Doosan Robotics said on the 6th that its next-generation collaborative robot project and intelligent welding solution project were both selected as state projects under the "K On-Device AI Semiconductor Technology Development Program" and the "Robot Industry Technology Development Program," jointly overseen by the Ministry of Trade, Industry and Energy and the Korea Evaluation Institute of Industrial Technology. Total research and development spending on the projects comes to 98.9 billion won, of which about 68.1 billion won is government funding.
The K On-Device AI semiconductor project aims to develop the foundational technology that strengthens a robot's perception, judgment and control capabilities — the equivalent of its brain. The robot industry project covers the development of applied technology that puts those capabilities to work at demanding manufacturing sites. The core of the technology is enabling robots to perceive, judge and control on their own without an external PC or cloud. To that end, a domestic AI system-on-chip being developed by AI chip startup Mobilint will combine a neural processing unit with real-time communication and control functions and be mounted on collaborative robots. The robots will make decisions on site and carry out tasks based on information from cameras, touch, force and torque sensors, and voice.
Doosan Robotics will develop three collaborative robot models with payloads — the maximum weight a robot can lift and move — of 5 kilograms, 10 kilograms and 20 kilograms equipped with the chip, along with an industrial humanoid capable of two-armed work and omnidirectional movement. The company plans to commercialize them from 2031 after conducting 72-hour nonstop demonstrations at two or more customer sites and obtaining international safety certification. It expects the move to expand collaborative robot supply into security-sensitive industrial settings such as nuclear power, defense and biotechnology, which have delayed automation over concerns about communication latency and data leaks.
The intelligent welding solution project, which combines collaborative robots with AI and digital twins, centers on automating highly difficult welding at the level of nuclear power plant components. Robots will recognize welding targets through cameras and correct their position on their own, while imitating and learning the movements of skilled workers to automatically generate welding paths and weaving motions. Weaving is a technique of moving the welding torch from side to side to make the weld width and quality uniform.
Doosan Robotics aims to match robot welding paths with those of skilled workers at a rate of 95% or higher and to reduce path error to within plus or minus 3 millimeters by the standards of a state-accredited testing institution. It also plans to automate multilayer welding of five or more layers, cutting work time by more than 50%.
Demonstrations will be carried out at Doosan Enerbility's (034020) nuclear power center. The technology will first be applied to tack welding, preheating and welding in the manufacturing of nuclear power plant components, then extended to shipbuilding, plants and defense after securing nuclear quality certification. The project runs for 45 months and is led by Doosan Robotics. DeepX, Sage and the Changwon National University Industry-Academic Cooperation Foundation will serve as joint research institutions, with Doosan Enerbility participating as a customer company.
The company said the projects carry significance beyond improving robot performance, as they mark a full-fledged attempt to integrate AI chips, software and robots using domestic technology.
"Labor shortages at manufacturing sites and the aging of skilled workers are structural problems that cannot be solved simply by adding automation equipment," said Park In-won, president of Doosan Robotics. "By combining on-device AI chips with our own collaborative robot platform, we will turn physical AI collaborative robots that learn and reproduce even the tacit knowledge of skilled workers into a new competitive strength for Korean manufacturing."
