Automatic Gaze Classification for Aviators: Using Multi-task Convolutional Networks as a Proxy for Flight Instructor Observation

Wilson, Justin; Scielzo, Sandro; Larson, Eric C.

doi:10.15394/ijaaa.2020.1499

Cited by 1 publication

(1 citation statement)

References 37 publications

(46 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…They have collected data with labels from the Air Force Personnel Center and applied different supervised ML models such as random forest, K-nearest neighbors and neural networks to predict the success patterns for pilots based on their historical testing records. Wilson et al [11] have studied the problem of gaze classification to evaluate the performance for aviators to analyze if the pilot trainee would scan too rapidly, omit, or fixatewhich are common errors when scanning the horizon, and cross-checking instruments. Traditionally, the examination for the gaze of pilot trainees relies on a manual check from flight instructors.…”

Section: Related Workmentioning

confidence: 99%

Artificial Intelligence in Pilot Training and Education – Towards a Machine Learning Aided Instructor Assistant for Flight Simulators

Yang

Lammers

et al. 2021

Communications in Computer and Information Science

View full text Add to dashboard Cite

The a viation industry was set to see unprecedented growth over the next two decades. Key occupations predicted to be in shortage included not only pilots, but also flight instructors. Undoubtedly, Covid-19 is currently having a huge impact on the industry. Nevertheless, the current environment further strengthens the need for pilots to maintain their training. Consequently, there is pressure to deliver high-quality training outcomes for an increasing number of pilots and trainees with limited resources available. Current simulator-based training schemes are limited by placing a significant reliance on the personal experience of flight instructors to assess pilot performance. Finding ways to increase the quality and efficiency of simulator-based training is therefore of high importance. With recent advances in artificial intelligence, it is possible to use machine learning techniques to extract latent patterns from massive datasets, to analyze pilot trainees' activities, and to provide feedback on their performance by processing hundreds of different parameters available on flight simulators. An ML-aided pilot training and education framework is needed that exploits the power of the ML techniques for more objective performance evaluation. In this paper, we describe a conceptual framework for such a system and outline steps toward the development of a full digital instructor system with the potential to overcome current limitations and enabling comprehensive and meaningful feedback that is tailored to the individual need of the trainee.

show abstract

Section: Related Workmentioning

confidence: 99%