MUNICH, Germany (Reuters) - For your average robot, chess may be easy compared to simple everyday activities like frying a sausage, but that may be about to change thanks to software being developed by scientists in Germany.
Researchers at the Technical University in Munich are using complex algorithms to teach robots to learn from their mistakes and perform routines humans can do in their sleep.
The aim is to instill perception, manipulation and reasoning capabilities, enabling the machines to learn from their own experience and taking them a step closer to developing the self-awareness that so far has been reserved for humankind.
That process “demands that ... they can foresee the consequences of their actions before they actually jump into action,” Michael Beetz, professor of computer science at the university, told Reuters.
“We can say that the robot is supposed to push a spatula under a pancake without damaging the pancake.”
Calculation models and controlling mechanisms programed into the hardware help robot TUM-Rosie understand the nature and function of the cooking spoon it is handling, while TUM-James uses real-time sensing to do things like slice bread.
Reporting by Marcus Nagle; Writing by Kalina Oroschakoff; Editing by John Stonestreet