In the latest technical developments, a robot has been developed at the Technical University of Munich that specializes in searching for lost items inside buildings.
To accomplish its mission, this robot relies on information obtained from the Internet, in addition to a detailed spatial map of the environment in which it operates.
The innovation, made by Angela Schwellig of the Learning Systems and Robotics Lab at the Technical University of Munich, was published in the journal IEEE Robotics and Automation Letters. The robot superficially resembles a mop mounted on wheels with a camera on top.
In order for the robot to find something, such as missing glasses in the kitchen, it first needs to explore the space and build a 3D map of the room. The camera takes 2D images, but each point (pixel) in these images contains information about depth. Through this data, an accurate spatial map of the environment is built with an accuracy of up to a centimeter, and it is continuously updated. In addition, the robot’s built-in computer identifies objects in the image and evaluates their importance to the human.
“We trained the robot how to recognize its surroundings. Our goal is to develop robots capable of self-guidance in any conditions. This basic understanding is essential for human-like robots in factories, or care robots in homes, and is important for all robots operating in constantly changing spaces,” says Professor Schwellig.
In this way, the robot can understand that the table or windowsill are suitable places to put glasses, while the stove or sink are not.
The researcher explained that the linguistic model understands the relationships between things, and this information is converted into a language that the robot understands. Numbers appear on the 3D map indicating the probability of the desired object being located at each point, and these probabilities are constantly updated. Experiments have shown that the robot explores potential locations 30% more efficiently than randomly searching around a room. AI is used in two ways: in image recognition and linguistic model operation. (Russia Today)
