InfiniteWorld: A Unified Scalable Simulation Framework for General Visual-Language Robot Interaction

TL;DR

InfiniteWorld presents a unified, scalable simulation framework for vision-language robot interaction built on Nvidia Isaac Sim, addressing fragmented asset interfaces and limited social interaction benchmarks. It introduces generation-driven 3D asset construction, a Real2Sim pipeline, and Annot8-3D for automated annotation, coupled with four benchmarks (Object Loco-Navigation, Loco-Manipulation, Scene Graph Collaborative Exploration, Open-World Social Mobile Manipulation) to comprehensively evaluate perception, planning, execution, and interaction. The framework enables unlimited asset expansion and realistic human-like interactions via LLM-driven prompts and multi-agent collaboration, showing strong performance gains for LLM-based planners in certain tasks and highlighting challenges in open-world social scenarios. Overall, InfiniteWorld offers a cohesive asset ecosystem, a scalable annotation platform, and systematic benchmarks to catalyze embodied AI scaling and cross-platform research.

Abstract

Realizing scaling laws in embodied AI has become a focus. However, previous work has been scattered across diverse simulation platforms, with assets and models lacking unified interfaces, which has led to inefficiencies in research. To address this, we introduce InfiniteWorld, a unified and scalable simulator for general vision-language robot interaction built on Nvidia Isaac Sim. InfiniteWorld encompasses a comprehensive set of physics asset construction methods and generalized free robot interaction benchmarks. Specifically, we first built a unified and scalable simulation framework for embodied learning that integrates a series of improvements in generation-driven 3D asset construction, Real2Sim, automated annotation framework, and unified 3D asset processing. This framework provides a unified and scalable platform for robot interaction and learning. In addition, to simulate realistic robot interaction, we build four new general benchmarks, including scene graph collaborative exploration and open-world social mobile manipulation. The former is often overlooked as an important task for robots to explore the environment and build scene knowledge, while the latter simulates robot interaction tasks with different levels of knowledge agents based on the former. They can more comprehensively evaluate the embodied agent's capabilities in environmental understanding, task planning and execution, and intelligent interaction. We hope that this work can provide the community with a systematic asset interface, alleviate the dilemma of the lack of high-quality assets, and provide a more comprehensive evaluation of robot interactions.

Figures (3)

• Figure 1: An overview of the proposed benchmark.
• Figure 2: Comparative of reconstruction results with and without post-processing. Visualizing the results shows that our post-processing method is very effective in resolving holes and removing floating meshes in the scene, such as the gaps around cabinets, sofas, and the floaters above the table in the red bounding boxes.
• Figure 3: Visualization of Benchmark 1 and Benchmark 2.