A team of researchers at the Technical University of Munich was able to develop a robot capable of tracking lost items in changing home environments, relying on linguistic models and 3D maps.
Scientists have described this robot, with a camera on top, as resembling a “broomstick on wheels,” and it is considered among the first robots to use image understanding to carry out a clearly defined task. This system aims to locate lost items upon request, by integrating knowledge taken from the Internet with a spatial map of the surrounding place.
In this regard, Angela Schulig, a professor in the field of robotics and artificial intelligence at the university and head of the Robotics Laboratory, stated that the robot has been trained to understand its surrounding environment and determine the appropriate places to place things, such as tables or window edges, while ignoring inappropriate places such as stoves or the sink.
The robot relies on artificial intelligence in two primary ways: first, it uses computer vision algorithms to detect objects and surfaces in the environment, and second, a linguistic model interprets the relationships between these objects and turns them into instructions that the robot can use to navigate.
In addition, the robot keeps a record of previous images and compares them with new data from the environment, which increases the accuracy of locating lost objects. Experiments showed that the robot is able to check potential locations 30% more effectively compared to a random search. When it detects a new item, it can identify the change with up to 95% accuracy and classify the location as a potential area for the missing item.
The team is currently working on developing an improved version of the robot to search inside drawers, cabinets and closed areas.
