TongSIM-Asset

# TongSIM

<div align="center">
<img src="Docs/logo.png" width="60%" alt="Pleias" />
</div>

<div align="center">

[](https://tongsim-platform.github.io/tongsim)
[](https://github.com/bigai-ai/tongsim)
[](https://arxiv.org/abs/2512.20206)

</div>

# GitHub

Visit [Github](https://github.com/bigai-ai/tongsim) : https://github.com/bigai-ai/tongsim

# What is TongSIM-Asset?

As artificial intelligence (AI) rapidly advances, especially in multimodal large language models, research focus is shifting from single-modality text processing to the more complex domains of multimodal and embodied AI. Embodied intelligence focuses on training agents within realistic simulated environments, leveraging physical interaction and action feedback rather than conventionally labeled datasets. To foster embodied AI research, we introduce [`TongSIM`](https://github.com/bigai-ai/tongsim), a high-fidelity, general-purpose platform for training and evaluating embodied agents.

TongSIM-Asset is the digital asset package of the TongSIM platform. The TongSIM-Asset package currently contains over 100 varied, multi-room indoor scenarios, and thousands of 3D interactive objects spanning more than 500 categories. By offering this open-source asset, TongSIM-Asset provides a unified and powerful tool for the development and testing of general embodied agents, enabling researchers from diverse AI fields to advance embodied intelligence research.

This dataset is intended for use in the TongSIM simulator for embodied in-home interaction tasks, such as object rearrangement, visual navigation, multi-agent embodied simulation, and various customized tasks.

Please note that the dataset is strictly for non-commercial academic research only. All materials are licensed under the [License](https://huggingface.co/datasets/bigai/TongSIM-Asset/blob/main/Docs/LICENSE).

### Dataset Contents (this repository):

TongSIM-Asset provides a project built on Unreal Engine version 5.6. The dataset contains 100 scenes are used as a test set for challenge assessment. All scenes can run above 60FPS.

- 3000+ objects covering 500+ categories.
- 100 scenes.
- 10+ agents, including adults and children.
- 100+ agent animations

<table align="center">
<tr>
<td align="center">
<img src="Docs/scenes.png" width="100%"/>
<br/><font color="#888">Scenes</font>
</td>
<td align="center">
<img src="Docs/object.jpg" width="100%"/>
<br/><font color="#888">Objects</font>
</td>
</tr>
<tr>
<td align="center">
<img src="Docs/human.png" width="100%"/>
<br/><font color="#888">Agents</font>
</td>
<td align="center">
<img src="Docs/animation.png" width="100%"/>
<br/><font color="#888">Animations</font>
</td>
</tr>
</table>

## Dataset Structure

```
├── Config
│ ├── */*.ini
├── Content
│ ├── Animations
| ├── *.uasset
│ ├── Characters
| ├── *.uasset
│ ├── Maps
| ├── *.uasset
│ ├── Objects
| ├── *.uasset
├── Docs
│ ├── *.md
│ ├── *.png
└── TongSimLite.uproject
```

### Install Unreal Engine 5.6:

TongSIM-Asset project built on Unreal Engine version 5.6 and requires to be installed first.

[**Vist the Unreal Engine Homepage**](https://www.unrealengine.com/en-US/download)

[**Vist the Unreal Engine Documentation**](https://dev.epicgames.com/documentation/en-us/unreal-engine/unreal-engine-5-6-documentation?application_version=5.6)

---

Citing TongSIM-Asset
---

Using TongSIM-Asset in your research? Please cite the following paper: [arxiv](https://arxiv.org/abs/2512.20206)

```
@article{sun2025tongsim,
title={TongSIM: A General Platform for Simulating Intelligent Machines},
author={Sun, Zhe and Wu, Kunlun and Fu, Chuanjian and Song, Zeming and Shi, Langyong and Xue, Zihe and Jing, Bohan and Yang, Ying and Gao, Xiaomeng and Li, Aijia and others},
journal={arXiv preprint arXiv:2512.20206},
year={2025}
}
```

