Towards automated hesitation detection during support-system enhanced industrial assembly

Abstract

Modern factories have to accommodate high flexibility, extreme customization and short product life cycles in a cost-effective way. This requires that the operators are provided with sufficient system support to aid their decision-making and hesitation. To investigate how the level of support can affect the operators’ behavior, the current study immersed 27 participants in a virtual reality factory where they were asked to complete three different assemblies with varying levels of system support (low, medium, or high). The support was provided by a collaborative robot (cobot). The participants’ experience was measured via a subjective marker of difficulty and an objective eye-tracking feature (gaze switches). The results showed that when the level of cobot support was low, participants found the assembly step more difficult and were gazing at the instruction screen more often compared to the medium and high support conditions. This suggests that the number of times operators look back at the instruction screen during a step could be a promising marker to automatically detect hesitation behavior in instruction-based assemblies. This study, therefore, presents the initial effort toward validating a behavioral marker of hesitation within this context.