CleanSeas

New robotic manipulation techniques and AI algorithms for the precise handling of objects in the sea using the example of ammunition dumps / UXO

Approximately 1.6 million tonnes of ammunition, mainly from the First and Second World Wars, are still lying in the North and Baltic Seas, the majority of which is conventional explosive and incendiary ammunition consisting of TNT or white phosphorus, but also a significant proportion of old chemical ammunition with warfare agents such as mustard gas/LOST, tabun, phosgene or arsenic-containing substances such as Clark I/II or adamsite. The advancing corrosion of these unrecovered contaminated sites poses an increasing danger to the ecosystem, for example through increasing toxic concentrations and correspondingly increasing exposure of the fish living there, but also through contact of fishermen or bathers with washed-up phosphorus clumps. The warfare agents are found in shipwrecks, in sea mines, in aerial bomb unexploded ordnance, but above all in munitions dumped by the Allied forces after the end of the war.

In the meantime, the consensus among the littoral states is that the long-standing strategy of leaving the contaminated sites at the bottom of the sea and letting them decompose over time is not a viable option. The current state of decay of many contaminated sites makes the problem increasingly acute. The Baltic Sea Parliamentary Conference (BSPC) rapporteur on contaminated munitions puts it this way: "Once the shells are completely gone, "it becomes virtually impossible to detect and recover". Previously often understood by riparian states as a military problem, it is increasingly understood as a task for society as a whole; concrete demands for testing environmentally compatible salvage procedures, also by various NGOs, are on the table; the problem is an explicit part of the coalition agreement of the governing parties. The main problems of the solutions presented so far are the enormous costs and the lack of qualified personnel (usually salvage divers).

The unresolved problem of contaminated ammunition sites has concrete negative effects, not only on the immediate dangers but also indirectly through the delay of projects in the area of construction and connection of offshore wind farms or other necessary infrastructures and thus causes enormous costs. These include delays in the expansion of offshore wind energy or in the area of green hydrogen, but also the threat to fish stocks and fisheries as well as the further use of maritime resources.

The following concrete objectives of the project are thus derived:

• the development of AI methods for learning motion and sensor models for precise navigation in the close-up range of critical structures
• AI-based 3D reconstruction of objects from sonar and camera data using generative networks to prepare object interaction
• Whole-body control for (semi-)autonomous contact sampling and manipulation of critical structures such as ammunition dumps.

Implementation of the above approaches in a demonstrator and evaluation of the system capabilities in the salt water test basin of the DFKI RIC. In addition to the aforementioned scenario of salvaging munitions waste, the approaches, technologies and robotic systems to be developed are highly relevant for the energy transition (maintenance of offshore wind farms and maritime hydrogen infrastructures), the more environmentally friendly maintenance of offshore oil and gas installations, as well as for the growing field of marine food production in aquaculture farms. Furthermore, the approaches are suitable for decisively improving the inspection and maintenance of critical infrastructures such as port facilities, dams or water-cooled power plants.