NC AI will join a national research and development (R&D) project based on physical AI with Hyundai Rotem.
NC AI said on the 28th that it formed a consortium with Hyundai Rotem and was selected as the final contractor for the national research and development project, "physical AI-based integrated simulator and modular robot system," commissioned by the Agency for Defense Development (ADD).
The project aims to organically control diverse, multiple unmanned robots and build a Digital Twin-based simulator and modular robot hardware to improve the efficiency of manned-unmanned combined weapon systems in future battlefields.
NC AI will lead development of the "world model," the brain of robots and a core technology to implement the next-generation robot foundation model (RFM). The world model is a technology that lets robots simulate across various environments by reflecting the physical laws of the real world and environmental changes, and turn those into training data.
The world model is regarded as the key to solving the "simulation-to-reality gap (Sim-to-Real Gap)" problem, in which unmanned robots malfunction in real environments due to physical variables or friction. NC AI also plans to build a pipeline to efficiently supply high-quality "large-scale synthetic data" needed for robot training.
In Mar., NC AI unveiled its world foundation model (WFM) technology that delivers similar performance while using about 25% of the graphics processing units (GPUs) compared with the global state-of-the-art (SOTA) model. While conventional world models first generate video and then infer through a vision-language model (VLM), NC AI adopted an architecture that directly derives robot behavior from information in the "latent space," the stage just before video generation.
Lee Yeon-su, CEO of NC AI, said, "Based on NC AI's own lightweight world model, we will develop a physical AI simulator that connects virtual and real to take the lead in accelerating Korea defense AX."