EndotLight participates in a project to develop 3D generation technology for physical AI model training.

3D AI specialized company Endotlight announced that it will participate as a joint research and development institution in the 'World Foundation Model Technology Development for Physical AI Model Learning' project hosted by the Ministry of Science and ICT and the National IT Industry Promotion Agency.

This project is a large-scale government project led by NC AI, with participation from domestic AI, robotics, and simulation organizations including Endolight, Samsung SDS, KAIST, MBC, Rainbow Robotics, and Real World. Endolight is responsible for developing generative AI-based 3D virtual asset technology and building an "E2E data lake."

End.Rite develops web-based tools that transform user-entered text or images into 3D CAD models. Unlike traditional visual 3D models, it generates "Sim-Ready" 3D assets that incorporate physical properties such as mass, friction, and collision meshes. Furthermore, by leveraging the "Omniverse Connector," developed in collaboration with NVIDIA, End.Rite provides an end-to-end pipeline that enables real-time integration of data generated with End.Rite's proprietary 3D CAD engine, End.Rite, with Omniverse and the NVIDIA Robotics Platform.

Endotrite's 'Sim-Ready Synthetic Data Generation Solution' TRINIX supports the global 3D standard OpenUSD and the URDF and MJCF formats for robot simulation, and can be used directly in the global robotics development environment through the Omniverse extension.

The ultimate goal of this project is to generate large-scale synthetic data for training robot vision-language-action (VLA) models in a virtual environment and load the entire process into a data lake using an automated, one-stop pipeline.

Park Jin-young, CEO of Endotlight, said, “Through this project, we will showcase engineering-level 3D creation technology that takes physical properties and lightweighting into account, and based on OpenUSD-based technology, we will establish ourselves as an essential AI infrastructure for the domestic and international robotics and digital twin markets.”


  • See more related articles