Design and Performance Analysis of a Real-Time Positioning Server Based on a Multi-MNO Cell Information Generation Model

Le résumé fourni par la source

This paper proposes a Transformer-based deep learning model that comprehensively utilizes cell information from multiple Mobile Network Operators (Multi-MNOs) and presents a design and implementation plan for a server-client system capable of real-time service. Conventional positioning technologies that rely on single-MNO information face significant limitations in urban and indoor environments, where performance varies significantly depending on base station distribution and network conditions. To address this problem, we designed a Transformer model that interprets the serving and neighboring cell information from a user's current MNO to infer the corresponding cell environment of other MNOs at the same location. This model enhances positioning accuracy by effectively learning the complex interrelationships between various cell signals through its self-attention mechanism. Furthermore, to enable the practical application of the proposed model in Location-Based Services (LBS), we designed a high-performance server architecture based on FastAPI, which is optimized for asynchronous processing suitable for real-time applications. This paper describes the model's architecture and training strategy using a large-scale dataset collected near Daejeon City Hall Station. It also details the overall implementation plan for a system that receives cell information from a client and returns a location in real-time. Experimental results show that the proposed model achieves a high average F1- score of 0.87 when predicting another MNO's cell information, laying the groundwork for future empirical testing and providing a blueprint for the development of next-generation LBS systems.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.