MessyNav: Zero-Shot Navigation via Manipulation in Messy Environments
MessyNav: Zero-Shot Navigation via Manipulation in Messy Environments
This research presents MessyNav, a zero-shot framework based on vision-language models (VLMs) that enables mobile manipulators to navigate cluttered real-world environments by performing interactive manipulation. The system uses explicit VLM-based affordance reasoning to decide whether to swipe, pick up, or avoid objects before task planning. Evaluated on a mobile manipulator in both indoor and outdoor settings, MessyNav improves the success rate by 27 percentage points over the strongest baseline in real-robot trials and shows statistically significant gains in offline evaluations without requiring task-specific training.
The study introduces MessyNav, a zero-shot navigation framework utilizing vision-language models for affordance reasoning to enable mobile manipulators to clear paths in cluttered real-world environments without task-specific training.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
or avoid an object and produce a high-level plan. Evaluated on a mobile manipulator across real-world indoor and outdoor environments, MessyNav improves success rate by 27 percentage points (pp) over the strongest baseline in real-robot trials and achieves statistically significant gains in a larger-scale offline evaluation, without task-specific training.
Open source S4
Source：arXiv Robotics — research abstracts · arxiv.org