Related Experiment Video
Updated: Aug 10, 2025
12:29
Coherence between Brain Cortical Function and Neurocognitive Performance during Changed Gravity Conditions
Published on: May 23, 2011
19.6K
Using machine learning methods and EEG to discriminate aircraft pilot cognitive workload during flight
Hamed Taheri Gorji1, Nicholas Wilson2, Jessica VanBree3
1Biomedical Engineering Program, University of North Dakota, Grand Forks, ND, USA. hamed.taherigorji@und.edu.
Scientific Reports
|February 13, 2023
Summary
This study used electroencephalogram (EEG) data to accurately measure pilot cognitive workload during flight. Machine learning models identified workload states, improving aviation safety and aircraft system design.
Area of Science:
- Aviation Psychology
- Neuroscience
- Machine Learning
Background:
- Pilot cognitive workload fluctuates during flight operations, potentially increasing error risk.
- Understanding and quantifying cognitive workload is crucial for aviation safety.
- Existing methods for assessing pilot workload may not fully capture dynamic changes.
Purpose of the Study:
- To investigate the use of electroencephalogram (EEG) for real-time assessment of pilot cognitive workload.
- To develop and validate a machine learning model capable of discriminating between low, medium, and high cognitive workload states.
- To explore the potential of EEG-based workload monitoring to enhance aircraft system design and flight automation.
Main Methods:
- Collected EEG data from ten collegiate aviation students during live-flight operations in a single-engine aircraft.
- Analyzed EEG signals using power spectral density (PSD) and log energy entropy across delta, theta, alpha, and beta sub-bands.
- Employed recursive feature elimination (RFE) and LassoCV for feature selection, followed by a stacking ensemble machine learning algorithm (SVM, Random Forest, Logistic Regression) with hyperparameter optimization and cross-validation.
Main Results:
- Identified 15 key features indicative of pilot cognitive workload states.
- The RFE-selected features achieved high performance in the stacking ensemble model: 91.67% accuracy, 93.89% precision, 91.67% recall, 91.22% F-score, and 0.93 ROC-AUC.
- Demonstrated the effectiveness of combining PSD and log energy entropy with machine learning for workload discrimination.
Conclusions:
- EEG, analyzed with advanced machine learning techniques, shows significant potential for accurately discriminating pilot cognitive workload levels.
- The findings support the integration of EEG-based workload monitoring into aircraft systems to improve automation and enhance flight safety.
- This research provides a foundation for developing intelligent systems that adapt to pilot cognitive states, mitigating risks associated with workload fluctuations.
