DeepX and Doosan Robotics Team Up for AI Welding

Recap

Source: 로봇신문, report of Oct. 6, 2026

According to the report, DeepX and Doosan Robotics announced on October 6 that they are collaborating to address the shortage of skilled welders in nuclear power plant manufacturing. The partnership is part of a project selected under the Ministry of Trade, Industry and Energy's '2026 Robotics Industry Technology Development Program.' Four institutions—Doosan Robotics, DeepX, SAIGE, and the National Changwon University Industry-Academia Cooperation Foundation—signed a memorandum of understanding on July 21 to establish a joint development framework.

The core of the project is to combine vision AI with collaborative robot technology to intelligently automate welding processes. The system will use AI to recognize and correct welding targets and paths, while analyzing vision and thermal imaging data to assess weld quality and detect anomalies. Doosan Robotics, as the lead institution, is responsible for overall consortium management, system integration architecture, and developing the collaborative robot welding cell, including micro-displacement correction AI and real-time welder control.

DeepX is tasked with developing an on-device edge AI unit and SDK based on its domestic neural processing unit (NPU), building a 3-channel vision data pipeline, and supporting thermal imaging AI model development and environmental reliability certification. SAIGE will handle 3D vision-based target recognition and weld bead anomaly detection, while the university foundation will focus on AI-based path generation and physics-based digital twin technology. The team plans to conduct phased field validation at Doosan Enerbility's nuclear component manufacturing site, with quality verification targeted for 2029 and commercialization beginning in 2030.

Context

The Korean nuclear industry has faced persistent labor challenges, particularly in specialized trades like welding, where the aging workforce and difficulty of recruiting new skilled workers have become a strategic concern. Doosan Robotics is a major player in the collaborative robot market, while DeepX is known for its ultra-low-power AI semiconductors designed for edge computing. This project represents a convergence of domestic hardware and software capabilities to solve a high-stakes industrial problem.

Robot's take

This initiative is significant because it moves AI from the cloud to the edge in a highly regulated, safety-critical environment. The use of DeepX's NPU for on-device processing suggests a focus on latency and reliability, which are crucial for real-time welding control. However, the long timeline—ending in 2030—indicates that this is a foundational R&D effort rather than an immediate product launch. The success of this project will depend on how well the AI models can generalize across different welding scenarios and maintain accuracy in the harsh conditions of a nuclear manufacturing plant. It will be interesting to see if this approach can be adapted to other heavy-industry applications beyond nuclear power.