Perception and Action Assistance for the Remote Control of Robotic Manipulation
Teleoperation systems offer tremendous potential in extending the reach of human abilities in domains such as healthcare, where remote manipulation can improve both access and efficiency. However, designing intuitive, efficient, and low-effort teleoperation interfaces remains a significant challenge, particularly for complex, high-precision tasks such as those in nursing. This dissertation explores the design and evaluation of control and perception assistances to enhance teleoperation performance, reduce physical and cognitive workload, and improve operator preference across a range of remote manipulation scenarios.
Through systematic user studies involving healthcare-relevant tasks and representative user populations, we first identify motion mapping as an intuitive and high-performing control interface. However, its benefits are offset by increased operator effort due to limited precision. To address this, we develop and evaluate action assistance strategies—including the separation of orientation and position control and environment-based motion scaling—which significantly reduce task completion time and body motions, while improving control precision and ergonomics. These approaches are extended to bi-manual teleoperation, demonstrating that both position and orientation supports yield workload and performance improvements. Additionally, assist-as-needed paradigms based on operator state (e.g., motion intent or physical activity) are shown to further improve performance and reduce frustration, highlighting the importance of adaptive interface design.
Recognizing that assistance may confuse the operator’s understanding of robot autonomy, we further explore perception assistance through Augmented Reality visual cues to improve awareness and user experience. We show that the effectiveness and preference for these cues depend on the level of autonomy in the interface, and that training can shift user preferences to align with developer-recommended designs.
In conclusion, this dissertation offers design ideologies for building teleoperation systems tailored to healthcare contexts. These findings underscore the importance of adaptable, intuitive interfaces and highlight the value of workload- and intent-aware assistances. The work lays a foundation for future teleoperation systems that are accessible, efficient, and supportive of diverse healthcare professionals performing complex remote tasks.
Advisor: Jane Li, Robotics Engineering (WPI)
Committee: Jing Xiao, Head of Robotics Engineering (WPI)
Berk Calli, Robotics Engineering (WPI)
Carlo Pinciroli, Robotics Engineering (WPI)
Karen Troy, Associate Department Head, Biomedical Engineering, (WPI)
Zoom: https://wpi.zoom.us/my/achyuthan?pwd=T0hrMGVidzVHSnJkMmM2RGlvd2xBdz09