Adaptive and transition-aware beam tracking for 6G mmWave systems with reduced overhead

Millimeter-wave (mmWave) and sub-terahertz (sub-THz) communications are essential to 5G-Advanced and 6G systems, yet they remain highly vulnerable to user mobility and line-of-sight (LOS) blockages. Conventional beam management as defined in 3GPP TS 38.211–214 relies on fixed-period sweeping of Synchronization Signal Blocks or Channel State Information Reference Signals, which becomes inefficient under varying mobility and propagation dynamics. This paper proposes an adaptive and transition-aware beam tracking framework that dynamically adjusts sweeping frequency according to both receiver speed and predicted LOS/NLOS transitions. Four hierarchical strategies are developed: the fixed-interval Reuse Strategy, its transition-aware variant, the Adaptive Reuse Strategy driven by a coherence-time-based physical model, and the Adaptive Hybrid Reuse Strategy. The framework is evaluated on synthetic ray-tracing datasets and the real-world DeepSense 6G scenario. Results show that Adaptive Reuse Strategy and Adaptive Hybrid Reuse Strategy maintain accuracy comparable to state-of-the-art deep learning baselines, such as DeepBT and GRU predictors, while reducing beam-sweeping overhead by up to 78%. A new composite score based on the Utopia-Distance principle is introduced to jointly assess accuracy and efficiency, revealing that the proposed adaptive strategies consistently occupy the Pareto-optimal region across datasets. The findings demonstrate that learning-driven adaptation following principles inspired by 3GPP mobility classes, provides a scalable path to efficient beam management in next-generation wireless networks.