Nota Moves to Connect Robots With Domestic NPUs
IT DAILY ·
✦ AI Summary
Nota has been selected as the lead organization for the humanoid field's third subproject in the Ministry of Trade, Industry and Energy and KEIT's "K-On-Device AI Semiconductor Technology Development Project."
The project aims to develop autonomous humanoids using domestic on-device AI semiconductors.
Nota will handle VLA lightweighting and Mobilint NPU-tailored optimization software, and will advance real-time operation and integrated environment verification.
Nota announced on the 8th that it had been selected as the lead organization for the third subproject in the humanoid field of the "K-On-Device AI Semiconductor Technology Development Project," which is being promoted by the Ministry of Trade, Industry and Energy and the Korea Evaluation Institute of Industrial Technology (KEIT).
The goal of the project is to develop autonomous humanoids using domestic on-device AI semiconductors. The performance required of autonomous humanoids is to recognize complex everyday environments and carry out human-level tasks.
LG Electronics is overseeing the overall project. LG Electronics will be responsible for developing the humanoid system.
Mobilint will handle the domestic NPU. Nota will be responsible for optimization software for AI models to run efficiently in NPU and real-world robot environments, and will lead the "Humanoid AI Model Optimization" project.
The technology Nota will handle includes lightweighting the vision-language-action model (VLA) installed in humanoids and optimizing it for Mobilint NPU. The goal is to enable fast and efficient operation inside the robot.
Nota plans to apply optimization to the vision-language-action model, and will reduce model size, computing load, and memory usage while minimizing performance degradation through token compression, quantization, pruning, distillation, and layer-specific optimization. Nota uses a balance among accuracy, computing load, and communication latency as its lightweighting criteria.
Nota plans to convert the lightweight model to match the computing architecture of Mobilint NPU and also build modules linked to a real-time inference environment and workloads. As a follow-up plan, it will verify the integrated environment of Mobilint NPU and the humanoid, and in the process check real-time operation performance and resource utilization efficiency.
Nota CEO Chae Myeong-su said this project is an example of applying the AI optimization technologies it has accumulated to humanoids based on domestic NPUs. He also said the company has technologies that efficiently connect AI models and hardware in mobile, edge, and data center environments, and added that it plans to expand technology applications and business opportunities across physical AI areas that require real-time on-device AI, starting with humanoids and extending to robots, automobiles, and industrial equipment.
Following its participation in the earlier "Independent AI Foundation Model" project, Nota is also taking part in the "Everyone's AI" project, through which it is providing optimization technology across the entire process from AI model development to actual service implementation.
Source: IT DAILY · Kwon Yeong-seok
Original: https://www.itdaily.kr/news/articleView.html?idxno=241470
References
This article was produced with the help of an automated content generation algorithm.
Source: IT DAILY
View originalThis article was summarized and organized by BizCrush based on the original article from IT DAILY. For exact quotations and full details, please refer to the original article.