Loading page…
The university’s October 6 account highlights reliability research, not completed certification. Air Space Intelligence says controllers would keep responsibility for real-time decisions.
Listen to this story
UC’s AI Bio Lab and Washington Dulles are pursuing research into whether aviation AI can be made reliable and certifiable, but professor Kelly Cohen described certification as a goal—not a completed safety finding. Separately, the FAA is considering Air Space Intelligence software for route and schedule planning, with a proposed rollout at Dulles and two other Mid-Atlantic airports. The Dulles research is not identified as validation of that vendor’s system, and controllers would retain real-time decision authority.
Air Space Intelligence says its software is already in use at airports in Alaska.
The proposed Mid-Atlantic expansion names Dulles, Ronald Reagan National, and Baltimore-Washington International airports.
Cohen told WKRC Local 12 that trusting aviation AI requires thorough verification of every part of the system.
University of Cincinnati researchers are working with Washington Dulles International Airport to test AI reliability, the university disclosed in an October 6 account of a Local 12 interview. Aerospace engineering professor Kelly Cohen called for thorough safety verification as the Federal Aviation Administration considers software to help traffic managers organize flight routes and schedules.
WKRC Local 12 interviewed Cohen about the risks and benefits of the FAA proposal. He directs UC’s AI Bio Lab, where he and his students work on trustworthy, safe AI for aviation. The lab’s collaboration with Dulles concerns AI reliability; it is not identified as a test of Air Space Intelligence’s particular software.
“What we are pushing together, my lab and the team, is certifiable,” Cohen told Local 12. That describes the research goal, not an achieved certification or a completed safety finding. His requirement for trusting aviation AI extends to verifying every part of it.
AI is having to get to these guarantees that it would be able to perform in a very safe manner. As Ronald Reagan said, trust but verify. So to be able to trust the AI, we need to make sure that the verification process of every part of it is very thorough.
Kelly Cohen, University of Cincinnati aerospace engineering professor, speaking to WKRC Local 12
The deployment proposal gives that research an operational backdrop. Contractor Air Space Intelligence is offering the FAA software to help traffic managers organize routes and schedules. The system is already in place at airports in Alaska. The company’s planned expansion would bring it to three Mid-Atlantic airports:
Air Space Intelligence says air traffic controllers would continue to be responsible for real-time decisions. The proposed division is between software supporting route and schedule planning and people retaining authority over live traffic—not an announced replacement of controllers with an automated decision-maker.
Loading discussion...
Join the conversation
Explain what else, if anything, would give you confidence.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.