Ghost Arm: Aligning Human and Robot Kinematics through AR Overlays in MoCap-Based Teleoperation of Robot Arm

Teleoperating a robot arm typically requires either positioning its end-effector directly or programming joint movements. The former limits posture and trajectory control, while the latter lacks intuitive, real-time feedback. Unlike controlling their own arms through integrated visual and proprioceptive cues, humans face difficulties with robot arms due to mismatched orientations and joint kinematic configuration. In this work, we investigate how different augmented reality (AR) visualisations can support teleoperation via motion capture (MoCap). Across two user studies (N = 24, N = 24), we evaluate how overlaying a virtual human arm alongside the robot arm affects user performance and experience. Our results show that the humanoid AR overlay facilitates learning the control mapping, but its benefits diminish once users become familiar with the task. We contribute empirical evidence and design insights for AR- and MoCap-based approaches to more effective robot arm teleoperation.