Humanoid Robot Industry Now Competing on "Platforms" and "Data Collection"
Humanoid Summit Seoul 2026 On-Site
Global Robotics Companies Demonstrate Technologies at COEX
Ecosystem Expansion Through Open-Source Models and Platforms
Enhancing Robot Manipulation Capabilities Through Work Data Collection
The humanoid robot industry is entering a full-fledged phase of industrialization, progressing beyond the stage of experimental proof. No longer confined to simply verifying factory applications, companies are now deploying robots in real production sites to perform tasks, and are using the data accumulated from these operations to further enhance performance. Companies are also actively unveiling open-source platforms for humanoid robots—open platforms that allow researchers and businesses to participate in technology development and application—thereby accelerating the expansion of the ecosystem for industrial adoption.
At the site of the global Robot & AI Conference "Humanoid Summit Seoul 2026," held at COEX in Seoul on the 22nd and 23rd, leading robotics companies from around the world—including Realworld, ROBOTIS, Unitree, and Persona AI—gathered to share technological advancements and strategies for commercialization.
Korean and Chinese Robotics Companies: "Ecosystem Expansion Through Open-Source"
The global robot and AI conference, "Humanoid Summit Seoul 2026," was held from the 22nd to the 23rd at COEX in Seoul. Photo by Joon Lee.
View original imageThe event highlighted open-source platforms and models aimed at broader participation from robotics developers and researchers. In his presentation, Byungsoo Kim, CEO of ROBOTIS, introduced the company’s proprietary open humanoid platform “CYCLO” and outlined the vision of evolving humanoids into generic platforms capable of hosting numerous functions—much like a smartphone. Kim explained, "With advances in AI, a single robot can now perform over 100 different tasks. This is the direction we foresee for humanoids as universal platforms."
ROBOTIS, through CYCLO, offers services that allow multiple robotics foundation models (RFMs) to be applied to robots and their performances compared. Kim expects that adding new AI features to a single hardware device will significantly expand the scope of robot applications. ROBOTIS is also making the core development software and relevant humanoid sources for CYCLO available as open source.
Xiaoli Chen, Director of Unitree Robotics Asia-Pacific, is giving a presentation at the global robot and AI conference, "Humanoid Summit Seoul 2026," held at COEX in Seoul on the 22nd. Photo by Junyi Park.
View original imageChina's leading humanoid manufacturer Unitree also announced its plan to expand the robotics ecosystem based on its own open-source platform “UnifoLM-WLA 1.0.” Xiaoli Chen, Director of Unitree Robotics Asia Pacific, stated in a presentation that the next goal is to evolve towards an open hardware platform and to drive broader participation from developers and researchers via open foundation models, robot models, and public datasets.
Director Chen noted that Unitree is exploring the potential for humanoid robot use in settings such as mining, subways, and electronics supply chains. The focus is on deploying robots in environments that are difficult for humans due to heat, noise, vibration, or dust, or in scenarios requiring repetitive tasks, allowing operations based on both voice commands and environmental recognition. "There’s no need for pre-programming or coding," he emphasized, highlighting the robots’ ability to work continuously for extended periods. He added that to ensure stable use in industrial environments, it is essential to boost hardware reliability and computing power, and to establish local systems for maintenance, training, and services.
The Robotics Sector is Racing to Gather Data
RealWorld is demonstrating a robot equipped with its own robotics foundation model RLDX-1 (Real Dex One) moving objects on a conveyor belt at its exhibition booth during the global robot and AI conference "Humanoid Summit Seoul 2026" held on the 22nd at COEX, Seoul. Photo by Junyi Park.
View original imageThroughout the exhibition, technologies that enable robots and robotic hands to learn from and be trained on real industrial work data attracted significant attention. On this day, Realworld demonstrated a robot equipped with its proprietary robotics foundation model RLDX-1 (Real Dex One) moving objects of various sizes and shapes on a conveyor belt at its booth. According to the company, instead of pre-programmed information for each object, the robot recognizes and manipulates different objects using model inference.
A Realworld representative explained, "A key characteristic of the model is its ability to use learned data to infer and grasp unfamiliar objects, even in new environments. We are developing the technology not only to handle simple tasks but also to break down complex operations into multiple stages and link them together for execution."
Realworld is pursuing joint development and on-site validation of a logistics-focused robot foundation model in partnership with CJ Logistics. The company is also collaborating with Lotte Hotel and others to use operational data for robot learning. Their plan is to develop specialized models tailored to company-specific worker data while simultaneously creating a universal model usable across varied environments. The target is to release these solutions within the next two years.
Watts Lab, a robot learning data solution company, is showcasing its glove-type device product line at its booth at the global robot and AI conference "Humanoid Summit Seoul 2026," held on the 22nd at COEX, Seoul. Photo by Junyi Park.
View original imageWatts Lab, a company specializing in robot learning data solutions, unveiled its glove-type device “AirGlove Atlas” that enables the capture of hand motion data for humanoid robot learning. When a worker wearing the glove performs real tasks, hand motion data is collected and can be reviewed or linked to robot control and training using dedicated software called “Spine.” The company is also developing a follow-up product, “Axis,” with enhanced tactile information gathering. Currently, Watts Lab is collaborating with several robotics firms, including ROBOTIS, Robros, and Wonik Robotics.
Hot Picks Today
"Two Spend 500,000 Won in a Flash, Yet Wait Two Hours?" ... 'Dreamland' Shunned by Kids and Foreign Tourists
- The Era of Getting 200 Million Won After 30 Years Is Over... A Japanese Bank's "Perpetual Department Head" Who Just Clocked In Cries Out
- Spain Hit by Its Worst Tragedy... Lee Kangin, Moreno and Real Madrid Join Soccer World in Mourning
- They Said, “Make It Your Home for Life” - Then Suddenly Kicked 200 Passengers Off... What Happened to the Around-the-World Cruise?
- "130 Rice Bowls Sold Out in a Day"... 'Unfiltered Face' Adds 8 Million Won to Night Market Sales: China's Bungeunwoo
A Watts Lab representative stated, "While robot hardware and models are important, data is currently considered the biggest bottleneck. We aim to alleviate this constraint by collecting high-quality data based on worker movements, even without a physical robot." The company plans to further advance its technology to gather not only motion data but also various types of sensory information, such as tactile data, for use in robot learning in the future.
© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.