Two MSc students from DTU Electrical Engineering have developed simulation tools that show how you can teach underwater robots not to panic.
During an expedition to the Antarctic in 2005, British scientists from the University of Southampton were unfortunate enough to lose their underwater robot underneath the ice. The robot is thought to be trapped 17 kilometres below the ice surface and scientists still do not know exactly what happened. The robot does not know what happened either—which is part of the problem.
Robots are expensive, as are the missions they are sent on. This is why every effort is being made to design robots that are ‘fault tolerant’—i.e. robots that can register a fault and independently assess whether they can continue their mission by compensating for the fault, or whether they should return to the research vessel.
Similarly, a fault-tolerant robot will also be able to avoid rotating on its own axis because one of its propellers is caught in seaweed. Instead it will be able to turn off the affected propeller, performing the mission without it. However, for the robot to carry out something meaningful—also in panic situations—it requires intelligent software that can function autonomously on the basis of the fault reports it receives.
A PhD project by Thomas Falkenberg and René Tavs Gregersen from DTU Electrical Engineering is playing an important role in this process. They are looking at the software requirements for a fault-tolerant robot—among other things using data from the underwater robot Minerva, owned by the Norwegian University of Science and Technology.
“It’s one thing building lots of extra propellers you can activate in an emergency, but you also need software that interprets the fault and tells the robot what to do. Out in the open sea, the robot will simply experience what’s known as a blow-out and return to the surface—but when you’re operating underneath the ice, that strategy is no longer an option,” says Thomas Falkenberg.
Indeed, ice will be the great challenge of the future, as oil exploration gets underway in the Antarctic. With billions of kroner at stake, robots have a key role to play, says René Tavs Gregersen: “If a robot dives to a depth of 200 metres and breaks down, it will have to return to the vessel, delaying the entire mission for several hours,” he explains.
“The ability to assess whether to continue with a minor fault will therefore ensure maximum utilization of resources. Conversely, in the event of serious or catastrophic failure, the robot can minimize loss, operating in a kind of safe mode to prevent inappropriate action several miles below the ice.”