Evaluating Vision-guided Shared Control in Assistive Robotic Manipulators for Individuals with Upper Limb Impairments

Research Objectives: To assess the performance efficiency and user acceptance of vision-guided shared (VGS) control compared to manual teleoperation in assistive robotic manipulators among power wheelchair users with upper limb impairments.

Design: Mixed methods

Setting: Controlled laboratory environment.

Participants: Twelve adults were recruited, with an average age of 52.42±18.86 years, 9 men and 3 women; 1 Asian American and 12 Caucasian adults; 4 with spinal cord injury, 3 with cerebral palsy, 2 with muscular dystrophy, 1 with post-polio syndrome, 1 with chronic inflammatory demyelinating polyneuropathy, and 1 with spinal stenosis.

Interventions: Each participant completed two multi-step kitchen tasks: a drinking task (6 steps) and a popcorn-making task (10 steps). For each task, participants operated two trials of both manual and VGS control. VGS is a discrete user-initiated switch where intelligent software assists automatically during the fine-manipulation portions of the tasks.

Main Outcome Measures: Task performance measures were recorded for times and success rate. After the control mode for each task, the System Usability Scale (SUS) and NASA Task Load Index (NASA-TLX) were rated to assess usability and workload. At study end, participants completed a semi-structured interview providing feedback on the control system and robotic arm usage. We analyzed SUS and NASA-TLX using the Wilcoxon signed-rank test with a 0.05 significance.

Results: There were no significant differences in SUS scores for both tasks, but NASA-TLX revealed significant differences between control modes. Manual operation times were 593.56±227.23 s for the drinking task and 1016.22±391.72 s for the popcorn task, compared to 291.75±85.94 s and 641.80±177.79 s with VGS control. Success rates for the manual control were 73.91%-100% for the drinking task and 56.52%-100% for the popcorn task. For VGS control, the success rates were 95.65%-100% and 79.17%-100%, respectively. The interview revealed 5 participants preferred manual control, 5 preferred VGS control and 2 had no preference. All participants think the robotic arm is useful for helping them reach and grasp objects and in daily living activities such as making coffee and food preparation. The main operation challenges were understanding the wrist orientation mode and depth perception.

Conclusions: VGS control reduced operation workload and can greatly reduce the task completion time, but not all users preferred it. Interview feedback suggested improving issues with depth perception and providing additional training with the wrist rotation mode. Further validation with a larger sample is needed.

Author(s) Disclosures: No conflict of interest.