Classifying Subjective Time Perception in a Multi-robot Control Scenario Using Eye-tracking Information

TL;DR

The paper tackles how to infer an operator's subjective time perception from eye-tracking data during multi-robot swarm control to enable online, closed-loop adaptation in ChronoPilot. It combines a robot-swarm experiment with 21 participants, 26 eye-tracking features, and two time-perception labels (duration estimation and PPOT), using Naive AutoML to automatically identify high-performing pipelines across multiple time-window sizes $t_w$. Results show high unseen-test accuracies, with better performance for shorter $t_w$ and robustness to swarm size and trial duration; personalized fine-tuning from a brief setup phase yields substantial improvements. This approach supports real-time operator-state feedback for adaptive robotic control and suggests avenues for broader multimodal, human-aware automation in large-scale multi-robot systems.

Abstract

As automation and mobile robotics reshape work environments, rising expectations for productivity increase cognitive demands on human operators, leading to potential stress and cognitive overload. Accurately assessing an operator's mental state is critical for maintaining performance and well-being. We use subjective time perception, which can be altered by stress and cognitive load, as a sensitive, low-latency indicator of well-being and cognitive strain. Distortions in time perception can affect decision-making, reaction times, and overall task effectiveness, making it a valuable metric for adaptive human-swarm interaction systems. We study how human physiological signals can be used to estimate a person's subjective time perception in a human-swarm interaction scenario as example. A human operator needs to guide and control a swarm of small mobile robots. We obtain eye-tracking data that is classified for subjective time perception based on questionnaire data. Our results show that we successfully estimate a person's time perception from eye-tracking data. The approach can profit from individual-based pretraining using only 30 seconds of data. In future work, we aim for robots that respond to human operator needs by automatically classifying physiological data in a closed control loop.

Figures (5)

• Figure 1: POV of participant during the robot experiment.
• Figure 2: Distribution of duration estimation labels (top: 2 classes; bottom: 3 classes), blue: questionnaire answers, red: thresholded labels black dashed lines: classification threshold.
• Figure 3: Distribution of the subjective PPOT labels (top: 2 classes; bottom: 3 classes), blue: questionnaire answers, red: thresholded labels, black dashed lines: classification threshold.
• Figure 4: Classification accuracies using the binary duration estimate label for experiments split by actively moving robots. The blue lines indicate the accuracy of the classifier for different $t_w$ while the red line indicates the majority class of the subdatasets. The shaded areas represent the standard deviation (over 100 independent repetitions).
• Figure 5: Classification accuracies using the binary duration estimate label for experiments split by experiment duration. The blue lines indicate the accuracy of the classifier for different $t_w$ while the red line indicates the majority class of the subdatasets. The shaded areas represent the standard deviation (over 100 independent repetitions).