Session: 16-01-01: Government Agency Student Poster Competition

Paper Number: 149674

149674 - The Impact of Robotic Slip Reflex Timing on Performance and User Experience for Body-Powered Prostheses 

Humans are capable of rapidly responding to grasp disturbances through unconscious low-level reflexes governed by our nervous system. When an object begins to slip out of our hands, local tactile feedback initiates muscle activity to increase grasp strength. These reflexes are impeded for upper-limb prosthesis users due to a lack of tactile feedback. As a result, researchers have developed slip reflexes for myoelectric prostheses, which are controlled through surface electromyography (sEMG) signals. Body-powered prostheses, a popular alternative to their myoelectric counterparts due to their inherent force and position feedback, typically rely on movement of the contralateral shoulder to open and close the hand. However, the implementation of robotic reflexes in body-powered prostheses remains unexplored. The addition of such reflexes is promising for grasp performance, but studies show that robotic assistance reduces the user’s sense of agency over a system and can lead to demotivation, dissatisfaction, and poor system acceptance.

In this study, we seek to improve slip reflex performance in body-powered prostheses without reducing user agency by tuning the delay of robotic assistance relative to the onset of object slip. We propose a robotic slip reflex design and test it with nine participants using a custom body-powered prosthesis emulator and virtual environment. Participants perform a simple slip reflex task 20 times for eight different pseudorandomly ordered conditions. The conditions include seven assistance conditions with delays ranging from 0 to 600 milliseconds, and one unassisted control condition. We assess task performance through object slip distance in each trial. After each condition, participants respond to a set of survey questions related to their sense of agency and perceptions of assistance, task performance, and robot cooperativeness. We analyze all data using mixed model regression, treating delay as an ordinal variable to enable pairwise comparisons.

Results show that increasing assistance delay generally results in a decrease in performance, perceived performance, and perceived assistance and an increase in agency and perceived robot cooperation. However, when the delay of assistance coincides with the average measured participant reaction time (~400 milliseconds), participants stop the slipping object quicker than when unassisted, without perceiving a decrease in agency—even though they perceive the presence of assistance and an increase in performance. These findings suggest that for physical human-robot interactions where users are aware of the robot’s assistive movements, such as in body-powered prostheses or exoskeletons, temporal alignment between the user and robot plays a significant role in task performance and user experience.

Presenting Author: Benjamin Davis University of California, Berkeley

Presenting Author Biography: Benjamin completed his undergraduate degree in Mechanical Engineering at The University of Texas at Austin in 2023 before starting his PhD in Mechanical Engineering at the University of California, Berkeley the same year. He is advised by Professor Hannah Stuart as a member of the Embodied Dexterity Group (EDG) . His interest lies in physical Human-Robot Interaction (pHRI), particularly regarding how humans perceive and adapt to varying levels of robot autonomy through their physical interface.

Authors:

Benjamin Davis University of California, Berkeley
Michael Abbott University of California, Berkeley
Hannah Stuart University of California, Berkeley