# TongSIM

<div align="center">
<img src="Docs/logo.png" width="60%" alt="Pleias" />
</div>

<div align="center">

[](https://tongsim-platform.github.io/tongsim)
[](https://github.com/bigai-ai/tongsim)
[](https://arxiv.org/abs/2512.20206)

</div>

# GitHub 仓库

访问 [GitHub仓库](https://github.com/bigai-ai/tongsim)：https://github.com/bigai-ai/tongsim

## 什么是TongSIM-Asset？

随着人工智能（AI）快速发展，尤其是多模态大语言模型（Large Language Model）领域，研究焦点正从单模态文本处理转向更复杂的多模态与具身AI（Embodied AI）领域。具身智能（Embodied Intelligence）聚焦于在逼真的模拟环境中训练智能体（Agent），依托物理交互与动作反馈，而非传统的标注数据集。为推动具身AI研究，我们推出了**TongSIM**——一款用于训练与评估具身智能体的高保真通用平台。

TongSIM-Asset是TongSIM平台的数字资源包。当前TongSIM-Asset套件包含100余个多样化的多房间室内场景，以及覆盖500余个类别的数千个可交互3D物体。通过开源这一资源包，TongSIM-Asset为通用具身智能体的开发与测试提供了统一且强大的工具，助力AI各领域研究者推进具身智能研究。

本数据集专为TongSIM模拟器中的居家具身交互任务设计，例如物体重排、视觉导航、多智能体具身模拟以及各类定制化任务。

请注意：本数据集仅可用于非商业性学术研究。所有材料均遵循[许可协议](https://huggingface.co/datasets/bigai/TongSIM-Asset/blob/main/Docs/LICENSE)。

### 数据集内容（本仓库）

TongSIM-Asset提供了基于虚幻引擎5.6（Unreal Engine 5.6）构建的项目。本数据集包含100个场景，用作挑战评估的测试集。所有场景均可稳定运行于60FPS以上。
- 3000余个覆盖500余个类别的物体
- 100个场景
- 10余种智能体，包含成人与儿童形象
- 100余种智能体动画

<table align="center">
<tr>
<td align="center">
<img src="Docs/scenes.png" width="100%"/>
<br/><font color="#888">场景</font>
</td>
<td align="center">
<img src="Docs/object.jpg" width="100%"/>
<br/><font color="#888">物体</font>
</td>
</tr>
<tr>
<td align="center">
<img src="Docs/human.png" width="100%"/>
<br/><font color="#888">智能体</font>
</td>
<td align="center">
<img src="Docs/animation.png" width="100%"/>
<br/><font color="#888">动画</font>
</td>
</tr>
</table>

## 数据集结构

├── Config
│ ├── */*.ini
├── Content
│ ├── Animations
| ├── *.uasset
│ ├── Characters
| ├── *.uasset
│ ├── Maps
| ├── *.uasset
│ ├── Objects
| ├── *.uasset
├── Docs
│ ├── *.md
│ ├── *.png
└── TongSimLite.uproject

### 安装虚幻引擎5.6

TongSIM-Asset项目基于虚幻引擎5.6构建，需先安装该引擎。
[**访问虚幻引擎官网**](https://www.unrealengine.com/en-US/download)
[**访问虚幻引擎5.6文档**](https://dev.epicgames.com/documentation/en-us/unreal-engine/unreal-engine-5-6-documentation?application_version=5.6)

---

## 引用TongSIM-Asset

若您的研究中使用了TongSIM-Asset，请引用以下论文：[arXiv](https://arxiv.org/abs/2512.20206)

@article{sun2025tongsim,
title={TongSIM: A General Platform for Simulating Intelligent Machines},
author={Sun, Zhe and Wu, Kunlun and Fu, Chuanjian and Song, Zeming and Shi, Langyong and Xue, Zihe and Jing, Bohan and Yang, Ying and Gao, Xiaomeng and Li, Aijia and others},
journal={arXiv preprint arXiv:2512.20206},
year={2025}
}