AI Goes To Flight School
In 2017, a shortage of pilots in the Air Force flight training reached a margin of over 2,000 pilots in the coming years, a problem echoed by other branches of the military.
Low recruiting numbers are compounded by declining retention of trained pilots, as civilian airline carriers (who are facing shortages of their own) have increased compensation to become more attractive options. In order to meet the country’s defense needs, there’s significant pressure to train pilots more efficiently and effectively.
On a fundamental level, aviation training in the Air Force, Navy, and Army hasn’t changed much since the 1920s. While the equipment has become more sophisticated, the education methods remain reminiscent of machine assembly:
Move each student along a conveyor belt of training using a building block approach, and discharge those who don’t meet the criteria. While this model ensures high-quality pilots, it also rejects candidates that may only need a little extra time in training beyond the standard “three strikes” policy.
The government has already invested substantial time and money into these student pilots that attrite, so there are significant incentives to increase retention by both adding needed personnel and improving ROI.
As training methods are adjusted to align with the modern classroom, artificial intelligence, virtual reality, and augmented reality (AR) have the potential to aid student pilots, instructors, and pilots fulfilling continuing education.
An Iterative Syllabus
AI can be harnessed to find insights into very large data sets on a periodic basis at affordable costs. A systematic, periodic flight syllabus review may reveal powerful suggestions for change.
Inputs could include not only the scores of specific maneuvers on event grade sheets but also other data such as undergraduate major, weather on the training sortie, and days from last flight.
Given the bandwidth of the senior flight instructors, this syllabus review process is currently incomplete and infrequent. The cultural change of being willing to differentiate the syllabus within reasonable boundaries versus the cookie-cutter approach would be far more economical and produce better-trained pilots.
Personalized Flight Training
Applying AI to the analysis of flight simulator data and individual flight training event data could yield even more awareness for the student in the training process.
Currently, simulators rely primarily on human supervision, and analysis of simulator data is limited to post-event debriefs of the recorded mission. Thus, training could potentially suffer from instructor bias or not provide enough information for the student to improve upon.
For example, a student could continually struggle to make timely decisions but not know which specific actions were being executed slower than normal.
Collection and analysis of biometric information from eye trackers and heart monitors could give a more complete picture of a trainee’s performance. Because AI excels at collating information from different sources, it could show that, for example, the student spends too much time on an instrument panel scan or focuses on the incorrect sensor.
With the available data, the instructor could spend more time helping the student prioritize their scan and ensuring they feel fully prepared before incorporating additional tasks.
There is also a large amount of individual flight training data that could provide key insights and thus faster learning to the flight students.
In Navy flight training, for example, each student completes hundreds of practice aircraft carrier landings at a runway prior to embarked aircraft qualifications. Using AI to quickly and routinely analyze throttle position and angle of attack during these practice sessions would add immediate value.
This analysis would also invariably reduce the disqualification rate of these expensive carrier landing qualification training evolutions and produce a better trained naval aviator.
Augmented reality powered by AI could also be used to substitute or enhance ground-based simulator training.
Specifically, AR mission rehearsals and “warm-up” flights could be used as refresher training on aircraft carriers where physical space is limited.
As cockpit technology progresses toward sensor fusion, this will inevitably come to fruition and increase mission readiness with limited additional cost.
Looking Ahead: Continuing Education
After completing their flight training and earning their wings, all pilots must log hours on various topics to stay up to date.
If AI created more personalized training, records could also be tracked over time. This would place more emphasis on areas in need of improvement. A more robust tracking system could also expose flight training weaknesses and how to overcome them.
This would be far more cost-effective than the current approaches to training and readiness.
What could this look like? Picture a chip that would stay with a pilot throughout their career (or lifetime), maintaining and analyzing training and proficiency records and eventually directing training.
This would enable organizations to break the time-scheduled proficiency model, replacing it with an as-needed paradigm. AI will make pilot training more efficient and effective in the coming years. The rate of adoption and depth of application will directly correlate to cost savings and increased output in DoD pilot training.
As a country, we must embrace AI to retain our global lead.