Training the Next Generation of Pilots with AI-Powered Adaptive Learning Applications
AI Redefined (AIR)’s Cogment Adaptive Learning Delivers Anytime, Anywhere Training to Fit Individual Student Needs
Electric vertical take off and landing (eVTOL) aircraft are an emerging technology in the aviation industry aimed at offering a quiet, low-emissions mode of short-range air transport. The development of this technology is happening at a time of unprecedented workforce shortages across the aviation industry, a phenomenon which is due in part to challenges of cost, time, and access to training resources, required in order for learners to become fully proficient. Systems Technology, Inc. (STI) and AI Redefined (AIR) have partnered to deliver a data-driven approach to eVTOL training to reduce training time and cost, increase efficiency, and meet regulatory safety standards. Integrating with simulation-based training, AI Redefined's Cogment Adaptive Learning platform uses AI to identify mistakes, provide feedback, and create novel training content targeted to each individual learner's needs. The design of this application will allow students to train on-demand and at their own pace, shifting training from an hours-based to a performance-based model, and providing a path toward better student outcomes in training programs to meet workforce needs.
An Emerging Industry
By 2050, it is projected that two out of every three people on earth will be living in an urban center. As populations grow, there will be increased pressure on existing transportation infrastructure, necessitating innovative approaches to alleviate congestion and reduce environmental impact. Electric Vertical Takeoff and Landing (eVTOL) aircraft offer a new dimension of mobility: these ‘air taxis’ allow for more flexible transit routes and a means of avoiding traffic congestion. eVTOLs are a quiet, low-emissions alternative to combustion engine-powered transport, offering a green transport option that reduces noise pollution.
While this new technology presents exciting opportunities for faster and greener transportation options, the eVTOL industry is emerging at a time of critical workforce shortages across the aviation sector. It is estimated that by 2028, the industry will have a need for 60,000 fully qualified pilots. This highlights the need for effective, efficient, and dynamic training solutions that have never been more critical.
A Key Partnership
In November 2023, AI Redefined and Systems Technology, Inc (STI) announced a partnership to develop training solutions for eVTOL aircraft.. STI is a leader in aerospace controls and simulated training assessment for more than 65 years, known for providing guidance & controls expertise to NASA, FAA, the US Department of Defense and other significant aerospace industry players. The goal of the partnership between STI and AIR is to revolutionize the training landscape for eVTOL pilots by harnessing the power of adaptive learning with advanced artificial intelligence techniques and state-of-the-art simulation technology.
This groundbreaking learning tool will provide eVTOL pilots with a highly immersive and realistic training experience, allowing them to hone their skills in a safe and controlled environment, while receiving highly personalized feedback and content recommendations aimed at focused skills development.
Cogment Adaptive Learning Changes The Training Paradigm
At the heart of this new tool is AI Redefined’s Cogment Adaptive Learning platform. Cogment Adaptive Learning integrates with any simulation-based learning environment to provide a personalized training experience for learners in safety-critical fields such as aviation or heavy machinery operation. As learners practice skills and techniques in flight simulation exercises, Cogment Adaptive Learning uses AI to identify mistakes, coach learners, provide feedback, and create novel training content tailored to each individual learner's needs.
Together, AIR and STI have developed a proof of concept demonstrating the use of Cogment Adaptive Learning tool for eVTOL pilot training on a hover-to-land maneuver. This is fully coupled with a piloted flight simulator, to demonstrate training in the realistic scenario. Cogment Adaptive Learning helps eVTOL pilot students learn in 2 ways:
Continuously assessing performance and providing feedback to give insight for skills improvement
Determining how to modify and recommend new content to give opportunities for practice on challenge areas
Providing Insight For Skills Improvement
As the student runs the simulation exercise, the AI-Evaluator watches student performance and provides assessment and feedback both ‘online’ and ‘offline’. The AI-Evaluator works by comparing the student performance to a reference agent trained with reinforcement learning to learn optimal task behaviour.
The Evaluator looks for areas where student performance deviates from the reference agent or violates specific task criteria in order to provide the student with both qualitative and quantitative feedback on how to improve performance in the future.
This component enables 3 key features to help learning:
Online coaching support, providing live feedback to correct behaviour that would lead to mistakes
Offline replay and debriefing, providing a full review of the complete training exercise, including a comprehensive summary of performance and areas for improvement
Origin of error analysis, looking at what behaviours led to down-stream consequences that resulted in errors
Recommended Content to Practice Challenge Areas
The AI-Director uses feedback from the AI-Evaluator to inform its recommendations for subsequent exercises to focus skills improvement. The AI-Director recommends content of the student’s next exercise to focus on the particular parts he/she was challenged by in previous runs.
The Director recommends variations on task parameters including the skills/techniques to be practiced, the task conditions (weather, initial conditions), and the level of help or coaching to be provided by the Evaluator in the student’s next exercise.
Advantages of Training with Cogment Adaptive Learning
The Cogment Adaptive Learning platform will allow students to train at their own pace, working through challenge areas and receiving feedback even when human instructors are not available. The goal of this platform is not to replace human instructors, but rather Cogment Adaptive Learning aims to make one-on-one time between students and instructors more efficient and impactful by allowing them to review previous lessons together and provide insights at a higher level.
By programmatically adjusting training content to target challenge areas, Cogment Adaptive Learning ensures a variety of scenarios for students to learn on, alleviating the bottleneck of limited operational resources to design exercises for an individualized learning experience. This application allows students to work through the necessary steps to master specific techniques before moving on to more complex exercises. The development of the training content, AI-behaviour, learning outcomes, and other details are being developed together with subject-matter experts, with experience in training and eVTOL domain.
The aim of this approach is to:
Improve student retention rates by enabling on-demand, anytime, anywhere access to training resources
Improve student learning outcomes by creating individualized training content oriented towards targeting challenge areas, enhancing competency development
Reduce the impact on operational resources by optimizing in-person training time between instructors and students. Instructors can be freed up to focus on more impactful coaching during 1:1 sessions with students
Reduce cost of creating pedagogical content by generating scenarios programmatically and validating with instructors
Enhance data management capabilities, including scenario creation, student performance analytics, archiving their training progression, etc. since all relevant interactions between students, instructors, and the various parts of the system (simulation, evaluation, direction) are logged with varying degrees of granularity.
Continuing to Expand the Cogment Adaptive Learning eVTOL training system
Initial evaluations of the Cogment Adaptive Learning eVTOL training system have been positive, focusing on direct comparison between traditional teaching and instruction methods and the new capabilities offered through the use of data driven approaches. Limited trials and assessments have been conducted by pilots using the simulation used for the landing task. This initial feedback has since led to developments in the user interface, evaluation scheme and student feedback.
Progressing in the project, currently AIR and STI are working to build out a wide variety of training modules which cover aspects of the complete eVTOL pilot training lifecycle, from learning systems and procedures, to specific maneuvers, all the way to complete flight missions. As the training landscape for future safe integration of eVTOL aircraft continues to develop, AIR and STI are continuing to develop the platform to solve the industry needs.
Shifting toward performance-based training over hours-based certification models will be a critical component of modernizing the training workflow to meet workforce demands. AIR and STI look forward to continuing to develop this training tool to meet the evolving needs of eVTOL pilot training programs worldwide.
