Abstract
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.