In what you can call a form of unnatural selection, the 'mother robot' built by a team of researchers from the University of Cambridge is able to builds evolving baby bots. The self-replicating robot is capable of doing this task by placing its 'baby bot' on the testing platform in order to evaluate whether its traits are worthy to be passed on to the next generation.
It's always been thought that evolution is a trait of life and that robots can't evolve like we can. However, according to reports, researchers have demonstrated that a sort of rudimentary artificial evolution is quite possible. Their research project involves a robot that is able to create new generations of smaller "baby" robots. More than this, the robot evaluates them in order to select the best fit to "live" on in a same way like Mother Nature does.
The study was led at the University of Cambridge by Fumiya Iida. The team of researchers in robotics created a "mother robot" that can assemble smaller robots by gluing pieces together in various configurations. Some of the pieces were equipped with motors. The 'mother robot' observed and evaluated how quickly the resulting "baby bots" could move.
Based on these criteria, the robot disassembled the slow bots and kept the designs of the fastest ones. The robot used this information to create a new "generation". This task was accomplished either integrating the successful designs into less successful ones or by mutating the successful designs to evolve into something even better.
According to Iida, this is a copy of the natural selection process, which is in fact based on reproduction and assessment. The Cambridge research team provided a news release informing about the details of their project. They explained that their 'mother robot' is doing something similar to natural selection and evolution, since the scientists could "actually watch the improvement and diversification of the species."
According to reports, after just a few generations, the new generations of baby robots were really more evolved, since they could run twice as fast as the first batch. This is not the first research of this type, however up to date this kind of studies were usually done virtually, using an artificial intelligence software application able to evaluate thousands or millions of possibilities in a computer simulation. But according to Iida, doing this kind of study in real life by using real robots is the only way to prove that it actually works.
The team of scientists may continue their future work by combining real-life testing with simulation. According to Iila they plan to use nature-inspired techniques to improve the efficiency of the children the open-access journal PLOS One.
