Implementation of a Transfer Learning and Fingerprint-Based Positioning System

The extensive deployment of 5G, has fueled the demand for location-based services. However, indoor positioning remains a persistent challenge, prompting extensive research efforts to address these issues. The advent of artificial intelligence (AI) technology has opened up new possibilities for positioning in the mobile communications domain. This paper proposes a transfer learning and fingerprint-based positioning system that leverages 5G signal strength information. Our proposition demonstrates the ability to rapidly adapt to environmental factors in real-world scenarios, significantly enhancing device positioning accuracy. To validate our proposed approach, we constructed a real network environment using commercial base stations, a commercial core, and commercial smartphones, and deployed the proposed transfer learning model in this real environment. Experimental results in real-world environments validate the effectiveness of our approach, achieving an average location estimation accuracy of over 90 %.