Towards Accessible Robot Control: Comparing Kinesthetic Teaching, SpaceMouse Teleoperation, and a Mixed Reality Interface

TL;DR

This study29 investigates how non-expert users control high-DOF robots using three interfaces—SpaceMouse teleoperation, a Mixed Reality interface, and kinesthetic teaching—across two 7-DOF manipulators and six real-world tasks. By combining objective measures (time, success) with subjective workload and usability assessments, the authors quantify a performance gap between teleoperation and direct manipulation, and reveal task-dependent differences in how each interface supports doing, learning, and understanding. Kinesthetic teaching offers the best objective performance but at greater physical cost, while SpaceMouse and MR provide comparable results with distinct strengths: SpaceMouse for direct, precise motion and MR for improved spatial understanding and reduced physical effort. The findings underscore the importance of user-centered evaluation in non-expert contexts and point to design directions such as adaptable interfaces and richer 3D visualization to close the usability-performance gap for everyday robot operation.

Abstract

Teleoperation interfaces are essential tools for enabling human control of robotic systems. Although a wide range of interfaces has been developed, a persistent gap remains between the level of performance humans can achieve through these interfaces and the capabilities afforded by direct human-guided robot control. This gap is further exacerbated when users are inexperienced or unfamiliar with the robotic platform or control interface. In this work, we aim to better characterize this performance gap for non-expert users by comparing two teleoperation approaches, SpaceMouse teleoperation and a Mixed Reality (MR) interface, against kinesthetic teaching as a non-teleoperation baseline. All three approaches were evaluated in a comprehensive user study involving two robotic platforms and six complex manipulation tasks. Quantitative results show that the SpaceMouse and MR interfaces performed comparably, with significant differences in task completion observed for only two tasks, and success rates declining as task complexity increased. Qualitative analysis reflected these trends, highlighting differences in Physical Demand and identifying interface attributes that influence users' ability to perform, learn, and understand. This study quantifies the limitations of current teleoperation methods and incorporates subjective feedback from 25 participants. The results highlight the critical need to design and rigorously evaluate teleoperation systems for non-expert users, particularly in contexts where autonomous robots are deployed in personal or everyday environments, to ensure usability, efficiency, and accessibility.

Figures (12)

• Figure 1: The three control methods investigated in this study. (Left: Kinesthetic Teaching, Center: SpaceMouse Teleoperation, Right: Teleoperation Interface.)
• Figure 2: The mixed reference frame paradigm (left), where SpaceMouse translation inputs are defined in the base (B) frame and rotation inputs in the tool (T) frame, and the base reference frame paradigm (right), where both translation and rotation inputs are defined relative to the base (B) frame. We used the mixed reference frame for the Kinova robot and the base reference frame for the xArm robot.
• Figure 3: The system architecture.
• Figure 4: Participants use far interactions (indicated by the white ray coming from the person's hand) to manipulate a virtual sphere and gripper. The physical robot moves in real time to match the end-effector pose defined by the virtual objects.
• Figure 5: Prior experience across four categories for participants in Groups 1 and 2. Group 1 completed four short-horizon tasks with the Kinova Gen3, while Group 2 completed two long-horizon tasks with the UFactory xARM 7.