RealMirror: A Comprehensive, Open-Source Vision-Language-Action Platform for Embodied AI

2025-09-18

AI Summary

Humanoid robot vision-language-action (VLA) research faces three key challenges: high data acquisition costs, the absence of standardized benchmarks, and the persistent simulation-to-reality (Sim2Real) gap. To address these, this work introduces the first standardized VLA simulation benchmark platform specifically designed for humanoid robots. Leveraging generative modeling and 3D Gaussian splatting, the platform constructs high-fidelity virtual environments and robot embodiments, enabling end-to-end data collection, training, and inference entirely in simulation—without requiring real-world hardware. Crucially, our end-to-end VLA system, trained solely in simulation, achieves zero-shot transfer to physical humanoid robots without fine-tuning, and demonstrates direct usability across diverse embodied tasks. This platform substantially lowers development barriers, accelerates VLA model iteration, and bridges the performance and deployment gap between simulation and reality.

Technology Category

• Intelligent Robots: Embodied AI
• Computer Vision: Language and Vision
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search

Abstract

The emerging field of Vision-Language-Action (VLA) for humanoid robots faces several fundamental challenges, including the high cost of data acquisition, the lack of a standardized benchmark, and the significant gap between simulation and the real world. To overcome these obstacles, we propose RealMirror, a comprehensive, open-source embodied AI VLA platform. RealMirror builds an efficient, low-cost data collection, model training, and inference system that enables end-to-end VLA research without requiring a real robot. To facilitate model evolution and fair comparison, we also introduce a dedicated VLA benchmark for humanoid robots, featuring multiple scenarios, extensive trajectories, and various VLA models. Furthermore, by integrating generative models and 3D Gaussian Splatting to reconstruct realistic environments and robot models, we successfully demonstrate zero-shot Sim2Real transfer, where models trained exclusively on simulation data can perform tasks on a real robot seamlessly, without any fine-tuning. In conclusion, with the unification of these critical components, RealMirror provides a robust framework that significantly accelerates the development of VLA models for humanoid robots. Project page: https://terminators2025.github.io/RealMirror.github.io

Problem

• Addresses high-cost data acquisition in humanoid robot VLA research
• Overcomes simulation-to-real world transfer gap without fine-tuning
• Provides standardized benchmark for embodied AI model comparison

Innovation

• Open-source platform for embodied AI VLA
• Generative models reconstruct realistic simulation environments
• Zero-shot Sim2Real transfer without fine-tuning