Genetic Algorithms Design and Manufacture Robots Without Human Intervention

Genetic Robots

Consisting of cylindrical tubes and ball-and-socket joints, Fraunhofer's Genetic Robots can be constructed by other robots to perform tasks based on a range of external factors.Fraunhofer IPA

In sci-fi lore, one of the great qualifying events leading up to the eventual war with and enslavement by our machines is the moment when robots begin replicating – that is, they begin manufacturing themselves without help from humans. If that's the case, then the latest news out of the Fraunhofer Institute should be particularly discomforting. Researchers there have created so-called genetic robots that are created fully automatically from a genetic software algorithm and a 3-D printer, no human intervention necessary.

The notion of genetic robots certainly isn't new, but the Fraunhofer team has reached a milestone by creating a computer algorithm that can take into account environmental factors, physical laws, the task at hand, and other external characteristics to design -– from scratch – a robot for the job. Using additive manufacturing (3-D printing), the algorithm can design a multitude of possible robots for a job and select the one it thinks is best, all from a few inputs either from a human or from another computer program.

Right now, those inputs and the robots are fairly simple – "build a robot that can efficiently move across this level surface," for instance – resulting in 'bots consisting of cylinder-shaped tubes, ball-and-socket-joints, and tiny actuators that drive them. But as they become more sophisticated, it's not a stretch to think that factories, homes, and (dare we say it?) militaries might be able to design task-specific, on demand robots.

So what's so "genetic" about these robots? The software takes into account the kinds of tasks and environmental factors a particular robot will face – like whether it will travel over solid ground, swim in water, climb stairs, etc. – but because it can design several possible outcomes from a given set of inputs the results vary, much as they do in the biological world. "The algorithm often spits out surprising variations – 'mutations' that would not necessarily have occurred to the designer," says Fraunhofer industrial designer and product developer Andreas Fischer in a press release.

So these 'bots can mutate over time, traverse land, AND swim in the water? Clearly the Fraunhofer team hasn't seen Terminator: Salvation.