Högskolan i Skövde

Laborious and repetitive tasks in industrial assembly stations have become one of the major causes of non-communicable diseases among industrial workers. As a solution, many companies have decided to automate these stations and replace the human worker with robots, often creating a sense of insecurity among workers and society.

Nevertheless, robots lack the ability and creativity to adapt dynamically in changing environments, a quality well-developed in humans. Thus, collaborative robots arise as an innovative solution to combine humans’ and robots’ strengths while maintaining a good production pace and a safe environment.

Within this framework, this study aims to enhance safety and efficiency in human-robot collaboration environments and avoid traditional fixed robot programming. To achieve this, proactive eye-gaze, a behavioural pattern where eye fixations precede actions, shows a promising potential as a predictive cue for human actions to control the robot movements, which, in combination with an ABB GoFa collaborative robot and the Pupil Labs Neon eye-tracking glasses, a replica of an electric vehicle car bumper assembly station has been constructed.

The developed system has successfully integrated robot dynamic motion, while maintaining safety measures, and a YOLOv11 model able to detect several targets in the assembly station. The result of this approach has been studied in a survey with 17 volunteers, proving the reliability of the system for proactive eye-gaze-based tasks. Besides, a total system latency of 500 ms has been measured, being well below the mean proactive eye-gaze interval of 795 ms, laying down the foundation for future research based on human-robot collaboration dynamic environments.