Assessment and Prediction of Remote Operators’ Mental Workload Through AoI Data in AV Scenarios

With the advancement of autonomous vehicles (AV), remote operation is pivotal in ensuring safe and efficient transportation, especially in partially automated systems. This research explores how mental workload influences ocular behavior in remote operators using a fleet management interface. Eye-tracking data from a user study manipulating task difficulty and frequency are analyzed using the Area-of-Interest (AoI) technique. Workload significantly influences AoI metrics, including fixation duration, fixation frequency, time-to-first fixation, visit frequency, dwell time, and stationary entropy. Difficulty and frequency independently affect visual strategies, with difficulty exhibiting broader influences across metrics. Predictive models obtained higher f1-scores than the majority class baseline. Binary workload (easy-slow vs. hard-fast) and frequency (slow vs. fast) states prediction achieved both above 80%, while difficulty (easy vs. hard) reached up to 72%. A four-class classification yielded 48% f1-score. These findings support the use of AoI metrics for workload assessment and contribute to the design of adaptive interfaces responsive to operator mental states.