Balancing Complexity and Accuracy: Trajectory-Enhanced Fingerprinting Localization With a Realistic Simulation Framework for 5G

This paper underscores the critical role of outdoor localization in wireless network operations and Internet of Things (IoT) applications. While Global Navigation Satellite Systems (GNSS), such as GPS, provide meter-level accuracy, their limitations in urban environments and high power consumption make them unsuitable for many use cases. Consequently, fingerprint-based localization methods leveraging existing sensor infrastructures have emerged as viable alternatives. This study investigates both traditional fingerprint-based algorithms and advanced deep learning techniques, noting their potential to enhance accuracy alongside challenges in data processing and collection. To address these issues, we propose a novel probabilistic model that enhances Received Signal Strength Indicator (RSSI)-based user location inference by incorporating user trajectories. This approach mitigates the effects of noise and signal fading. We further developed a comprehensive algorithm integrating this probabilistic model with system-level parameters to accommodate diverse real-world deployment scenarios. Performance evaluations demonstrate that our trajectory-enhanced method achieves accuracy improvements of 10% to 35% in urban environments compared to RSSI-only, Weighted k-Nearest Neighbor (WKNN), and neural network-based approaches. These enhancements rely solely on User Equipment (UE) signal strength from the serving cell and benefit from rapid training processes. Additionally, we introduce a simulation framework build on top of OMNet++, Veins, and Simu5G platforms, enabling realistic and generic validation of localization algorithms. This framework allows researchers and telecommunications operators to efficiently evaluate algorithm performance across diverse scenarios, reducing costs and deployment efforts by eliminating the need for extensive hardware resources.