Professor receives flight data grant

Plane crash encourages development of active system to prevent further sensor failure

Associate Professor Carlos Varela of RPI’s computer science department received a grant from the U.S. Air Force Office of Scientific Research to aid in his development of new programming techniques for better data management systems and their application to air travel.

Varela, who currently teaches Programming Languages, received his B.S. with honors, his M.S., and his Ph.D from the University of Illinois at Urbana-Champaign. He also works with various scientific journals, holding the positions of associate editor and information director of the ACM Computing Surveys journal as well as guest editor of Scientific Journal.

The motivation behind this endeavor, Varela mentioned, was the fatal crash of Air France Flight 447 in June of 2009. He asserts that the incident was in large part due to sensor failures, arguing that much of the problem was caused by the freezing over of a pitot tube—a device which measures air pressure in order to provide pilots with a measurement of airspeed. Varela, a licensed pilot himself, states that, when this freezing occurs, it can appear as though the plane is slowing down. Although pilots are trained to angle the nose of the aircraft downward in such a situation, he acknowledges that it can be difficult to prevent oneself from pulling back on the control yoke when altitude is being lost.

This action, Varela believes based on reviewing the black box-recorded cockpit conversation, is what ultimately caused the crash. When the angle between the nose of the plane and the relative wind extends past a critical point, he says, the wings can no longer provide lift, and the plane plummets to the earth.

All of this, he concludes, was the result of failed interpretation of erroneous data provided by the multiple sensors and meters installed on aircrafts. It is this problem that Varela hopes to address with his research.

His goal is “to come up with better programming techniques to manage data streams” that are both “spatial and temporal” in nature. The system will also model the relationship between certain variables, such as an aircraft’s groundspeed, its air speed, and the speed of the wind through which it is flying. Should the system work, it would help pilots “detect that a particular data stream is providing erroneous data.” Such a system could have prevented the crash of Flight 447.

Currently, flight plans are based on information obtained prior to the flight. These data streams, such as weather forecasts, Varela explains, are considered static. His system, however, will involve active data streams. The programming in aircrafts will be required to handle multiple sources of constantly changing information. “If weather patterns change on the day of a flight,” Varela says, “these changes can be put into the system and distribute updated information to pilots dynamically.” Similarly, if pilots observe adverse conditions while en route to their destinations, they can provide this information to the system, and flight plans can be adjusted accordingly.

Varela’s proposed system essentially involves a programming model, which is capable of reasoning with multiple data streams. It will, he says, be an extension of an existing logic programming language, such as Prolog. Combining his work with the capabilities of such a language, Varela believes the system will be able to “detect errors with a high probability.” If airlines were to implement his system, pilots will be flying based on active flight plans which can change while the plane is still in the air.

The grant Varela received provides him with $100,000 in funding for one year with more available for a second year based on performance. Having acquired this grant, he hopes it will allow him to “network with members of the Air Force in the future.”

As for the future of the research, Varela believes that his active data system could be “very applicable to other fields and domains.” He speculates it could even change the field of computing. Currently, programming is based around passive data access. Active data access, which Varela says would be possible with his work, could result in an overhaul of modern computing systems.