Robots that know their bodies well

A pair of engineers from the Technical University of Munich in Germany have discovered that it is possible to provide robots with a certain degree of proprioception.

Oliver Thansan
Oliver Thansan
04 February 2024 Sunday 09:32
4 Reads
Robots that know their bodies well

A pair of engineers from the Technical University of Munich in Germany have discovered that it is possible to provide robots with a certain degree of proprioception. Thus, through machine learning techniques, artificial devices can have a sense that tells them their situation, their movement and the action of their different parts.

In this way, scientists have confirmed that robots can perceive the position of their joints or the effort they have to carry out depending on the tasks given to them. In summary, the authors of the article in which this novelty is reported, Fernando Díaz Ledezma and Sami Haddadin, consider that machines can know the specific characteristics of their bodies.

Robots can usually move easily thanks to cameras and pressure sensors. The data provided by these elements is processed so that the activity carried out is appropriate. Neither animals nor people behave this way, since they are physically aware of themselves. That is, they have proprioception.

Now, professors at the Munich Institute of Robotics and Machine Intelligence who have led this project have reduced the distance that separates humans from plastic and metal artifacts. The main purpose was that the sensors added would make it possible for these robots to have specific information about their components.

Through these pieces, they should know where a knee or elbow is, and in what direction it should be flexed and with what intensity. Always, in every moment. Díaz Ledezma and Haddadin have achieved this by superimposing sensors and perfecting the interpretation of the material they generate, which has ended up producing better self-awareness.

They have even managed, to a large extent, for the devices to learn to understand their structure and organization, both internal and external. In addition, they have carried out tests on different types of machines, for example, on a six-legged spider robot, on a humanoid and on an articulated arm. As the researchers conclude, the result has been successful in all cases.