OREANDA-NEWS.Thus, engineers have demonstrated that such robots can be repaired even in poor conditions. A group of researchers from Japan built a robot, instead of limbs it had pieces of branches. Moreover, there were only three such branches, and their mobility was severely limited, but thanks to artificial neural networks and deep learning algorithms the robot still learned to crawl. 

Instead of limbs, a new robot received three branches of irregular shape, and then the neural network evaluated various movements in terms of their effectiveness: the farther it was to crawl, the better. Over time, the structure assembled from sticks and servos began to creep in a given direction, and without human intervention.