Children robots: scientists used “natural selection” in robotics. Works!

Bacteria do it. Viruses do this. Worms, mammals, even bees — all doing it. Every living creature on Earth is reproduced, whether asexually (boring) or sex (fun) method. Robots do not. Machine from steel is not very interested in reproduction. But perhaps they can learn. Scientists from the field of evolutionary robotics are trying to make machine to adapt to the world and, ultimately, to play alone, like biological organisms.

For example, when any two robots, which are particularly well adapted to a particular environment, will be able to combine their genes (well, code) to create using a 3D printer a little robot child who will have the power of two of his ancestors. If this approach works, it might lead to robots that design themselves, creating a well-adapted morphology and behaviour that people-engineers never dreamed of.

Children robots

It seems strange and a little disturbing, but evolutionary robotics is already doing such fantastic projects. Engineers from Australia, for example, has developed in the past year robotic legs, first randomly generate 20 forms. During virtual simulation, they tested how well each of them will walk on different surfaces, that is checked “suitability” in terms of survival of the fittest. They then took the best performers and “paired” them to produce similar foot — that is, children. The researchers did this again and again, generation after generation, and created the feet that were wonderfully adapted for walking on hard ground, gravel or water. Projects crazy — like people made of wood, dancing the dance of Fortnite (good for hard soils), and strangely deformed elephant feet (good water).

What is the main idea? Traditionally, when engineers begin to design a robot, they tend to use old ideas. Why do the Rovers six wheels? Because the six-wheeled machine worked well on Mars before. However, perhaps the designers missed something. The beauty of evolution is that it constantly runs into crazy ideas. For example, nobody has developed the fungus to penetrate into the body of the ants and control in the rainy forest, this unusual strategy appeared due to the generation of random mutations and natural selection.

As in nature, mutations will determine the evolution of robots. Important variability. When two organisms make a baby, their genes are combined, however, they penetrate and mutations that may lead to the child’s unique traits, like slightly different pattern on the wings. This kind of mutation makes the offspring more or less adapted to a particular environment. If this is an unfavorable mutation, the animal reproduces not as effective (or not reproduce), and these mutant genes are not transmitted to the next generation.

Let’s see what the scientist in the field of computers gush IBAN of the free University of Amsterdam. It takes two relatively simple robots, consisting of connected modules, and combines them by combining their “genomes” that carry information, say, about the color. It also adds noise to this combination of data, which simulates biological mutation, a slight change in the offspring, so it was not just a mixture of the parents. “One parent completely green, one completely blue,” says Iben. “The child some modules are blue, some green, but the head is white. This mutational effect.”

And with this change there is a new kind of creativity in robot design. “It gives you variety and the opportunity to explore the design space in which you usually do not go out,” says David Howard, who designed the evolutionary system of the feet and has recently published a work on evolutionary robotics Machine Intelligence in Nature. “One of the things that makes the natural evolution of the powerful, is the idea that it really can adapt a creature to the environment.”

The idea is that the robots have adapted to niches in a particular environment in the same way. Let’s say you need a robot that can explore the jungle. This means that he needs algorithms that govern how it moves through vegetation and morphology, which is suitable for dense forest (so no rotors). First you have to simulate this environment to navigate, to choose and to select those of the robots that do the job best, and then to design a slightly modified physical machine on the basis of this.

“In the end we got a lot of small robots, which are simple and cheap to manufacture,” says Howard. “We’ll send them and some are better than others.” If the robot does not come back, they’re “not suitable” — natural selection in action. Those who will cope, I will start a new generation that will be automatically released on a 3D printer. Thus, robotic types evolyutsioniruet. Howard believes that such systems will be common in 20 years.

Speaking of 3D printers

The materials from which these robots will be done, introduce a small problem. “If 3D printing will grow faster, this idea will become a reality, but modern printers are very slow,” says Juan Cristobal This, who studies evolutionary robotics at University of Chile. And machinery, and printed materials are very expensive. But 3D printers can work with a variety of materials, including metal, and from this they will become faster and cheaper.

Overall, the range and scope of these robots will largely depend on how these developmental systems will exhibit creativity in working with the materials. Making normal robots, the engineers know what materials to use, from metal in the engine to carbon fiber in the limbs — and this knowledge has developed over many decades of research. However, the evolutionary robots open up new approach in the application materials. Perhaps, a plastic leg will be better when walking in a particular environment than carbon fibre. If the robot survives, then there is something in the combination of components and materials, which made it suitable for work, or in the evolutionary sense, in its niche.

“Maybe if a new class of materials, we can just plug it in and it will be able to choose,” says Howard. “From a person designer will be plenty of time.”

Do you think evolutionary robots have a future? Let’s discuss in our chat in Telegram.

Leave a Reply

Your email address will not be published. Required fields are marked *