Instinct, borrowed from animals, helps the robot to walk
The advanced biped robot can adapt to unfamiliar terrain. The “instinct” helps him in this, which is based on a theory that explains how animals move.
The experimental sample is 23 centimeters in height and is called the Ranbot. This is not just a pair of metal legs, entangled in bulky electric motors and a mass of wires. Ranbot walks and, sometimes, stumbles. The robot walks in a circle and is mounted on a special spar. The experiments are carried out in the laboratory of Florentine Wörgetter at the University of Götingen in Germany.
Ranbot - record holder. He is the fastest two-legged robot in the world.
')
The latest version of the bot can be adapted to an unknown, new for him difficult terrain. The robot uses a "hierarchical control structure." In other words, different levels of control can be triggered when environmental requirements increase.
The lowest level of the hierarchical management structure is called “local chains”. He performs basic robot motion control. At this level, data coming from the hinges is used to make corrections. Then, with the help of special commands, the corrections are transmitted to the hinges.
This simple loop with feedback allows the robot to move steadily, provided that the surface on which it goes is relatively flat.
Trial and error method
The basic control level cannot provide steady movement when an unfamiliar piece of terrain is encountered, for example a climb. This can unbalance the robot and it will fall (as in the video).
In this case, the researchers introduce a higher level control mechanism. When the robot becomes unstable, the infrared sensor gives the command to use another computer. It replaces the mechanism of local chains and makes the robot learn to overcome unknown terrain by trial and error.
This approach is easy to implement and is characterized by the speed of calculations when compared to methods that other robots use. For example, a Honda robot called Asimo must constantly analyze and evaluate the movement of all of its hinges and sensors. He does this to calculate each next step. For such calculations require significant computer power.
Hierarchical control is a theory that explains how humans and animals manage to move on different surfaces with relative ease.
Saving data
The theory of hierarchical control was first proposed in the 30s of the last century by the Russian psychologist Nikolai Bernstein . In animals, local feedback occurs between the muscles, nerves and the spinal cord. The brain performs higher level control.
Mechanisms in the body of animals are much more complex than in the robot Ranbot. Animals have a variety of mechanisms not described by hierarchical control theory, and they play an important role.
Vorögötter believes that sooner or later, robots will learn to imitate living beings more than they do now. For example, a person remembers key movements that he learned before. Unlike humans, modern robots store very little data.
“In the future, robots will be able to constantly store large amounts of data. There are no fundamental reasons for limiting the ability of robots to store data, ”Wörgetter told New Scientist magazine.
The experimental sample is 23 centimeters in height and is called the Ranbot. This is not just a pair of metal legs, entangled in bulky electric motors and a mass of wires. Ranbot walks and, sometimes, stumbles. The robot walks in a circle and is mounted on a special spar. The experiments are carried out in the laboratory of Florentine Wörgetter at the University of Götingen in Germany.
Ranbot - record holder. He is the fastest two-legged robot in the world.
')
The latest version of the bot can be adapted to an unknown, new for him difficult terrain. The robot uses a "hierarchical control structure." In other words, different levels of control can be triggered when environmental requirements increase.
The lowest level of the hierarchical management structure is called “local chains”. He performs basic robot motion control. At this level, data coming from the hinges is used to make corrections. Then, with the help of special commands, the corrections are transmitted to the hinges.
This simple loop with feedback allows the robot to move steadily, provided that the surface on which it goes is relatively flat.
Trial and error method
The basic control level cannot provide steady movement when an unfamiliar piece of terrain is encountered, for example a climb. This can unbalance the robot and it will fall (as in the video).
In this case, the researchers introduce a higher level control mechanism. When the robot becomes unstable, the infrared sensor gives the command to use another computer. It replaces the mechanism of local chains and makes the robot learn to overcome unknown terrain by trial and error.
This approach is easy to implement and is characterized by the speed of calculations when compared to methods that other robots use. For example, a Honda robot called Asimo must constantly analyze and evaluate the movement of all of its hinges and sensors. He does this to calculate each next step. For such calculations require significant computer power.
Hierarchical control is a theory that explains how humans and animals manage to move on different surfaces with relative ease.
Saving data
The theory of hierarchical control was first proposed in the 30s of the last century by the Russian psychologist Nikolai Bernstein . In animals, local feedback occurs between the muscles, nerves and the spinal cord. The brain performs higher level control.
Mechanisms in the body of animals are much more complex than in the robot Ranbot. Animals have a variety of mechanisms not described by hierarchical control theory, and they play an important role.
Vorögötter believes that sooner or later, robots will learn to imitate living beings more than they do now. For example, a person remembers key movements that he learned before. Unlike humans, modern robots store very little data.
“In the future, robots will be able to constantly store large amounts of data. There are no fundamental reasons for limiting the ability of robots to store data, ”Wörgetter told New Scientist magazine.
Source: https://habr.com/ru/post/31163/
All Articles