A South Korean consortium led by NC AI and LG Electronics has been selected for a government-funded project to build autonomous humanoid robots powered by domestically developed on-device AI semiconductors. The project, announced September 3, falls under the K-On-Device AI Semiconductor Technology Development Project run by the Ministry of Trade, Industry and Resources and the Korea Planning & Evaluation Institute of Industrial Technology (KEIT).
The goal is to create humanoids that can perform complex tasks in everyday, unstructured environments - such as homes - without relying on external servers. On-device AI processes all data locally on the robot's own chips, reducing latency and eliminating cloud dependence.
What the consortium will build
The group will jointly develop high-performance, low-power AI semiconductors and the control technologies that let humanoids move in real time within physical spaces. LG Electronics and other participants will integrate robot hardware, training data, AI models, and the domestic chips into a single development framework.
Initial use cases target general-purpose humanoids capable of organizing and transporting objects, serving, and handling tools. The consortium also plans to expand into industrial and public-sector applications, including manufacturing support, logistics, disaster response, facility inspections, and work in hazardous areas.
NC AI's role: motion data and world models
NC AI will build the initial training data for learning-based general-purpose motion tracking and a human-to-robot retargeting pipeline. Retargeting converts human movements into motions a humanoid can physically execute, accounting for differences in body proportions, joint limits, foot support, and balance constraints.
Rather than collecting data piecemeal in physical environments, NC AI will use digital twin simulations to generate large-scale motion datasets across varied task scenarios. The company will combine this with its own world model technology, which aims to let robots predict environmental changes, make independent decisions about multi-step tasks, and execute them sequentially.
Why this matters for IT and development professionals
For developers and engineers working with AI, this project signals a concrete push toward AI for Software Developers in physical systems. The pipeline NC AI is building - from simulation-based training data generation to on-device inference - mirrors patterns already familiar in software AI but applies them to the harder constraints of real-world physics and hardware.
On-device AI semiconductors remove the network dependency that makes cloud-based robotics brittle. Anyone working on edge AI, embedded systems, or robotics middleware will recognize the integration challenges here: model optimization, real-time control loops, and sensor fusion all running on low-power domestic silicon. The project's scope across AI for IT & Development spans motion data pipelines, digital twin environments, and autonomous decision-making - a full stack that demands both software and hardware expertise.
Your membership also unlocks:
