Pilot mental workload analysis in the A320 traffic pattern based on HRV features.

Area of Science:

• Aviation Psychology
• Biomedical Engineering
• Machine Learning

Background:

• Pilot mental workload is crucial for flight safety, especially during demanding phases like takeoff and landing.
• Assessing workload accurately is vital for preventing cognitive overload and ensuring safe operations.

Purpose of the Study:

• To evaluate pilot mental workload across various flight phases using heart rate variability (HRV) and machine learning.
• To develop a reliable framework for real-time pilot workload monitoring and prediction of cognitive overload risks.

Main Methods:

• Collected heart rate data during simulated A320 traffic pattern flights.
• Utilized selected HRV features (Min_HR, SDNN, SD2, Modified_csi) and machine learning classifiers (RF, KNN, GBDT, XGBoost).
• Trained and evaluated models for pilot mental workload level classification, comparing performance with and without feature selection.

Main Results:

• The XGBoost model with selected HRV features achieved 66.67% accuracy and 58.33% F1-score, significantly improving upon using all HRV features.
• HRV suppression correlated with high-workload phases (landing) and lower performance scores.
• HRV recovery and peak performance were observed during low-workload phases (cruise).

Conclusions:

• Selected HRV features combined with machine learning provide a reliable method for assessing pilot mental workload.
• This framework enables real-time monitoring and prediction of cognitive overload, enhancing flight safety during critical operations.