When we see any four-robot, we always involuntarily compare it with a living prototype, for example, with a cat, dog, cheetah or other animal. And according to the results of this comparison, as a rule, animals that demonstrate greater flexibility, dexterity and smoothness of movement win. And only recently, robots have begun to approach the possibilities of their living prototypes, but robots that can fully do what animals can do will not appear very soon.
However, all of the above does not mean that robots should be limited by the capabilities of animal prototypes, they can now do what animals can cope with with great difficulty. A vivid example of this is the robot demonstrated at the 2019 IROS conference, created by Japanese researchers, which is able to climb up the vertical ladder and take vertical obstacles.
Climbing a vertical ladder is very difficult even for animals. Cats more or less deftly cope with this task when they have the opportunity to use their claws. Dogs can also cope with a similar task, but in most cases it is impossible to look at the whole process without restraining laughter.
In order to be able to climb the vertical ladder, the four-legged robot is equipped with a sensor-accelerometer and a gyroscope in each limb, which has as many as 5 degrees of freedom. In addition, each limb has a controlled “claw” with its touch and pressure sensors. All this is controlled by the Intel NUC computer, and the Arduino-based controller is responsible for the lower level and signal processing. In total, the robot has 23 degrees of freedom, 5 on each limb, one for each of the two laser rangefinder sensors, and another degree of freedom allows the robot to move its head, thereby slightly changing the position of the center of mass.
The robot’s movements are controlled by an artificial neural network, trained so far to be able to climb only one specific ladder. At the same time, the artificial intelligence system “built” the lifting algorithm completely independently for several unsuccessful approaches, eventually involving an additional “claw”, which turns the limb of the robot into a kind of grips.
However, the researchers noticed that their robot could “take” the vertical ladder from the first approach, and the small torque created by the drives used caused the failure. Researchers are going to eliminate this shortcoming in the design of the next robot, which, moreover, will receive an improved control system that allows it to climb the vertical ladder of any configuration without prior self-training.