In This Story
With his research career barely off the ground, Shahab Aref has already made air travel safer for millions. A PhD student in the Systems Engineering and Operations Research Department, Aref created a tool used by the Federal Aviation Administration (FAA) to evaluate new flight procedures safety.
A paper Aref wrote with advisors John Shortle and Lance Sherry on synthetic flight trajectory generation earned both best in session and best in track awards at the 44th Digital Avionics Systems Conference. The project, supported by the FAA and fueled by a practical need, got off the ground due to years of research by Shortle and Sherry.
“When a new flight procedure is proposed, they need to make sure it is safe,” Aref explained. Doing so requires enormous volumes of data—sometimes tens of millions of flight scenarios—to fully understand risk across varying conditions. Generating that data traditionally is not feasible. “It’s not possible to ask a pilot to go sit down in a flight simulator and fly a procedure 30 million times,” he said.
To address this limitation, Aref and his collaborators turned to AI. Using a model called a variational autoencoder, they developed a method to generate synthetic flight trajectories, simulated flight paths that replicate real-world patterns.
The key breakthrough in Aref’s work was ensuring the realism of the synthetic data. “Any other paper that used AI to generate flight trajectories just used different statistical distributions,” Aref said. Instead, his team introduced a new approach using physics-based metrics.
By incorporating factors such as course change, bank angle, rate of descent, and deceleration, the team could evaluate whether each individual trajectory was not only statistically similar to real data, but physically flyable. Moreover, rather than comparing clusters of data, they assessed each flight path on its own merits, bringing a new level of precision and reliability to the process.
After the paper’s publication, the research moved beyond theory into practice. “We converted that method to a tool,” Aref said, describing a software platform the team designed for the FAA. The tool enables users to generate synthetic flight trajectories for specific airports and evaluate them using the same physics-based metrics developed in the research. It is now actively used by the FAA.
“Aref exhibited creativity and perseverance in overcoming several challenges in the research,” said Sherry. “It has been a privilege to collaborate with him.”
Aref began his PhD in Spring 2023 and is now approaching his fourth year, with his dissertation closely tied to the intersection of AI and aviation safety. Seeing his research adopted by a federal agency so early in his career has been a defining experience. “I’m really glad that this is my project,” he said. “That’s really impactful.”