This paper presents a measurement-driven case study on early radio link failure (RLF) warning as device-side network sensing and analytics for proactive mobility management in 5G non-standalone (NSA) railway environments. Using 10~Hz metro-train measurement traces with serving- and neighbor-cell indicators, we benchmark six representative learning models, including CNN, LSTM, XGBoost, Anomaly Transformer, PatchTST, and TimesNet, under multiple observation windows and prediction horizons. Rather than proposing a new prediction architecture, this study focuses on quantifying the feasibility of early warning and the trade-offs among observation context, prediction horizon, and alarm reliability under real railway mobility. Experimental results show that learning models can anticipate RLF-related reliability degradation seconds in advance using lightweight features available on commercial devices. The presented benchmark provides practical insights for sensing-assisted communication control, such as proactive redundancy activation and adaptive handover strategies, aligning with the 6G vision of integrating sensing and analytics into mobility control.

翻译：本文提出一种基于测量数据的早期无线链路故障预警案例研究，作为设备侧网络感知与分析技术，用于5G非独立组网铁路环境中的主动移动性管理。利用包含服务小区与邻区指标的10Hz地铁列车测量轨迹，我们在多种观测窗口和预测时域下对六种代表性学习模型进行基准测试，包括CNN、LSTM、XGBoost、Anomaly Transformer、PatchTST和TimesNet。本研究并非提出新的预测架构，而是重点量化真实铁路移动场景下早期预警的可行性，以及观测上下文、预测时域与告警可靠性之间的权衡关系。实验结果表明，学习模型能够利用商用设备可获取的轻量级特征，提前数秒预测与无线链路故障相关的可靠性下降。所提出的基准测试为感知辅助的通信控制（如主动冗余激活和自适应切换策略）提供了实用参考，符合6G将感知与分析融入移动性控制的技术愿景。