TL;DR
This study explores how power wheelchair users prefer to interact with robotic arms, finding that while autonomous features are appreciated, users strongly prefer maintaining manual control via joystick for a sense of independence and accomplishment. The research highlights the need for future assistive systems to offer granular situational feedback and collaborative control, rather than full automation.
What Problem Does This Paper Address?
As robotic arms become more integrated into assistive technologies for individuals with mobility impairments, understanding user preferences for control is crucial. The challenge lies in balancing the potential benefits of autonomous functions (like increased efficiency and reduced cognitive load) with the user's desire for agency and control. Previous research has shown that full software autonomy isn't always preferred, but there's a need for deeper insight into how users want to share control and what specific features contribute to a positive user experience. This paper specifically investigates control authority preferences among potential future users of Assistive Robotic Manipulators (ARMs) for power wheelchair users.
Methods
The study employed a mixed-methods approach involving 10 power wheelchair users with no prior robotic arm experience. Participants engaged in a kitchen task (popcorn making) across three scenarios:
- User-Initiated Autonomous Software (AutonSof): Participants used a shared control system where predefined autonomous actions could be triggered via a touchscreen.
- Manual Joystick Control: Participants performed the task solely using a joystick.
- Interactive Assistance Scenario: Participants could request assistance or switch between manual and autonomous control at any point, with an investigator acting as the software.
Data collection included task performance metrics (completion time, mode switches, success rates), standardized questionnaires (NASA-TLX for workload, SUS for usability), and semi-structured exit interviews. Thematic analysis was used to interpret qualitative data from interviews.
Datasets
Not discussed in the paper. The study used a custom popcorn-making task and existing questionnaires.
Key Insights and Ideas
- The "Driver" Mentality: Users strongly prefer to feel like they are "driving" the system at all times, even when autonomous features are available. This sense of agency is paramount.
- Independence and Accomplishment: Manual joystick control, despite potentially higher workload, fosters a greater sense of independence and accomplishment, which users highly value.
- Autonomy as a Backup, Not a Replacement: While impressed by autonomous capabilities, users see them more as a helpful backup for when they are fatigued or stuck, rather than a primary mode of operation.
- Granular Feedback is Key: Users desire more detailed, real-time feedback about the robot's state and environment (e.g., "How close is the gripper?", "Is the lid centered?"). This situational awareness is crucial for effective collaboration.
- Learning by Doing: Participants expressed a preference for learning and understanding how the system works through manual control, reinforcing their desire to be in charge.
Key Results
- Workload: Manual control resulted in significantly higher mental demand compared to AutonSof, likely due to frequent mode switching.
- Usability: Both AutonSof and manual control achieved good System Usability Scores (SUS), indicating generally positive usability.
- Task Performance: Manual control generally took longer and involved more mode switches than AutonSof. However, the "grasp lid" and "place lid" steps were challenging in both modes.
- Preference: Seven out of ten participants preferred manual joystick control, while all had a positive impression of autonomous capabilities. Three participants preferred the interactive assistance scenario.
- Assistance Requests: Participants sought assistance for various reasons, including visual obstruction, fatigue, and the need for precision. They also requested situational feedback more than direct task takeover.
Limitations and Future Work
- Population Bias: Participants were recruited from the National Veteran Wheelchair Games, an event that may attract more outgoing and socially engaged individuals with potentially better upper-limb function than the broader population of ARM users.
- Limited Study Time: The constrained schedule of the NVWG limited the number of trials and may not fully reflect long-term use and adaptation.
- Study Order Effect: The order of testing (AutonSof first, then manual) might have influenced preferences. Future studies could counterbalance this.
- Future Work: The authors suggest further research into:
- Longer-term studies to understand how novelty wears off and the true benefits of autonomy emerge.
- Developing interactive feedback systems that adapt to user ability and style.
- Implementing variable granularity in action selection and adaptive state feedback.
- Designing modular systems that can seamlessly integrate interactive components with user input to maintain the user's sense of control.
- Exploring how to effectively convey information to and from the system in a way that meets diverse user needs.
Summary updated