Without a doubt, four-legged robotsmoving on the ground is much more confident than the two-legs. They can easily get up after a fall, while a stumbled robot with two legs can fall and wallow on the floor for several minutes trying to get up and stand up again. Researchers from Switzerland, who developed the four-legged robot dog ANYmal, figured out how to make it better. Instead of teaching him new movements in reality, they allowed him to study on his own inside a computer simulation.
The result of the new approach to learning is impressive: the robot began to move 25% faster, and if it is strongly pushed, it easily gets to its feet, maintains balance, or completely rebounds from impact. ANYmal has learned all this in a computer simulation with different situations that he had to cope with by trial and error. According to Jemin Hwanbo, ETH Zurich employee, the robot consists of many mechanisms and faces countless situations, so it would be almost impossible to develop an algorithm for each of them manually.
To create a virtual space forWorkouts required a regular computer. The training took place thousands of times faster than in the real world, so the developers saved a lot of time and effort without straining the programmers. The researchers noted that computer modeling can be used not only in the case of ANYmal, but also with other four-legged robots.
According to Professor Chris Atkeson fromCarnegie Mellon University, the ETH Zurich team made a great contribution to robotics - reduced the cost of programming, which is highly valued in this field.
When this robot becomes very smart, it will get angry. The video recording where he is being kicked is already ready, so he will be the first to come to us when the robots revolution begins - the professor joked.
Many probably wondering - why do we needfour-legged robots? The answer was given in November 2018 - they can be very useful in places where ordinary people would simply refuse to work. One of these places is the sea power station in the North Sea, where automation is very boring and much work.
What do you think, than such robots can be dangerous to humans? Share your bold assumptions in the comments, and you can discuss the news in our Telegram-chat.