Static Is Not Enough: A Comparative Study of VR and SpaceMouse in Static and Dynamic Teleoperation Tasks

TL;DR

This study addresses how teleoperation interface choice affects the quality of demonstrations for imitation learning by comparing a VR controller and a SpaceMouse across static and dynamic tasks in a within-subjects design. It introduces an open-source VR teleoperation interface and shows that VR yields higher success rates, faster task completion, and lower workload, particularly in dynamic tasks where dynamics like impulse and momentum matter. The work highlights the limitations of velocity-based inputs for dynamic manipulation and provides practical insights for data collection in real-time robot control, with implications for constructing high-quality dynamic-task datasets. Overall, the findings advocate for pose-based VR control in demonstration collection and offer an open-source tool to facilitate broader research and application in dynamic teleoperation scenarios.

Abstract

Imitation learning relies on high-quality demonstrations, and teleoperation is a primary way to collect them, making teleoperation interface choice crucial for the data. Prior work mainly focused on static tasks, i.e., discrete, segmented motions, yet demonstrations also include dynamic tasks requiring reactive control. As dynamic tasks impose fundamentally different interface demands, insights from static-task evaluations cannot generalize. To address this gap, we conduct a within-subjects study comparing a VR controller and a SpaceMouse across two static and two dynamic tasks ($N=25$). We assess success rate, task duration, cumulative success, alongside NASA-TLX, SUS, and open-ended feedback. Results show statistically significant advantages for VR: higher success rates, particularly on dynamic tasks, shorter successful execution times across tasks, and earlier successes across attempts, with significantly lower workload and higher usability. As existing VR teleoperation systems are rarely open-source or suited for dynamic tasks, we release our VR interface to fill this gap.

Figures (5)

• Figure 1: Overview of the study design, including the two teleoperation interfaces (VR, SpaceMouse), four randomized tasks (two static and two dynamic), and evaluated measures of user task performance and user experience.
• Figure 2: Workflow of our VR teleoperation interface. Controller pose is tracked and mapped to incremental end-effector commands, with button inputs handling gripper control and calibration.
• Figure 3: 4 Tasks used in our study, including success criteria. T1 and T2 are static tasks. T3 and T4 are dynamic tasks requiring continuous motion.
• Figure 4: User task performance across four tasks. A: Average success rate for VR and SpaceMouse across tasks. B: Average completion time across tasks under both interfaces. C: Cumulative success rate over the five allowed attempts for each task. VR consistently achieves higher success rates, yields significantly shorter completion times, and reaches successful completion earlier across static and dynamic tasks. Significance levels are indicated as: $*\,p < .05,\ **\,p < .01,\ ***\,p < .001$.
• Figure 5: User subjective feedback. A: SUS scores. B: NASA-TLX scores. C: NASA-TLX scores for each individual dimension. VR shows higher usability and significantly lower workload across both static and dynamic tasks.