RESEARCHERS USE FUZZY LEARNING FOR AIRCRAFT CONTROL

COLUMBUS, Ohio -- Researchers here have taken the first steps toward developing a system using "fuzzy logic" to help pilots regain control of aircraft following a major system malfunction.

The system, which is under development, would assist pilots in accommodating for failures which may lead to airplane crashes, said Kevin Passino, an assistant professor of electrical engineering at Ohio State University.

"Investigations of many aircraft accidents find that even with some of the most severe unanticipated failures, the aircraft could have been saved if the pilot had taken proper actions in a timely fashion," Passino said.

"But because a catastrophic event often happens fast, and given the level of stress and confusion during these incidents, it is understandable that a pilot may not find the solution intime to save the aircraft."

Passino and Stephen Yurkovich, an associate professor of electrical engineering at Ohio State, wanted to develop a control system that would automate pilot expertise and take action to save the aircraft as soon as an error is noticed.

Fuzzy logic is a method to give inexact operating instructions to machines. It takes into account the vagueness present in many situations, such as an emergency on an airplane.

This type of system might have helped pilots save American Airlines Flight 191, which crashed on May 29, 1979 at Chicago-O'Hare International Airport. An engine and hydraulic system malfunction on the DC-10 wasn't detected in time for pilots to save the plane.

Studies of the crash on a simulator suggested the plane could have been flown successfully despite the problems, Passino said. A fuzzy logic system may have been able to warn pilots of the failure of the hydraulic system in time for them to make necessary adjustments. Or, a fuzzy logic system may be able to make the necessary adjustments in the aircraft controls to save the flight.

Researchers have examined fault-tolerant aircraft control systems using conventional controls, but this is the first time fuzzy logic has been applied to this problem, Passino said.

Fuzzy logic may be a preferred method because conventional control systems are based on mathematical differential equations, which sometimes do not facilitate the translation of human problem-solving techniques into a computer algorithm.

"Fuzzy control systems use a linguistic approach which allows us to express the desired control actions in words," he said.

Passino and Yurkovich have tested their theories in a simulator they developed with researchers from Wright Laboratories. Results indicate a fuzzy control system may also be able to perform differently for various aircraft system failures.

"If you have a failure, you can't expect an aircraft to perform as well as it did before," Passino said. "This system automatically recognizes that and tries to seek the level of performance that is appropriate to the level of failure that has occurred."

The next step in the research is to study the stability of a fuzzy learning control system and decide how the system would be integrated in an aircraft and used by a pilot, Yurkovich said.

"We want to be able to design a controller that we know will stabilize a system and keep it stable," he said. "The development of stability theory for fuzzy control systems is only in its infancy. There's a lot more work to be done."

The researchers plan to design more challenging failure scenarios to test their control system. A final phase of the work would be to field test the system in commercial and military aircraft.

The research is funded by the National Science Foundation and was published in a recent issue of Proceedings of the IEEE.

Back to Archive
Go to Current Month News Research Stories
Go to Current Month Newsfeature Stories
Go to Current Month News Cancer Report Stories

Contact: Kevin Passino, (614) 292-5716

Written by Kelli Whitlock, (614) 292-9475