The maintenance of deep-sea installations such as oil and gas pipelines, telecommunications cables or offshore wind turbines is extremely complex and cost-intensive. If defects occur, usually humans dive down into the depths to carry out repairs on the installations. To minimize the risks for divers and prevent damage, robots are increasingly being used to monitor the condition of the plants and detect weak points at an early stage. So far, they cannot manage without human help: the mostly autonomous vehicles are accompanied by large supply ships that take care of launching and recovering the systems before and after the mission. However, the availability of suitable ships is not always guaranteed, especially in urgent operational and emergency situations.
The focus of the project CIAM ("Cooperative Development of a Comprehensive, Integrated Autonomous Solution for Underwater Monitoring"), which started on May 1, 2021, is on the development of a new generation of unmanned submersible robotic systems controlled by artificial intelligence. Thanks to intelligent energy management, these systems should be able to achieve particularly long mission durations and high ranges of up to 500 kilometers to carry out coast-to-coast operations entirely without human support. The BMWi is funding the project as part of the Maritime Research Program, which is intended to strengthen the innovative power and competitiveness of the maritime sector in Germany and at the same time promote climate and environmental protection.
With its research departments Robotics Innovation Center and Plan-Based Robot Control, DFKI is part of the project consortium led by ROSEN Technology and Research Center GmbH, which consists of nine partners from industry and research. Within the joint project the partners will design, build and test two innovative AUVs that are deep-sea capable, yet relatively lightweight, and feature unprecedented autonomy. The Robotics Innovation Center, which has been developing intelligent hardware and software solutions for AUVs for many years, is playing a key role in the realization of vehicle autonomy: this includes front-seat navigation (low-level control), which combines all the information collected by the vehicle sensors to derive the AUV's current position, course, and speed, and adapts these to the environmental conditions detected, such as the current situation.
In addition, the Bremen research department is contributing its expertise to the development of a localization algorithm for back-seat navigation (high-level control) that enables the AUV to be positioned with minimal energy consumption. For this purpose, various system components can be switched on or off depending on the navigation quality during certain mission sections. For example, the robot can navigate for a certain time using only its internal compass and an inertial measurement unit. As soon as it deviates too much from the desired course, additional sensors are automatically activated to provide more precise navigation data. The sensors are switched on and off by an intelligent power distribution system that measures the power currently flowing and deactivates individual components if a limit value is exceeded. In this way, the energy consumption of the entire system can be considerably reduced, allowing for significantly longer operation times. Deployment and retrieval of the AUV from a ship, which is thus required much less frequently, is realized via a docking station. For this, the Robotics Innovation Center contributes an intelligent algorithm that enables the system to dock autonomously.
To safely navigate underwater and autonomously inspect underwater infrastructures, robots must not only comprehensively perceive their environment sensorially, but also be able to interpret it. The DFKI research department Plan-Based Robot Control is responsible for the implementation of semantic environment representation in CIAM. This is based on the one hand on inspection data from inside a pipeline, and on the other hand on the environmental information acquired during an external inspection of the same pipeline. In these data, the Osnabrück researchers search for mission-relevant features and objects that are recognizable from both the outside and the inside, e.g., branches or sacrificial anodes, and thus annotate the inspection data with semantic knowledge. The resulting semantic maps are then merged into a consistent, overall georeferenced map that can support mission planning and AUV localization as well as aid in object identification.
The technologies developed in the project will initially be tested under controllable laboratory conditions. For this purpose, the partners have various test basins at their disposal in the DFKI's Maritime Exploration Hall in Bremen: a 20 m³ water basin, which can be darkened and clouded to create deep-sea-like conditions, is particularly suitable for testing imaging sensors. Navigation and autonomous docking can be tested in a 3400 m³ saltwater test basin, which has a water depth of 8 meters. In order to take into account aspects such as currents, solar radiation, interaction with flora and fauna, as well as greater travel distances and diving depths, the systems will subsequently also be put to the test outside the laboratories in realistic environments.
CIAM project partners:
- ROSEN Technology and Research Center GmbH – coordinator
- FormLED GmbH
- balticFuelCells GmbH
- German Research Center for Artificial Intelligence GmbH (DFKI)
- Berlin University of Technology
- HafenCity University Hamburg
- University Hospital Freiburg