Cognitive Workload Assessment of UAV Operators Using Synergy of Machine Learning and Reasoning

Detecting and predicting the UAV operator’s cognitive overload is a critical part of the mission effectiveness. Despite impressive achievements in neurocognitive science, cognitive workload assessment and prediction in real-time remain unsolved problems. In this work, we investigate an ensemble of machine learning and machine reasoning approaches towards this solution. The synergy of these two AI areas enables the assessment of operator overload. We conduct an experiment using several machine learning techniques to detect levels of cognitive workload using directly measured physiological, neurophysiological, and environmental features. We use these results to build risk assessment models based on probabilistic causal graphs. Using the data distributions, we infer and assess the risk of operator overload and overall performance. We provide a cognitive workload risk monitoring and mitigation scheme based on the proposed learning and reasoning paradigm.