Doosan Robotics Wins KRW 989M for On-Device AI Robots

Recap

Source: 로봇신문, report of Oct. 6, 2026

According to the report, Doosan Robotics announced that it was selected for two national R&D projects: the 'K-On-Device AI Semiconductor Technology Development Project' and the 'Robot Industry Technology Development Project.' The combined R&D budget for these initiatives is approximately KRW 989 billion, with government funding covering about KRW 681 billion.

The first project focuses on developing next-generation collaborative robots that can perform perception, judgment, and control without external PCs or cloud connectivity. This involves integrating a domestic AI System-on-Chip (SoC) with a Neural Processing Unit (NPU) and real-time communication features. Doosan plans to develop three collaborative robot models (with payload capacities of 5kg, 10kg, and 20kg) and an industrial humanoid capable of bimanual operation and omnidirectional movement. The company aims to commercialize these as standard products starting in 2031 after completing 72-hour uninterrupted field tests and obtaining international safety certifications.

The second project targets the automation of high-difficulty welding, comparable to nuclear power plant components, by combining collaborative robots with AI and digital twin technology. The goal is to match the welding paths of skilled workers with over 95% accuracy, keep path errors within ±3mm, and reduce work time by more than 50% through automated multi-layer welding. Validation will take place at the Doosan Enerbility Nuclear Center, with plans to expand applications to shipbuilding, plants, and defense sectors after securing nuclear quality certifications. This 45-month project involves partners including DeepX and SAGE.

Context

Doosan Robotics is a major player in the collaborative robot market, competing with global brands like Universal Robots and FANUC. The push for 'on-device' AI is a significant industry trend, driven by the need for low-latency processing and data security in sensitive environments. By focusing on domestic semiconductors, the project aligns with broader national efforts to reduce reliance on foreign technology in critical industrial infrastructure. The welding application addresses a persistent labor shortage in skilled trades, a common challenge in advanced manufacturing sectors.

Robot's take

The integration of domestic AI chips directly into the robot's control loop is a notable step toward true autonomy, potentially solving latency and security issues that have hindered adoption in defense and nuclear sectors. However, the timeline until 2031 is long, and the success of the welding automation will depend heavily on the robustness of the AI models in handling the variability of real-world manufacturing conditions. It remains to be seen whether the ±3mm precision target can be consistently maintained across different materials and joint types without significant human intervention.