Robots learn to be self-sustained
Researchers of Bielefeld University in Germany are doing interesting trials. They are trying to teach robots to learn and understand the properties of objects that surround them without the aid of preinstalled database.
For this purpose, scientists constructed several robotic arms that resemble human limbs in terms of their look and functionality. Every arm selects an object, tests and studies it. Afterwards manipulators must reveal which object they took and how to interact with it. Some objects are very soft and can be broken if clasped tighter, other have a specific shape so that it is not easy to hold them. Researchers want to teach robots to work out all these consequences in advance. In the future, it will allow machines to adapt to environment and change their behavior depending on the situation.
Scientists use artificial intelligence technologies, combining them with other methods. In particular, they analyzed human actions in the process of learning unknown objects and used the obtained data to teach robots.