Identifying the impact of robot speed and task time on human-robot collaboration through facial feature analysis

Abstract

The increasing involvement of collaborative robots (cobots) has led human workers to perform more value-added tasks, but with greater mental demands. In order to properly design tasks that protect worker long-term health, it is important to be able to detect and quantify stressors (robot speed and task time) that increase mental workload in human-robot collaboration (HRC). In this work, HRC task conditions (robot speed and task time) were classified based on changes in facial features, a non-intrusive stress indicator that has rarely been investigated for HRC. Twenty participants performed an assembly task in a seated posture under both high and low robot speeds, and for a prolonged duration. The results showed stress level and mental workload were higher at high robot speed compared to low speed. For the high-speed setting, a higher stress level was observed at the end of task compared to the beginning. For task classification, a random forest model was able to classify task conditions for robot speed and task time with accuracies greater than 97%. The lip corner movement was the primary facial feature change across classification tasks. These results support the use of facial feature changes to detect worker response to stressful conditions in HRC.