Fatigue and Performance Modeling of Sleep-Deprived Soldiers

Soldiers with chronic sleep restriction have reduced cognitive performance that can result in dangerous and deadly behavior. Current mathematical models used to predict performance are based on the Two-Process sleep regulation and have been well-validated for single episodes of sleep deprivation, but not for chronic sleep restriction and recovery. An innovative new modeling approach will be developed based on hybrid neural networks, where the existing circadian pacemaker model will be retained and combined with a neural network mapping of the fatigue and performance metrics. The model will first be trained to duplicate existing Two-Process models, and then will be altered as needed and training directly from experimental data. A technical issue that will be resolved is how to include sleep and wake protocols in the model structure and training. An experiment will be designed in Phase I to increase the database for chronic sleep restriction studies, using tasks relevant to Army personnel. The experiment will be conducted in Phase II. The new models when validated can be used to determine unit readiness and to schedule sleep regimens. Real time versions combined with physiological measurements can potentially be used to adapt to and monitor individual soldiers.