The Kocevje investment is about 5 billion yen and lifts manipulator capacity to 850 units a month. A separate agentic robot system uses Google DeepMind's Gemini Robotics ER 1.6.
Doosan Robotics wins two Korean Physical AI projects worth KRW 98.9 billion
EditorialBy TrustList Editorial
About KRW 68.1 billion of the budget is government money. One project builds cobots with on-device AI chips, the other an AI welding system for Doosan Enerbility.
About Doosan Robotics wins two Korean Physical AI projects worth KRW 98.9 billion
Doosan Robotics wins two Korean Physical AI projects worth KRW 98.9 billion
6 October 2026: Doosan Robotics has been selected for two government-funded Physical AI research projects with a combined budget of about KRW 98.9 billion, of which about KRW 68.1 billion is public funding. The programmes are run jointly by the Ministry of Trade, Industry and Energy and the Korea Planning & Evaluation Institute of Industrial Technology.
Not yet independently verified. The budgets and technical targets come from Doosan Robotics' own announcement alone. We will update this when it can be confirmed, and remove this note.
Cobots that think on the chip
The first project sits under the K-On-Device AI Semiconductor Technology Development Program and runs for 54 months, with Doosan Robotics leading. It aims at next-generation collaborative robots built on on-device AI chips, so that perception, decision and control all happen locally, without an external PC or the cloud.
The consortium includes Aidin Robotics, Mobilint, the Korea Electronics Technology Institute, the Korea Institute of Industrial Technology and SAIGE, plus university foundations from Korea, Dongguk, Sejong and Yonsei universities.
The planned products are collaborative robots with payloads of 5, 10 and 20 kg, and an industrial humanoid with dual-arm manipulation and omnidirectional mobility. For validation, Doosan will run 72-hour continuous-operation demonstrations at two or more customer sites and pursue international safety certification. Standard products are targeted from 2031.
A welding cobot trained by welders
The second project falls under the Robot Industry Technology Development Program and runs for 45 months. It covers intelligent welding with cobots and AI. Vision locates the weld targets, and imitation learning from skilled welders generates the torch paths and weaving patterns.
Partners are DeepX, SAIGE and the Changwon National University foundation, and the end user is Doosan Enerbility. A national accredited testing institute will verify the targets:
- more than 95 percent similarity to skilled-welder paths
- path error within plus or minus 3 mm
- automated multi-pass welding of five or more layers
- processing time cut by more than 50 percent
The demonstration will take place at the Doosan Enerbility Nuclear Power Center, covering tack welding, preheating and welding of nuclear plant equipment. The company lists shipbuilding and plant engineering as later applications. Doosan Enerbility, the end user, is based in Changwon.
One platform behind both
Doosan says it plans a single integrated Physical AI platform for its cobots, with the two projects feeding it. The programmes run for 54 and 45 months respectively, so both stretch over several years, and standard products from the first one are targeted from 2031.
Inwon Park, President of Doosan Robotics, says: "Labor shortages and the aging of skilled workers in manufacturing are structural challenges".
A second Doosan move in Europe
Separately, Doosan Robotics opened an expanded Frankfurt branch on 19 May 2026. The company says about 200 guests attended, and the branch offers local spare parts and training sessions at least twice a month.
Related on TrustList:
- Yaskawa builds a robot hub in Slovenia and pairs Motoman with Gemini
- P&G takes Siemens-built AI quality inspection to plants worldwide
- Universal Robots launches Gen 7 platform with AI-ready arms
Sources
Categories & features
TrustList Weekly
The week in software and IT, in one email
The news that matters to buyers, new rankings and our own research. Every Thursday, free, and easy to leave.
You may be interested in
Siemens says the Visual Inspection Cockpit is commissioned five to ten times faster than bespoke vision systems. It runs on Industrial Edge with Nvidia GPU industrial PCs.
The Teradyne company redesigns controller, pendant and tool flange around PolyScope X, and adds three g-Series arms; the CB7 Core controller ships at launch, other availability is not given.
Hitachi plants in the Ibaraki region serve as the proving ground, and customer deployments are planned from fiscal 2027. FANUC also announced a Google collaboration in May.
More than 84% of votes cast at Churchill Capital Corp XI backed the merger. Closing is expected on 4 November and trading on 5 November, with about $465 million of gross capital before expenses.
The testbed sits inside Hyundai's Metaplant near Savannah and moves to a building about ten times larger in 2027. Hyundai Motor Group plans 25,000 Atlas units.
The Copenhagen case could produce a ruling covering 17 of the 18 Unified Patent Court states. The Elite Robots dispute ended by confidential agreement with no admission of liability.
The tractor maker's commercial agreement puts Digit on parts sorting and tote handling at Hesston; Agility, whose SPAC listing goes to a shareholder vote on 2 November, says it has nine committed customer sites.
More on TrustList
Everything here links back to the same verified catalogue. Pick your next stop.
- More South KoreaThe ranking for this subject
- CompaniesAgencies, consultancies and IT service providers, ranked by verified reviews.
- ProductsSoftware and SaaS with pricing, features, integrations and alternatives.
- AwardsAnnual recognition decided by verified reviews and an independent jury.
- LaunchesNew products and releases, voted up by the community every day.
- AI ModelsBenchmark scores and community ratings for every major model.
- RequestsBuyers describe what they need; vendors respond directly.
- PeopleReviewers, authors and makers with public profiles.
- ComparePut up to four listings side by side before you shortlist.