Logo: University of Southern California

CS Undergraduate Team Wins FAA Challenge Competition Award

Voice recognition software listens to air control dialog, "and under certain circumstances, we can detect mistakes.”
Bob Melisso
June 14, 2010 —

An undergraduate team has won FAA recognition - and cash - for its design of a safer runway system.

From left: Clarence Li, Farid Nobakht,  Denis Tulskiy. On team, not shown; Ryan Berti, Ryan Brown, Tommy Holford, Michael Hsu, Pradyumna Kejriwal
“People make mistakes," said David Wilczynski, a lecturer in the Viterbi School computer science department. "The Federal Aviation Administration is trying to figure out ways to help. We built a system that actually listens to what the controllers are telling the pilot and, under certain circumstances, we can detect mistakes.”

Every year 64 million planes take off and land safely, but as airports get busier and busier the FAA is focusing on how to make the critical times when planes depart and arrive safer.

The term the FAA uses for mistakes during these crucial maneuvers are called “incursions” – for example, a plane taxis onto an active runway where another plane is about to take off, or a plane is preparing to land on a runway that already has a plane on it.

“The simplest example is that a controller may have just given clearance to a plane for takeoff or landing and then -- not too much later -- he gives clearance to another plane to cross the same runway. He’s made a mistake,” said Wilczynski.

The student project, called the Airport Communication-based Incursion Detector (ACID), hopes to catch such mistakes. ACID won third place in the FAA’s Runway Safety/Runway Incursions Challenge. The award came after a review by a panel of experts from FAA, industry and academia. A media backgrounder describes all the winning projects.

The student engineers in the design competition were Ryan Berti, Ryan Brown, Tommy Holford, Michael Hsu, Pradyumna Kejriwa, Zhongyuan "Clarence" Li, Farid Nobakht and Denis Tulskiy. All will share equally in the cash prize of $1,000.

Wilczynski’s students built a system that has a model of the airport built into it. Using that virtual airport map, the software can monitor an airport controller’s verbal orders to pilots and figure out what he or she intends.

Dave Wilcynski
David Wilczynski: developed team coaching instructional technique with Michael Crowley
“Our system listens to what the controller is saying and can detect errors by having a model of the airport and understanding the intentions of the controller.”

“So, we start from understanding the speech—that gives us a generic parsing – from that parsing we have a system that takes that and that is turned into rules that can detect incursions.”

If the orders conflict with safe operation rules programmed into the system – then the software warns of possible dangerous incursions of one plane into the safe operating space of another.

One of the big hurdles in any system that tries to understand speech is the weakness of speech recognition software. However, the limited language used in conversations between control towers and pilots makes the challenges a little easier.

“The conversations between pilots and controllers are in a very stilted language – it makes it perfect for this type of application,” says Wilczynski.

Plus, “we got some help from consulting with LAX (Los Angeles Airport) on the project.”

Below: a video dramatizes the incursion problem, with real voices:

LAX now uses an early warning system to warn off pilots that a runway will be in use by an incoming plane, using the very bright runway landing lights triggered by a radar system that detects a plane once it gets within range of the airport.

Wilczynski says that when he and his class spoke with LAX airport controllers, they suggested that the system they were building could be even more useful if it could understand when a controller told a plane’s pilot what its runway assignment was.

The software would consult its internal map of the airport, figure out which runway the controller had designated and then turn on that runway’s landing lights – a signal to other pilots that the runway is “in use” and to stay off it. This warning would come well before the airport’s current radar-based runway light system would be triggered – giving a larger margin of safety to the airport runway operations.

This project was a team effort. It’s an approach that Wilczynski has developed with another computer science faculty member, Michael Crowley. The pair routinely combine class projects to work on larger problems as teams.

Their team curriculum approach has been recognized by the Accreditation Board for Engineering and Technology (ABET)  the organization that reviews and accredits engineering curriculum across the country.

ABET called the team approach of Wilczynski and Crowley as having some of the most “innovative” ideas they have seen in undergraduate education.

Wilczynski believes that the team approach to solving a problem is a key to educating the next generation of engineers.

“We have team oriented projects… that’s what engineering is in the real world – you don’t sit there by yourself doing homework—people don’t give you little problems and say here -- do this and don’t talk to anyone else. That’s not the way the world works.”