Cognitive workload quantification for air traffic controllers: an ensemble semi-supervised learning approach

• Main Authors: Yu, Xiaoqing, Chen, Chun-Hsien, Yang, Haohan
• Other Authors: School of Mechanical and Aerospace Engineering
• Format: Article
• Language: English
• Published: 2025
• Subjects: Engineering; Workload quantification; Semi-supervised learning
• Online Access: https://hdl.handle.net/10356/182319
• Institution: Nanyang Technological University

The human-automation collaboration has garnered widespread attention due to the significant challenges remaining in achieving full automation in air traffic control (ATC) systems. Accurately identifying the cognitive workload of air traffic controllers (ATCOs) is critical to ensure a seamless transition between human operators and automated control. However, previous studies have predominantly focused on discrete workload classification, which proves challenging to implement in real-time scenarios where workload levels fluctuate continuously. To tackle this problem, this study proposes a semi-supervised learning approach to quantitatively estimate the workload of ATCOs in a continuous value. Specifically, a co-training approach that integrates ensemble learning has been developed to collaboratively extract representations from EEG signals and eye movement data, leveraging both labeled and unlabeled samples. A self-collected dataset based on ATC scenarios has been constructed to evaluate the effectiveness and accuracy of the proposed method. The results demonstrate that our model can accurately provide continuous workload estimations for ATCOs and outperforms the baseline models. The validity of these estimations is further confirmed by comparing them with the task difficulties outlined in the experimental design. The findings of this study can facilitate the development of adaptive automation within future human-AI collaborative ATC systems, thereby enhancing overall operational efficiency and air safety.