Using Context to Optimize a Functional State Estimation Engine in Unmanned Aircraft System Operations

As UAS operations continue to expand, the ability to monitor real-time cognitive states of human operators would be a valuable asset. Although great strides have been made toward this capability using physiological signals, the inherent noisiness of these data hinders its readiness for operational deployment. We theorize the addition of contextual data alongside physiological signals could improve the accuracy of cognitive state classifiers. In this paper, we review a cognitive workload model development effort conducted in a simulated UAS task environment at the Air Force Research Laboratory (AFRL). Real-time workload model classifiers were trained using three levels of physiological data inputs both with and without context added. Following the evaluation of each classifier using four model evaluation metrics, we conclude that by adding contextual data to physiological-based models, we improved the ability to reliably measure real-time cognitive workload in our UAS operations test case.