Learning-based Radio Link Failure Prediction Based on Measurement Dataset in Railway Environments

2025-11-12

AI Summary

To address the challenge of early wireless link failure (RLF) prediction in 5G non-standalone (NSA) railway scenarios, this paper proposes the first lightweight RLF prediction framework leveraging real-world train measurement data sampled at 10 Hz. Methodologically, we systematically evaluate six time-series models—CNN, LSTM, XGBoost, Anomaly Transformer, PatchTST, and TimesNet—under multiple window sizes and prediction horizons. Notably, we present the first empirical validation in railway environments demonstrating TimesNet’s superior F1-score for 3-second-ahead prediction and CNN’s optimal trade-off between accuracy and low latency at 2-second horizon. The framework relies solely on standard service-cell and neighbor-cell measurements accessible via commercial user equipment, requiring no additional hardware. Experimental results show that high-confidence RLF warnings can be generated 2–3 seconds prior to actual link failure, enabling deployable, real-time reliability assurance for 5G train communications.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

In this paper, a measurement-driven framework is proposed for early radio link failure (RLF) prediction in 5G non-standalone (NSA) railway environments. Using 10 Hz metro-train traces with serving and neighbor-cell indicators, we benchmark six models, namely CNN, LSTM, XGBoost, Anomaly Transformer, PatchTST, and TimesNet, under varied observation windows and prediction horizons. When the observation window is three seconds, TimesNet attains the highest F1 score with a three-second prediction horizon, while CNN provides a favorable accuracy-latency tradeoff with a two-second horizon, enabling proactive actions such as redundancy and adaptive handovers. The results indicate that deep temporal models can anticipate reliability degradations several seconds in advance using lightweight features available on commercial devices, offering a practical path to early-warning control in 5G-based railway systems.

Problem

• Predicting radio link failures in 5G railway environments using measurement data
• Benchmarking six machine learning models for early RLF prediction accuracy
• Enabling proactive network actions through reliable failure anticipation in railways

Innovation

• TimesNet achieves highest F1 score
• CNN provides accuracy-latency tradeoff
• Deep temporal models predict reliability degradations