Researchers from Miguel Hernández University in Spain have developed an AI localization system that helps robots quickly recover from loss of position.

The problem of the "kidnapped robot" represents a major challenge in robotics, occurring when an autonomous robot suddenly loses its localization, for example, after being moved or stopped. Researchers at the Miguel Hernández University in Elche, Spain, have created a new localization system based on artificial intelligence, called MCL-DLF (Monte Carlo Localization – Deep Local Feature), which helps robots recover quickly, even in changing environments. The system combines 3D LiDAR scanning with deep learning models to identify relevant features of the environment and to maintain multiple position hypotheses simultaneously. This allows robots to orient themselves similarly to humans, using large structural elements to determine a general area and smaller details to refine their position. MCL-DLF proves to be robust in dynamic environments, constantly maintaining updated localization estimates, which is essential in variable conditions, such as seasonal changes or lighting variations. Tests conducted on the university campus showed greater accuracy and more consistent performance compared to traditional methods.