Next-generation collaborative robot and intelligent welding solution to use domestic AI chips
Collaborative robot to operate without cloud connectivity
High-precision welding automation targets nuclear power plant components
Doosan Robotics announced Tuesday that it has secured two government-funded research projects related to physical AI.
The two projects — a next-generation collaborative robot initiative and an intelligent welding solution initiative — were selected under programs jointly overseen by the Ministry of Trade, Industry and Energy and the Korea Evaluation Institute of Industrial Technology (KEIT): the K-On-Device AI Semiconductor Technology Development Program and the Robot Industry Technology Development Program, respectively.
The combined research and development budget for the two projects totals approximately 98.9 billion won ($72.7 million), of which about 68.1 billion won will come from government funding.
The first project, focused on developing next-generation collaborative robots based on on-device AI semiconductors, aims to enable robots to perform perception, decision-making and control on their own — without relying on an external PC or cloud connection.
A domestically developed AI system-on-chip integrating a neural processing unit for AI computation alongside real-time communication and control functions will be embedded directly in the collaborative robot. The robot will process inputs from cameras, tactile sensors, force-torque sensors and voice to make on-site decisions autonomously.
Doosan Robotics plans to develop three collaborative robot models equipped with the chip — with payload capacities of 5 kilograms, 10 kilograms and 20 kilograms — as well as an industrial humanoid robot capable of dual-arm operation and omnidirectional movement.
The robots will undergo 72-hour continuous operation trials at two or more deployment sites and obtain international safety certification before being commercialized as standard products from 2031. Once commercialized, the company expects to expand supply of collaborative robots into security-sensitive environments such as nuclear power plants, the defense industry and biotech facilities.
The project runs for 54 months, with Doosan Robotics serving as the lead institution. Participating organizations include Aidin Robotics, Mobilint, the Korea Electronics Technology Institute, the Korea Institute of Industrial Technology, Sage, and industry-academic cooperation foundations at Korea University, Dongguk University, Sejong University and Yonsei University.
The second project centers on developing an intelligent welding solution that combines AI and digital twin technology with collaborative robots to automate high-precision welding at the level required for nuclear power plant components.
The robot will use cameras to identify welding targets and self-correct its positioning, while learning to replicate the movements of skilled welders to automatically generate welding paths and weaving motions — a technique in which the welding torch is moved side to side to ensure consistent weld width and quality.
Doosan Robotics is targeting a match rate of 95 percent or higher between the robot's welding path and that of a skilled welder, with a path error margin within plus or minus 3 millimeters as measured by a nationally accredited testing body. The company also aims to automate multilayer welding of five or more passes, cutting work time by more than 50 percent.
Validation trials will be conducted at Doosan Enerbility's nuclear energy center. The technology will first be applied to tack welding, preheating and welding processes in nuclear component manufacturing, with plans to expand into shipbuilding, plant construction and defense applications after obtaining nuclear quality certification.
The project runs for 45 months, with Doosan Robotics as the lead institution. DeepX, Sage and the industry-academic cooperation foundation at Changwon National University will serve as co-research institutions, and Doosan Enerbility will participate as the end-user company.
"The structural shortage of workers and the aging of skilled labor on the factory floor cannot be solved simply by adding more automated equipment," said Park In-won, president of Doosan Robotics. "We will combine on-device AI semiconductors with our own collaborative robot platform to develop physical AI collaborative robots that can learn and reproduce even the tacit knowledge of skilled workers, making them a new source of competitiveness for Korean manufacturing."
keg@heraldcorp.com
