tell them the three top choices, and the difference between those choices would be something only a human could tell.”

Most recently Balakrishnan has developed algorithms to prevent congestion on airport runways. Large hubs like New York’s JFK can experience significant jams, with up to 40 planes queuing up at a time, each idling in line—and generating emissions—before finally taking off.

Balakrishnan found that runways run more smoothly, with less idling time, if controllers simply hold planes at the gate for a few extra minutes. She has developed a queuing model that predicts the wait time for each plane before takeoff, given weather conditions, runway traffic, and arriving schedules, and she has calculated the optimal times when planes should push back from the gate.

In reality, air traffic controllers may also be balancing “human constraints,” such as maintaining a certain level of fairness in determining which plane lines up first. That’s why a large part of Balakrishnan’s work also involves talking directly with air-traffic controllers and operators, to understand all the factors that impact their decision making.

“You can’t purely look at the theory to design these systems,” said Balakrishnan. “A lot of the constraints they need to work within are unwritten, and you want to be as nondestructive as possible, in a way that a minor change does not increase their workload. Everybody understands in these systems that you have to modernize. If you’re willing to listen, people are very willing to tell you about what it looks like from where they are.”

When Balakrishnan was at Stanford, she shifted her focus from fluid dynamics to air traffic and control-related problems, first looking at ways to track planes in the sky.

“That got me interested in how the rest of the system works,” she said. “I started looking at all the different decisions that are getting made, who’s deciding what, and how do you end up with what you see eventually on the data side, in terms of the aircraft that are moving.”

After graduating from Stanford, she spent eight months at NASA’s Ames Research Center, where she worked on developing control algorithms to reduce airport congestion and optimize the routing of planes on the tarmac.

In 2007, Balakrishnan accepted a faculty position in MIT’s Department of Aeronautics and Astronautics, where she has continued to work on developing algorithms to cut down airport congestion. She’s also finding practical ways to integrate those algorithms in the stressful and often very human environment of an airport’s control tower.

She and her students have tested their algorithms at major airports including Boston’s Logan International, where they made suggestions, in real time, to controllers about when to push aircraft back from the gate. Those controllers who did take the team’s suggestions observed a surprising outcome: The time-saving method actually cleared traffic, making it easier for planes to cross the tarmac and queue up for takeoff.

“It wasn’t an intended consequence of what we were doing,” said Balakrishnan. “Just by making things calmer and a little more streamlined, it made it easier for them to make decisions in other dimensions.”

Such feedback from controllers, she says, is essential for implementing upgrades in a system that is projected to take on a far higher volume of flights in the next few years.

“You’re designing with the human decision-maker in mind,” said Balakrishnan. “In these systems, that’s a very important thing.”

Continue reading »