The Economist –
For years navies have employed human divers, dolphins and sea lions to search for explosives attached to the hulls of warships by a scuba-diving enemy.
While these mine-finding tactics work, they are less than ideal.
Divers can be killed or injured and marine mammals are extremely costly to maintain on a boat.
The mines are also getting smaller and harder to detect.
The idea of using aquatic robots to search for the mines instead is alluring, but it is difficult to teach machines how to navigate around hulls without crashing into them or getting lost.
Franz Hover and Brendan Englot at the Massachusetts Institute of Technology (MIT) have come up with a way to improve things by using a two-step process.
Programming robots to scan hulls would have been relatively easy if limpet mines were still as large as watermelons.
The robot would simply be told to maintain a safe working distance and swim back and forth using its sonar cameras to generate an image of the hull’s topography.
If the generated image perfectly matched that of a clean hull stored in the robot’s memory, the robot would know that there were no mines attached.
If not it could raise an alarm. Yet nowadays mines can be as tiny as a small iPod, which might not blast a hole in the hull but if carefully placed could disable a ship’s propellers.
The problem, then, is one of definition.
Sonar scans done at a safe distance of 10 metres create a rough image known as a data-point cloud. But this lacks the detail to spot small explosive charges.
The addition of vision cameras may not do much to assist since harbour waters are often murky.
To work around these problems, Dr Hover theorised that the data-point cloud could be used not to spot mines but as a guide to help the robot take a closer look at the hull.