Active feedback helps drones fix and check tasks with better success
Fly-by-Code: Embodied Coding Agents for Aerial Manipulation with Active Visual and Physical Feedback
RoboticsComputer Vision and Pattern RecognitionMachine Learning
Summary
Drone robots often can’t see everything they need when carrying things and don’t always know why a task failed. The authors created a system where the drone can actively look from new angles and test parts of a task in simulation to gather missing clues. This extra checking helps the drone fix its plan and better know when a job is really done or not. When tested in a simulation and real world, this method made completing drone tasks much more reliable.
What this means in practice
- •For drone developers: Create drone programs that can diagnose and fix task errors using active simulated checks and new visual viewpoints for better real-world performance.$Commercial implications: This enables commercial drone software companies to sell smarter programming tools that improve complex aerial delivery and inspection services.
- •For robotics control engineers: Use active feedback techniques to improve assessment of task success and program updates during physically interactive robotic tasks with limited sensor views.
Authors
Jaewoo Lee, Jeongyeon Seo, Sihyun Cho, Gyeongrak Choe, Yutong Wang, Bavin Saravanan, Jia-Bin Huang, Furong Huang, Sebastian Scherer, Guanya Shi, H. Jin Kim, Seungjae Lee, Dongjae Lee
Abstract
Embodied coding agents construct robot policies from natural-language instructions by generating executable programs and refining them from execution feedback. This refinement assumes that execution feedback reveals whether the task succeeded and, after a failure, provides sufficient evidence to guide program revision. Aerial manipulation violates this assumption in two ways. Payload constraints restrict onboard camera coverage and leave task-critical outcomes outside the robot's view (partial observability), while strong dynamic coupling and limited control margins make physical interactions fragile (limited interaction robustness) and leave failure causes unresolved. Re-examining the same feedback cannot recover observations that were never collected or separate failure causes that the available evidence does not distinguish. To acquire this missing evidence, we introduce Active Feedback (AF), an active "View and Probe" framework. When execution feedback leaves unresolved questions, the agent synthesizes a separate diagnostic program that runs in simulation to inspect unseen outcomes from informative viewpoints (View) or test hypotheses about failure causes through targeted physical interactions (Probe). Across six simulated aerial manipulation tasks with 60 trials per condition, AF reduces false completion declarations from 25 to 5 and raises final task success from 31.7% to 75% compared with terminal feedback under the same coding-agent configuration and task-execution limit. 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.