Flight Crew Physiological Data for Cognitive State Monitoring

Physiological data was collected from pilot and copilot pairs during controlled experiments and a full flight simulation to train models for detecting cognitive states. The data captures three induced states: Channelized Attention, Diverted Attention, and Startle/Surprise. Sensor recordings were sampled at 256 Hz.

Use Cases

• Train machine learning models to classify Channelized Attention, Diverted Attention, and Startle/Surprise states from 256 Hz physiological sensor data.
• Analyze physiological signal patterns between baseline and induced cognitive states like Diverted Attention during monitoring tasks.
• Benchmark model performance by training on controlled single-state experiments and testing on the full LOFT flight simulation data.
• Investigate pilot-copilot pair dynamics by correlating physiological responses within shared crew IDs across different cognitive loads.

Strengths

• Data includes a structured test set from a full Line Oriented Flight Training simulation with take-off, flight, and landing phases.
• Sensor data was collected at a high sample rate of 256 Hz for detailed physiological analysis.
• Experiments were designed to induce three distinct, labeled cognitive states (CA, DA, SS) for supervised learning.

Limitations

• The dataset contains physiological noise and artifacts, as acknowledged for data from real people.
• The training set consists of only three controlled experiments, which may limit the diversity of scenarios for model training.
• Specific details on the number of pilot pairs, recording duration, and total data volume are not provided.

Provenance

Source

National Aeronautics and Space Administration, Langley Research Center

Collection Method

Collected from pilot/copilot pairs in and out of a flight simulator during controlled experiments and a full flight simulation.

Freshness

Last updated in January 2026.