The UK鈥檚 Royal Navy wants to use artificial intelligence to guide a fleet of robot submarines, plotting the best course to find and destroy underwater mines.
Mines are a cheap and effective form of naval warfare. Laid by aircraft, ships or submarines, they lurk in shallow water until a vessel passes nearby and triggers a magnetic sensor, setting off the explosive.
Ensuring waterways are safe requires regular sweeps. At present, the UK searches for mines in the Gulf using crewed vessels. But from next year, these will be supplemented by uncrewed boats and small robot submarines.
Advertisement
Planning minesweeping routes is difficult because there are so many interacting variables. Tides and currents, the weather, seabed contours, the number of wrecks and other metallic objects that may confuse the search all influence how long it will take to survey a given area.
So, the Royal Navy and data analytics firm Envitia are developing a system called Route Survey Tasking & Analytics (RSTA) to help for mine-hunting vessels to search more efficiently. RSTA will use artificial intelligence to take all the variables into account and generate search plans rapidly without human effort.
While current systems just record possible mines as dots, with RSTA the entire area will be mapped with high-resolution sonar images. This means it will be easy to tell from a distance if an area contains a new potentially dangerous object.
The Royal Navy is testing robotic submarines including the Remus 600. These are torpedo-like craft 3 metres long able to operate independently for up to 24 hours on battery power. They locate mines on the seabed with sonar. Each one can search several square kilometres a day, depending on conditions.
Once a mine is located, the robot submarines can destroy it using a directed explosive charge. Mines may also be 鈥渆xploited鈥� 鈥� opened up and inspected.
RSTA will attempt to find the most efficient minesweeping routes, bypassing difficult areas so that a clear channel can be found quickly and sea-going traffic isn鈥檛 interrupted. Each mine-clearing trip will be fed back into the system allowing it to constantly improve via machine learning.
鈥淪trides in AI-enabled data processing are making it increasingly possible to control large numbers of unmanned vehicles and sort critical data on the locations of mines from false positives,鈥� says Sidharth Kaushal at defence think tank RUSI.
RSTA should be in service by 2022. The project is part of the UK Ministry of Defence鈥檚 鈥淗uman-machine teaming鈥� plans, in which increasing amounts of thinking will be carried out by machines.
The next application is likely to be in anti-submarine warfare, says Mark Atkinson of the Royal Navy, but it could spread to many other areas. These could include cyber-operations, controlling swarms of unmanned aircraft and helping to improve strategies, he says.
