Monitoring pilots' mental workload in real flight conditions using multinomial logistic regression with a ridge...

Area of Science:

• Cognitive neuroscience
• Aerospace engineering
• Machine learning

Background:

• Pilot mental workload is critical for aviation safety, as fluctuations can lead to errors.
• Monitoring mental workload is essential for preventing aviation accidents.

Purpose of the Study:

• To develop and validate a machine learning model for accurately monitoring pilot mental workload.
• To improve upon previous methods for pilot workload detection using advanced algorithms.

Main Methods:

• Electroencephalogram (EEG) data were collected from pilots under Visual Flight Rules (VFR) conditions.
• Data cleaning involved the Riemannian artifact subspace reconstruction (rASR) filter.
• Feature selection used an information gain (IG) attribute evaluator, identifying 25 optimal features.
• Fifteen classifiers were evaluated, with multinomial logistic regression with a ridge estimator selected for its performance.

Main Results:

• The selected model achieved a mean accuracy of 84.6% on a dataset of 17 subjects.
• Multinomial logistic regression with a ridge estimator demonstrated significant classification accuracy (p < 0.05).
• The model effectively detected pilot mental workload in real flight scenarios.

Conclusions:

• Advanced machine learning, specifically multinomial logistic regression with a ridge estimator, can effectively detect pilot mental workload.
• This approach offers improved accuracy for monitoring cognitive states in aviation.
• Further research should address limitations such as controlling environmental variables and workload stationarity.