EmbodiedGen V2: An Agentic, Simulation-Ready 3D World Engine for Embodied AI

2026-07-08

AI Summary

Existing approaches struggle to automatically assemble generated 3D assets into task-level simulation environments, heavily relying on manual intervention and thereby limiting the scalability of embodied intelligence for closed-loop learning. This work proposes the first unified, generative, editable, and reusable simulation-ready 3D world engine that enables end-to-end automated construction of complete task environments from individual assets. By integrating a unified simulation representation, stateful Vibe Coding, interaction-aware semantic modeling, and a generative asset pipeline, our method supports deployment across simulators and seamless transfer to real robots. Evaluated on navigation and manipulation tasks, it achieves significant performance gains: 83.3% of generated environments are directly simulation-ready, reinforcement learning success rates improve from 9.7% to 79.8%, and real-robot task success increases from 21.7% to 75.0%.

Abstract

We present EmbodiedGen V2, a generative 3D world engine for building executable sim-ready environments for embodied intelligence. Sim-ready 3D asset generation has advanced rapidly, yet assembling such assets into policy-ready task environments remains largely manual, limiting scalable closed-loop learning. EmbodiedGen V2 addresses this gap through a unified sim-ready representation that connects cross-simulator assets, interaction affordances, task-driven worlds, large-scale multi-room scenes, and stateful Vibe Coding into a generative, editable, and reusable simulation pipeline. The generated environments support manipulation, navigation, mobile manipulation, cross-simulator deployment, and embodied policy training. In evaluation, the asset pipeline achieves 96.5% human acceptance and 98.6% collision success, and 83.3% of task-driven worlds are directly usable for downstream simulation without manual modification. Online reinforcement learning with generated environments further improves simulation success from 9.7% to 79.8%, and transfers to real robots with task success increasing from 21.7% to 75.0%. These results establish EmbodiedGen V2 as scalable simulation infrastructure for training, evaluating, and deploying embodied policies.

Problem

• embodied AI
• sim-ready environments
• 3D world generation
• task-driven simulation
• scalable learning

Innovation

• sim-ready 3D generation
• embodied AI
• task-driven environments
• cross-simulator compatibility
• generative simulation pipeline