NdotLight and ROBROS Collaborate on Commercializing Humanoid AI Training Data
NdotLight participated in the event as a member of NVIDIA Inception, NVIDIA’s startup acceleration program. During the exhibition, the company demonstrated its proprietary TRINIX solution, showcasing Text-to-CAD and Sim-Ready asset generation technologies at the NVIDIA booth.
Under the partnership, NdotLight will utilize TRINIX to automatically generate Sim-Ready 3D assets embedded with articulated joint structures and physical properties. The company will also support integration with robot simulation platforms such as NVIDIA Isaac Sim, digital twin environment development, and simulation data pipeline construction.
TRINIX is a neuro-symbolic solution that combines a 3D CAD engine with generative AI. It enables end-to-end creation of 3D assets from text prompts, images, and existing CAD data, automatically incorporating physical properties, collision meshes, and joint structures. The platform supports industry-standard formats including USD, URDF, and MJCF, allowing seamless integration with major robotics simulation environments.
ROBROS will lead the development of humanoid robot hardware and software, including AI algorithms, reinforcement learning-based control systems, training scenario design, data requirement definition, and the deployment and validation of simulation-trained models on physical robots.
Humanoid robot reinforcement learning typically requires millions of simulation iterations. However, building simulation-ready environments and generating suitable 3D assets have long been recognized as major bottlenecks in robotics development.
By combining NdotLight’s Sim-Ready 3D asset generation technology with ROBROS’ humanoid robotics platform and AI control capabilities, the two companies aim to significantly reduce these bottlenecks. The collaboration is expected to shorten robot training cycles in virtual environments while advancing the commercialization of simulation data essential for humanoid robot deployment.
“We aim to accelerate the entire AI training pipeline for humanoid robots through the convergence of our technologies and establish a leading position in the global digital twin and robotics simulation data market,” said Jinyoung Park, CEO of NdotLight.
“By integrating NdotLight’s TRINIX-based Sim-Ready 3D asset generation technology with ROBROS’ humanoid hardware, software, and AI control technologies, we will shorten robot learning cycles and develop humanoid solutions capable of delivering real value in industrial environments,” said Seungjun Roh, CEO of ROBROS.