Cognitive Load Estimation through Eye-Tracking in Industrial Tasks

Abstract

Understanding and managing cognitive workload is critical for safety and efficiency in industrial settings, yet practical measurement techniques are limited on the factory floor. We present a vision-based approach to estimate cognitive load using eye-tracking, hand tracking, and object detection, avoiding intrusive sensors like EEG. We evaluated this method in a real-world assembly task performed by an expert under two conditions: an organized workspace and a disorganized workspace. A mobile eye-tracker recorded the worker’s gaze, while computer vision detected hands and tools in use. The disorganized condition elicited higher visual workload, evidenced by more frequent fixations and saccades, broader gaze dispersion, and more attention to irrelevant areas, despite little change in physiological proxies such as pupil size and blink rate. These results demonstrate that our non-intrusive, vision-only system can distinguish cognitive workload differences in an industrial task, laying the groundwork for in-situ workload monitoring without requiring cumbersome biosensors.