[The Upcoming K-Robot Era]④Korea’s Competitive Edge Lies in ‘Data’... "Accumulation and Utilization Enable Software Differentiation"

ChatGPT Image Generation.

원본보기 아이콘

Korean robotics company Robotis has established a subsidiary in Uzbekistan and is embarking on the construction of a large-scale data factory. This move appears to be based on the assessment that "data" is the area where Korea, with its manufacturing strengths, can differentiate itself from the United States and China.

A representative from Robotis stated, "The United States excels in software-based technologies, but due to its lack of a robust manufacturing environment, there is high demand for data. However, since it cannot outsource data to China, high-quality data from Korea is needed." He added, "By expanding data factory operations in Uzbekistan following our domestic operations, we will gather 'action data,' further enhance our competitiveness, and accelerate our global expansion."

Park Ilwoo, Robotics Program Director (PD) at the Ministry of Trade, Industry and Energy, said, "The Korean government aims to become one of the 'top three humanoid robot powers' through ecosystem cooperation centered on the 'MAX (Manufacturing and AI Transformation) Alliance,'" and added, "We are currently discussing with the industry various policies, such as securing data for humanoid robot training, internalizing core component technologies like actuators and robot hands, and expanding real-world demonstrations in industrial robot sites."

On January 6, the opening day of CES, a representative from the Korean robotics company Robotis demonstrated a robot that recognizes hand gestures with a camera and imitates them at the Las Vegas Convention Center (LVCC) in Nevada, USA. Photo by Yonhap News

원본보기 아이콘

The global competition among robotics companies to secure data is fierce. China recognized the importance of data early on and is currently operating data factories in seven locations nationwide to collect data, which is essential for advancing artificial intelligence (AI). The Chinese government is focusing its national capabilities on gathering data for use in robots and has announced a policy for 'Innovation and Development of Humanoid Robots,' aiming to introduce robots to 1,000 industries and 100 sub-industries by 2028.

In particular, Chinese companies are actively collecting vast amounts of data through their own data factories. AgiBot, a humanoid robotics company based in Shanghai, has created a space of over 4,000 square meters, simulating environments such as homes, restaurants, industries, shopping malls, and offices, and collects and analyzes all human activity data occurring in these spaces to train its robots. By recreating real-world environments and focusing on data collection, AgiBot released its robot AI model as open source in December 2024, and within just three years of its founding, has emerged as the world’s leading humanoid robotics company.

On the 4th, at the 2026 Smart Factory and Automation Industry Exhibition held at COEX, a Chinese humanoid robot is posing. From the left, Aegibot G2, Unitree G1, Reju Quavo4 Pro. Photo by Yonhap News Agency

원본보기 아이콘

원본보기 아이콘

Experts predict that, going forward, countries and companies possessing both "on-site data acquisition capabilities" and "data generation capabilities" are highly likely to gain the upper hand in the robotics industry. Unlike video data, robot data is extremely expensive to collect and accumulates at a very slow pace. Additionally, there are safety concerns, difficulties in ensuring environmental diversity, and a lack of standardized data structures, making it inevitable that those who possess robot data will have a competitive edge. Given the structural difficulties and numerous constraints in acquiring robot data, there is growing importance placed on data generation capabilities using methods such as simulation, synthetic data, and world model-based techniques.

Kim Ikjae, Director of the AI and Robotics Research Center at the Korea Institute of Science and Technology (KIST), stated, "While simulation and generated data are important, it remains essential to acquire real-world data from the field." He pointed out, "The physical world inherently includes complex factors such as friction, contact, exceptional situations, and interactions with humans, which cannot be fully learned through virtual environments alone."

Looking at the evolution of generative AI, current AI models are still built on data-dependent structures, and as more large-scale data is trained, their performance and generalization capabilities improve, allowing them to operate reliably in diverse environments. Director Kim emphasized, "The same applies to physical AI," adding, "The key challenges in the era of physical AI are not only how much data is collected, but also how quickly real-world experiences can be converted into formats suitable for learning."

As the United States and China compete for leadership in the robotics industry, data is also seen as the field where Korea, with its manufacturing strengths, can differentiate itself.

Son Dongsin, Committee Member at the Future Robotics Lab at LG CNS, commented, "In actual industrial settings, it is necessary to have data that enables robots to perform tasks under human-level strict requirements. It is difficult to achieve human-level proficiency in real-world tasks using only simulation-generated data, so on-site data acquisition is essential. This is an area where Korea, with its accumulated manufacturing experience and know-how across diverse sectors, can secure a competitive advantage."

An Seongwon, Director of AI Policy Research at the Software Policy & Research Institute, said, "Ultimately, for robots to respond to countless variables, it is critical for pre-trained models to have learned from as much data as possible." He added, "To break into global markets such as the Middle East, we need to develop and pursue specialized models for manufacturing, finance, and healthcare-fields in which we excel and have amassed significant data."

Related article: [Exclusive] "Largest-scale Data Factory"...Robotis to Build Bigger than China's No. 1

© The Asia Business Daily(www.asiae.co.kr). All rights reserved.