Caging Vision-Language Navigation for Legged Robots via Barrier-Certified Safety Architecture
Caging Vision-Language Navigation for Legged Robots via Barrier-Certified Safety Architecture
This research introduces a safety architecture that augments pretrained Vision-Language Models for legged robots with physical grounding and explicit collision-avoidance guarantees. By combining a high-level VLA planner with a low-level layer utilizing Signed Distance Fields and Discrete Exponential Barrier Function constraints, the system ensures safe navigation under stated assumptions. The modular design is compatible with various VLM/VLA models and robot platforms, and it is validated through extensive real-world experiments on both humanoid and quadrupedal robots.
The paper proposes a modular safety architecture that cages Vision-Language Action models for legged robots by integrating high-level VLA planning with low-level physical grounding using Signed Distance Fields and Discrete Exponential Barrier Functions to provide provable collision-avoidance guarantees, validated on both humanoid and quadrupedal platforms.
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
- code: not_reported
- weights: not_reported
- dataset: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
odels and agnostic to the robot platform. The proposed system is validated with extensive real-world experiments on humanoid and quadrupedal robots, demonstrating versatile, efficient, and safe vision-language driven navigation capabilities.
Open source S5
Source：arXiv Robotics — research abstracts · arxiv.org