Using AI in aviation training could help ease pilot shortage

A significant shortage of commercial pilots is one of the major problems faced by Canadian airlines today, contributing to the recent delays and cancellations plaguing the airline sector. One Calgary-based company is exploring how artificial intelligence can help build a platform to ease that shortage by training pilots faster and more efficiently.

Delphi Technology Corp. is working with two professors at the University of Alberta to develop a virtual training course that aims to reduce the time and cost required to train new commercial pilots. The partnership will explore integrating ML into the platform to better identify a trainee's strengths and weaknesses.

The original concept for the machine learning component was developed with support from Amii. Afterwards, Delphi has continued to develop the project with U of A professors Carrie Demmans Epp and Matt Taylor; Taylor is also a Fellow & Canada CIFAR AI Chair at Amii.

According to Delphi CEO Alan Tay, Canadian airlines have been dealing with a shortage of commercial pilots for years. It's an issue that he is deeply familiar with, having spent 25 years as both a commercial and military pilot, as well as a flight instructor.

The issue has become much worse since the COVID-19 pandemic. Faced with travel bans and lack of demand, many pilots and other airline workers took early retirement or left the industry. Now, with air travel once again ramping up, there aren't enough qualified workers to replace them. Tay estimates that to return to the levels of air travel seen before the pandemic, Canada would need to train thousands of new pilots and other aviation professional workers.

It's not a quick fix. That training is both time-consuming and expensive — it involves hundreds of hours of instruction and time spent in the air. According to one Calgary flying club, obtaining a pilot's licence in Canada can cost students between $25,000 and $35,000.

You also need people to teach them. That's where Tay sees a major bottleneck — a lack of instructors. He says it is hard to attract people to the role; the pay is often low and the job is often seen as unglamorous.

"It is the lowest part of the ecosystem in commercial aviation career paths. No one wants to be an instructor," he says. "Nobody wants to be on the ground. Everyone wants to be in the big jet. You want to be in the sky."

'Not like a cookie cutter'

Delphi's virtual training platform, called VR City, aims to take some of the strain off of the system. Tay says that one barrier to pilot training is that typically it is all done in person. Trainees will meet in a classroom setting in large groups for instruction. Later in their training, they'll also spend hundreds of hours in flight with an instructor to gain practice.

By moving that instruction and training to a virtual platform, Tay says flight instructors can teach more potential pilots in the same amount of time. That will free them up for more advanced instruction that needs to be done face-to-face, as well as conducting the validation and testing to make sure trainees have qualified for their licence.

Delphi has partnered with the University of Alberta to integrate machine learning into the virtual platform to make it more responsive to student needs. One of the major barriers in pilot training is that it is often a one-size-fits-all approach: trainees spend the same amount on the same subjects. But Tay points out that many potential pilots come in with previous experience. An aviation mechanic, for example, who wishes to become a commercial pilot probably doesn't need as much instruction on the mechanical aspects of a plane.

The partnership with Demmans Epp and Taylor will build a model that will assess each trainee's strengths and weaknesses, much like a human instructor. It could help identify areas where the trainee needed more training, reducing wasted time and resources.

"It's not like a cookie cutter where everyone goes through the same thing," Tay says. "It's able to help us to assess a person's competence level, just like any other instructor."

Delphi is currently developing the platform and working with aviation regulators to approve the training process. The rules around how pilots are taught to fly jets are understandably strict and complex, so Tay says it is a slow process. But they have already seen some success — the European Union Aviation Safety Agency has approved similar technology for use in pilot instruction. The company is currently in talks with aviation authorities globally.

"We are working with the system, not around the system," he says.

While the focus was on training pilots, Tay says the virtual training program is expanding to other jobs in aviation and aerospace, including mechanics and airline attendants. Delphi will soon adapt their training platform to other industries, like museums, healthcare and manufacturing.

Find out how to bring your business' artificial intelligence ideas into the real world. Check out our Industry page to learn how Amii can help.