Robolution Joins National Shipbuilding AI Welding Project

Recap

Source: 로봇신문, report of Oct. 6, 2026

According to the report, Robolution (CEO Yoon Kyung-seok) has joined a national physical AI research and development project focused on autonomous welding in ship manufacturing. The company will serve as a joint research institution for a specific task under the Ministry of Science and ICT's initiative to develop and globally validate Large Action Models (LAM) for human-AI collaboration.

The project, titled 'Fusion Data Collection and Validation Technology for Curved Outer Plate Structural Welding,' runs from August 2026 to December 2030. It has a total budget of 17.5 billion won, with 10.5 billion won provided as government R&D funding. Neuromeka is the lead institution, coordinating with academic and industrial partners to develop autonomous welding technologies for curved blocks.

Robolution’s role involves building the physical system that executes AI decisions. The company will develop its own Welding Procedure Specifications (WPS), create test specimens and jigs, and integrate gantries, robots, welders, and sensors into a unified autonomous welding system. The goal is to enable robots to explore weld lines, determine paths and conditions, and self-correct in response to real-time changes in the manufacturing environment. Key performance targets include tracking curved shapes within ±2mm, improving weld quality by over 10%, and achieving a 90% autonomous welding success rate. The project also aims to accumulate over 100,000 synthetic data points for training.

Context

Shipbuilding remains a critical sector for South Korea, but the industry faces labor shortages and the need for higher precision in complex curved structures. Traditional automated welding often relies on pre-programmed paths that struggle to adapt to the thermal deformation and assembly errors inherent in large steel blocks. This project represents a shift toward 'physical AI,' where AI models do not just analyze data but directly drive the physical movements of industrial equipment. Neuromeka, Robolution’s parent company, has been a prominent player in the Korean robotics scene, known for its humanoid and industrial robot developments. This collaboration leverages Robolution’s expertise in industrial and collaborative robot manufacturing to bridge the gap between AI software and hardware execution.

Robot's take

This project is significant because it moves beyond simulation to validate AI-driven control in a high-stakes, real-world industrial setting. The challenge of adapting to 'thermal deformation' and 'assembly errors' in curved blocks is a classic problem in robotics that often plagues pre-programmed automation. If Robolution can successfully integrate the 'cognition-judgment-action' loop in a gantry-robot system, it could set a new standard for autonomous manufacturing. However, the four-year timeline and the complexity of integrating multiple subsystems (gantry, robot, sensors) present substantial engineering hurdles. The success of this project will depend heavily on the robustness of the AI model in handling the noisy, dynamic environment of a shipyard, which is far less controlled than a typical factory floor.