AI-Supported Underwater Vision for Robotic Welding Under Real Port Conditions

In the DATIpilot community project SPorT-UWS, Fraunhofer IGD Rostock is testing the underwater vision system developed in the predecessor project MARIOW for the first time under real operating conditions in the ports of Bremen. The project focuses on a robust camera system for monitoring the welding process, improved calibration methods to increase accuracy and repeatability, and the evaluation of additional sensor technologies for future deployment scenarios.

From Laboratory Prototype to Port Deployment

The predecessor project MARIOW resulted in an initial prototype for semi-automated underwater welding, which was tested under controlled laboratory conditions. The DATIpilot community project SPorT-UWS represents the next step: the complete system will be validated in harbor basin tests in the ports of Bremen under real-world operating conditions.

Poor visibility, water currents, and changing lighting conditions present challenges that could not be fully replicated in the laboratory environment. The insights gained during these field tests will directly contribute to further optimization of the system.

Contribution of Fraunhofer IGD Rostock

Within the project, Fraunhofer IGD Rostock is responsible for testing and optimizing the underwater vision system. The main focus is on a robust stereo camera system for reliable monitoring of the welding process, as well as improved calibration procedures between the camera, welding torch, and robotic arm. The goal is to significantly increase positioning accuracy and repeatability.

In addition, the project investigates whether supplementary sensors such as a CTD probe (Conductivity, Temperature, and Depth), a hydrophone, and an oxygen sensor can be used to derive welding parameters and thereby enable more precise process control in future applications.

Benefits and Expected Outcomes

Testing under real-world conditions will provide a reliable data foundation for all project partners, particularly for autonomous weld path planning and more precise robotic guidance. The experience gained with the camera system, calibration methods, and additional sensors will also lay the groundwork for future projects aimed at implementing AI-based control of the underwater welding process.