Fly-by-Code: Embodied Coding Agents for Aerial Manipulation with Active Visual and Physical Feedback
Embodied coding agents construct robot policies from natural-language instructions by generating executable programs and refining them from execution feedback.
ProofPaper ↗
Key points
- Aerial manipulation violates this assumption in two ways.
- To acquire this missing evidence, we introduce Active Feedback (AF), an active "View and Probe" framework.
- Using shared robot APIs, task programs refined with AF in simulation run on a physical aerial manipulator without task-program source-code modification and complete multi-stage manipulation tasks in the real world.
- These results show that executable diagnosis fills gaps in execution feedback for embodied coding agents, improving both completion assessment and program refinement in aerial manipulation.
Sources (1)
- [1]Fly-by-Code: Embodied Coding Agents for Aerial Manipulation with Active Visual and Physical FeedbackarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 8, 06:00 PM
Embodied coding agents construct robot policies from natural-language instructions by generating executable programs and refining them from execution feedback.
Aerial manipulation violates this assumption in two ways.
Extractive summary: sentences quoted from the sources.