Robot learns to clean by demonstration
Researchers at TU Wien have taught a robotic arm to learn by imitation with the help of a technically prepared cleaning sponge.
Cleaning robot at work.
(Image: ACIN / TU Wien)
Researchers at TU Wien have developed a process that enables robots to learn activities such as cleaning or sanding by imitating humans. This involves collecting data from a sponge with force sensors and tracking markers and then feeding it to a neural network.
Robots make our everyday lives easier. They are used in industry to reliably perform repetitive tasks. Things become difficult when the shape of the workpiece changes or the work has to be carried out with varying intensity.
If the robot arm is to clean or paint an object with corners and cats, the effort required to "precisely define these things in fixed rules and predefined mathematical formulas" is high, as the researchers at TU Wien describe a common problem. To stay with the example of the cleaning robot: It has to take into account that the sponge is used at different angles and with different force.
Empfohlener redaktioneller Inhalt
Mit Ihrer Zustimmung wird hier ein externes YouTube-Video (Google Ireland Limited) geladen.
Ich bin damit einverstanden, dass mir externe Inhalte angezeigt werden. Damit können personenbezogene Daten an Drittplattformen (Google Ireland Limited) übermittelt werden. Mehr dazu in unserer Datenschutzerklärung.
The human cleans, the robot follows suit
For precisely such applications, the researchers are using the approach of allowing robots to learn by imitation and experience, just like humans. To do this, a human cleans a sink with a specially equipped sponge. Force sensors and tracking markers on the sponge generate data which, after statistical processing, is used to train a neural network.
"The robot learns that you have to hold the sponge differently depending on the shape of the surface," explains doctoral student Christoph Unger, "and that you have to apply a different force on a highly curved area than on a flat surface." The highlight: the robot learns this itself using a new type of learning algorithm, even though only part of the washbasin was cleaned in the demonstration.
Videos by heise
The researchers see potential for the future in mobile robots that work in smaller areas such as workshops and share their accumulated experience with each other. In this "federated learning" process, the basic knowledge of the respective applications would be shared with other robots.
Elsewhere, researchers are already working on helpful robot prototypes that interact with humans by voice.
(hoh)
