Human learning assessment in teleoperated manipulation using a UR3 robotic arm

Le résumé fourni par la source

This study investigated human learning behaviour during teleoperated manipulation using a Universal Robots UR3 collaborative robotic arm. A repeatable experimental framework was developed in which twelve participants performed an identical object manipulation task over five consecutive trials. Performance was evaluated using objective metrics including task completion time, command count, warning count, and a normalized Final Performance Score (FPS).The experimental results demonstrated a clear improvement in participant performance with repeated practice. The average FPS increased from 0.713 in the first trial to 0.875 in the fifth trial, indicating that participants gradually became more proficient in controlling the robot. Although a temporary reduction in performance was observed during the fourth trial, the strategy analysis revealed that this decline corresponded to a period of exploration in which many participants adopted alternative command execution patterns. Once more effective strategies were identified, performance improved significantly in the final trial.A logistic learning model was successfully employed to represent the learning progression, showing that participant performance followed a gradual improvement pattern rather than a simple linear trend. Furthermore, the proposed Human Cognitive Factor (HCF) incorporated participant fatigue into the evaluation process, providing a more comprehensive assessment of teleoperation performance by considering both learning and cognitive state.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.