A Comparative Study of a Shared-Autonomous Control System versus a Manual Control Interface: Task Performance and Usability in the Control of Dexterous Prosthetic Hands

In human–robot interaction, both robotic performance and usability must be achieved. However, interacting with complex systems with many degrees of freedom often leads to cumbersome and unintuitive control. Autonomy has been introduced to address these challenges, but it can reduce the user’s sense of agency—an essential factor for robotic systems that aim to function as seamless extensions of the human body. In this study, we investigated how autonomy influences task performance, usability, and user experience in controlling a dexterous prosthetic hand, compared with a conventional manual interface. For this comparison, we implemented a shared-autonomous control interface that enabled seamless use of versatile grasp postures, and a manual interface requiring explicit posture selection while preserving a high sense of agency. Ten able-bodied participants performed pick-and-place tasks with both interfaces. With shared-autonomy, tasks were completed 26.3% faster, with less hesitation. Although manual control was initially rated higher for learnability, participants increasingly preferred shared-autonomy over time, reporting reduced cumbersomeness and smoother operation. User confidence strongly correlated with performance, usability, and perceived seamlessness. Our results suggest that shared-autonomy can improve interaction with complex robotic systems while supporting sense-of-agency-relevant indicators, including confidence and operational seamlessness, especially as users gain confidence in the system.