Reimagining service delivery in the agentic era with Google Public Sector
cloud.google.com·21h ago
AI Research
arXiv CS.AI·Jun 9·5 min read
ReCoVLA: VLM-Guided Reward Compilation for Failure Recovery in Vision-Language-Action Policies
TL;DR
Vision-language-action (VLA) policies struggle with recovery from unexpected failures. ReCoVLA is a framework that uses a vision-language model (VLM) to identify failure modes and generate structured rewards for recovery training.
✦ Why It Matters
Engineers can leverage ReCoVLA to improve the reliability of robotic systems in unpredictable environments.
Key Takeaways
- Utilize ReCoVLA for robust failure recovery in robotic tasks.
- Leverage external VLMs to improve reward selection.
- Train residual policies in simulation for real-world deployment.
- Achieve higher success rates in manipulation tasks with ReCoVLA.
How It Works
ReCoVLA operates by freezing a pretrained VLA policy and employing a vision-language model (VLM) to analyze failure scenarios. The VLM predicts recovery descriptors and generates a reward mask, which guides the training of a residual policy in simulation.
This decoupling of high-level understanding from low-level control allows for flexible adaptation across different VLA systems.