What is EmbodiedGen
EmbodiedGen is a generative 3D world engine and toolkit for Embodied AI applications. It can quickly generate high-quality, low-cost, and physically plausible 3D assets and interactive environments, helping researchers and developers build testing environments for embodied agents. EmbodiedGen includes multiple modules, such as generating 3D models from images or text, texture generation, articulated object generation, scene and layout generation, etc., supporting the creation from simple objects to complex scenes. The generated 3D assets can be directly used for robot simulation and in URDF format, providing powerful tool support for embodied AI research.
Main Features of EmbodiedGen
-
Image-to-3D Conversion: It can generate physically plausible 3D assets from input images.
-
Text-to-3D Generation: It generates 3D assets of various geometries and styles based on text descriptions.
-
Texture Generation Feature: It generates visually rich textures for 3D meshes.
-
Complex Scene Construction: It supports the creation from simple objects to complex scenes, and can generate high-quality 3D assets with real-world scale and compliant with the Unified Robot Description Format (URDF).
-
Intelligent Layout Generation: It provides intelligent layout generation capabilities, supporting downstream tasks in training and evaluation.
-
Physical Property Support: The generated 3D assets have sealed geometric structures and physically plausible properties, and can be directly applied to robot simulation and description formats.
Technical Principles of EmbodiedGen
-
Application of Generative AI: EmbodiedGen is based on generative AI technology and can generate 3D models from images or text descriptions.
-
Multi-Module Collaboration: EmbodiedGen includes six key modules: image-to-3D, text-to-3D, texture generation, articulated object generation, scene generation, and layout generation. The modules work together to generate diverse 3D worlds from simple objects to complex scenes.
-
Physical Realism and Real-World Scale: The generated 3D assets have sealed geometric structures and physically plausible properties, and can be directly applied to robot simulation and description formats such as URDF (Unified Robot Description Format).
-
Dynamic Environment Generation: The environments generated by EmbodiedGen are dynamic and can generate and modify environments in real time according to the AI's behavior.
EmbodiedGen Project Address
- Project Website：https://horizonrobotics.github.io/robot_lab/embodied_gen/index.html
- Github Repository：https://github.com/HorizonRobotics/EmbodiedGen
- arXiv Technical Paper：https://arxiv.org/pdf/2506.10600
Application Scenarios of EmbodiedGen
- Robot Simulation and Training: EmbodiedGen can generate 3D assets with physical plausibility and real-world scale, which can be directly applied to robot simulation and description formats such as URDF (Unified Robot Description Format).
- Autonomous Driving and Drones: The dynamic 3D environments generated by EmbodiedGen can be used for simulation training of autonomous driving and drones. By simulating complex road and terrain conditions, it helps autonomous driving systems and drones better adapt to real-world scenarios.
- Virtual Social Interaction: Users can control virtual avatars through VR devices for socializing, meetings, and other activities.
- Healthcare and Rehabilitation: 3D environments generated by EmbodiedGen can be used for simulation and training in the medical and rehabilitation fields. Surgical operation simulation training is conducted through virtual environments.