Robots know to 'give birth' and evolve themselves

Information may sound like science fiction, but researchers have actually created robots that are capable of spawning and evolving without human intervention.

Successfully create "mother" robots that "spawn" themselves and evolve

Robot experts have built a "mother" machine that can produce its own "children" and test which sub-robots best perform certain tasks. The "mother" robot then uses these results to declare the design of the next " child ", so that the preeminent characteristics are passed on from one generation to the next, similar to the way of natural selection process in animals.

The robot that gives birth to babies is the creative product of a team of engineers at Cambridge University, led by Dr. Fumiya Iida. Dr. Iida and colleagues designed the " mother " robot and programmed it to be able to build a simple, movable robot.

From this point, no need for any human intervention or computer simulation, the " mother" robot produces its own composed of 1 - 5 plastic blocks and a small side motor. in.

In each separate experiment, the mother robot designed, built and tested a generation of 10 children, using the information gathered from this generation to declare the design for the next generation. As a final result, superior features have been passed down through generations of robots and the best robots in the last generation perform a predetermined task twice as fast as the best robots in the first generation. .

"Natural selection is basically reproduction, evaluation, reproduction, evaluation and repeating that process. Overall, this is what the mother robot is doing. We can really observe. The improvement and diversification of species, ' explains Dr. Iida.

According to Dr. Iida, every conceived robot comes from a unique "geology", a combination of 1-3 "ge n" (essentially plastic blocks) that contains all the information information about shape, structure and movement commands.

Just like in nature, evolution in robots takes place through a "mutation process" in which components of a mutated gene or single genes are added or subtracted and " bred " (a new gene is formed through mixing genes from two individuals).

In order for the mother robot to determine which child is the best, each baby robot is tested for how far it can move from the starting position, for a certain period of time. The most successful individuals in each generation remain the same in the next generation to preserve their capabilities, while mutations and breeding are carried out in less successful offspring.

The team discovered that it helps to create design variation and improve the ability of robot generations over time.

Their research focuses on how robots can be improved, such as intelligence or movement, through learning from nature. A robot now requires 10 - 100 times more energy than an animal when doing the same thing.

" There is still a long way to go before we build robots that look, act and think like humans. However, what we have is a lot of feasible technologies, helping them. I put some aspects of biology into the technical world, "Iida said.