🤖 AI Summary
This work addresses the challenge of generalizing control and ensuring safe execution for general-purpose robots across heterogeneous morphologies—such as humanoids, mobile manipulators, and fixed-base arms—in real-world environments. The authors propose a staged vision-language-action framework that integrates a five-phase curriculum learning strategy, a unified embodied perception-action interface, pretraining with multimodal foundation models, and a safety-enhancement mechanism during inference. Innovatively combining reinforcement learning policy alignment, temporally aligned data processing, and out-of-distribution detection, the approach significantly improves task success rates, robustness, and long-horizon execution efficiency. Extensive experiments in both simulation and real-world robotic platforms demonstrate the method’s effectiveness and performance gains in cross-platform deployment.
📝 Abstract
We introduce Green-VLA, a staged Vision-Language-Action (VLA) framework for real-world deployment on the Green humanoid robot while maintaining generalization across diverse embodiments. Green-VLA follows a five stage curriculum: (L0) foundational VLMs, (L1) multimodal grounding, (R0) multi-embodiment pretraining, (R1) embodiment-specific adaptation, and (R2) reinforcement-learning (RL) policy alignment. We couple a scalable data-processing pipeline (3,000 hours of demonstrations) with temporal alignment and quality filtering, and use a unified, embodiment-aware action interface enabling a single policy to control humanoids, mobile manipulators, and fixed-base arms. At inference, the VLA controller is enhanced with episode-progress prediction, out-of-distribution detection, and joint-prediction-based guidance to improve safety and precise target selection. Experiments on Simpler BRIDGE WidowX and CALVIN ABC-D, as well as real-robot evaluations, show strong generalization and performance gains from RL alignment in success rate, robustness, and long-horizon efficiency.