Managing human cognitive load, especially in teams, is important for ensuring high task performance. We hypothesize that robots which can monitor human cognitive load in real-time in mixed initiative teams and adapt their interactions based on it will lead to better team performance compared to robots that are unaware of human cognitive load. In this paper, we introduce an online cognitive workload detection algorithm based on changes in human pupil size that is able to track human cognitive workload during task performance. The algorithm is integrated into a cognitive robotic architecture to allow the system to adapt its behavior based on inferred workload. The system is evaluated in a mixed initiative human-robot team experiment where two humans and two autonomous robots had to collaborate in order to achieve high performance in a complex dual-task setting. The results confirm the operation of the proposed integrated system and demonstrate that online adaptation of autonomous robot behavior to human cognitive load can lead to better team performance compared to non-adaptive robots.
@inproceedings{aygunetal26icsr,
title={Online Human Workload Detection for Behavior Adaptation in Autonomous Robot Teammates},
author={Ayca Aygun and Helena Fu and Evan Krause and Matthias Scheutz},
year={2026},
booktitle={Proceedings of the Annual Meeting of the Cognitive Science Society},
url={https://hrilab.tufts.edu/publications/aygunetal26icsr.pdf}
}