GLIDE
GLIDE illustrates guardrail refinement from guarded teleoperation through the wine-serving task. In this task, bimanual teleoperation moves the bottle and glass independently, creating several failure points: glass drift and uncontrolled bottle tilt during the pour, bottle–glass misalignment, and end-effector motions that destabilize either grasp.
To prevent these failures, GLIDE limits hand motion to preserve the glass grasp, amplifies wrist rotation for controlled pouring, and maintains bottle–glass alignment by bounding their relative position. This lets the operator focus on timing the pour instead of constantly correcting the grasp. Appendix B provides the full guardrail details for all three tasks.
Initialization is described in Section 3.1, trajectory-based self-refinement in Section 3.2, and deployment for guarded data collection and policy deployment in Section 3.3.
Conclusion and Limitations
The results are encouraging but leave several opportunities for future work. First, the runtime guardrails use only proprioception to filter commands, although the coding agent uses recorded video and numerical trajectories for offline diagnosis. Adding visual and tactile signals at runtime could cover slipping grasps, object motion, and liquid state.
Second, guardrails already condition on robot state, but uncertainty or task-phase estimates could improve intervention timing. Future studies can extend GLIDE to long-horizon tasks, mobile manipulation, and more complex robot systems.
Frequently Asked Questions
What task demonstrates GLIDE’s guardrail refinement? The wine-serving task demonstrates guardrail refinement from guarded teleoperation.
What failure points occur during the wine-serving task? Failure points include glass drift, uncontrolled bottle tilt, bottle–glass misalignment, and end-effector motions that destabilize either grasp.
How does GLIDE prevent these failures? GLIDE limits hand motion, amplifies wrist rotation, and bounds the relative position of the bottle and glass.
What are important directions for future work? Future work includes adding visual and tactile runtime signals, improving intervention timing, and extending GLIDE to longer-horizon and mobile manipulation tasks.