Robot simulates laboratory rat behavior
The behavior of the robot, controlled by a special program, surprisingly resembles the behavior of rodents. The program simulates a part of the rat's brain and is used in exactly the same classical experiments as those that are placed on live rodents.
The software that controls the robot is borrowed from nature and models a part of the brain called the hippocampus. This part of the brain helps the rat to map the environment and is activated when the rat is in a familiar environment.
Alfredo Witzenfeld (Alfredo Weitzenfeld), a specialist in robotics at the Technological Institute of Mexico City, carried out work on the reprogramming of the robot dog AIBO, produced by the Japanese company Sony.
A robot placed in a labyrinth learns to find a way to “reward” in such a way that strikingly resembles the behavior of real living rodents. In doing so, he uses landmarks for navigation.
Rat in the maze
Witzenfeld discovered that the robot could recognize the places he had visited before, distinguish similar parts of the maze from each other, and also approximately determine their location by being placed in an unfamiliar part of the maze. All these actions are able to perform the robot after a single training session.
“The uniqueness of our work lies in the fact that we are trying to reproduce on robots the experiments that were carried out with laboratory rats,” Witzenfeld said in an interview with the magazine New Scientist.
The task of the scientists was to repeat with the robots the classical experiments with the water maze , carried out by Richard Maurice in 1980. These experiments were designed to shed light on the solution of spatial problems using neurology.
Witzenfeld collaborates with scientists, neurobiologists, who experiment with real living laboratory rats. He said: “Our goal is to expand and improve existing models by testing new hypotheses on robots. We plan to conduct relevant experiments on live rodents in order to understand the mechanisms of spatial memory and learning. ”
New approaches and perspectives
One of the important tasks in the navigation of robots is to teach machines to create maps of their surrounding space, and to determine their location (Simultaneous Localization And Mapping - SLAM).
“We believe that our experiments in the future will lead to new discoveries in the field of SLAM and robot training,” added Vitsenfeld.
Chris Melhuish, director of the laboratory of robots (Bristol Robotics Laboratory) noted that while other researchers conduct their experiments by computer modeling, experiments with robots are carried out in a real environment. This can be of great value when creating more reliable software for robots.
Witzenfeld agreed: “Such an approach complicates our task, but gives a better understanding of the complexity that we will encounter in real and artificial systems.”
')
There is also a translation on the Membrane website .
The software that controls the robot is borrowed from nature and models a part of the brain called the hippocampus. This part of the brain helps the rat to map the environment and is activated when the rat is in a familiar environment.
Alfredo Witzenfeld (Alfredo Weitzenfeld), a specialist in robotics at the Technological Institute of Mexico City, carried out work on the reprogramming of the robot dog AIBO, produced by the Japanese company Sony.
A robot placed in a labyrinth learns to find a way to “reward” in such a way that strikingly resembles the behavior of real living rodents. In doing so, he uses landmarks for navigation.
Rat in the maze
Witzenfeld discovered that the robot could recognize the places he had visited before, distinguish similar parts of the maze from each other, and also approximately determine their location by being placed in an unfamiliar part of the maze. All these actions are able to perform the robot after a single training session.
“The uniqueness of our work lies in the fact that we are trying to reproduce on robots the experiments that were carried out with laboratory rats,” Witzenfeld said in an interview with the magazine New Scientist.
The task of the scientists was to repeat with the robots the classical experiments with the water maze , carried out by Richard Maurice in 1980. These experiments were designed to shed light on the solution of spatial problems using neurology.
Witzenfeld collaborates with scientists, neurobiologists, who experiment with real living laboratory rats. He said: “Our goal is to expand and improve existing models by testing new hypotheses on robots. We plan to conduct relevant experiments on live rodents in order to understand the mechanisms of spatial memory and learning. ”
New approaches and perspectives
One of the important tasks in the navigation of robots is to teach machines to create maps of their surrounding space, and to determine their location (Simultaneous Localization And Mapping - SLAM).
“We believe that our experiments in the future will lead to new discoveries in the field of SLAM and robot training,” added Vitsenfeld.
Chris Melhuish, director of the laboratory of robots (Bristol Robotics Laboratory) noted that while other researchers conduct their experiments by computer modeling, experiments with robots are carried out in a real environment. This can be of great value when creating more reliable software for robots.
Witzenfeld agreed: “Such an approach complicates our task, but gives a better understanding of the complexity that we will encounter in real and artificial systems.”
')
There is also a translation on the Membrane website .
Source: https://habr.com/ru/post/31152/
All Articles