Smart robot uses 3D vision to locate lost objects 30% more efficiently

Researchers at the Technical University of Munich (TUM) have developed a search robot that uses a combination of three-dimensional (3D) mapping and language models to locate lost objects in homes and dynamic environments with 30% greater efficiency than random searching. The robot, equipped with a camera that captures depth data, builds a constantly updated 3D spatial map of its surroundings accurate to within centimeters. It integrates computer vision algorithms to detect objects and surfaces, while a language model interprets relationships between these elements to guide the robot’s navigation and decision-making. This allows the robot to understand suitable locations for items, such as recognizing that glasses might be placed on a table but not on a stovetop. The system also features memory capabilities, enabling it to compare current visual data with past images to detect changes with about 95% accuracy and identify likely locations of missing objects. Although the current prototype focuses on searching open spaces, the team is working on enhancements to enable the robot to search inside closed areas like drawers