Meta-Learning-Based Scheme for Cross-Scenario Path Loss Prediction

As the sixth generation mobile communication system (6G) evolves, its envisioned paradigms have fragmented communication scenarios into specialized domains, requiring tailored physical-layer designs. Existing channel models face cross-scenario generalization deficiencies, while deep learning (DL)-based methods rely on large training datasets and lack adaptability to new scenarios. This paper presents a generalizable cross-scenario channel modeling scheme, including a Meta-learning-based Channel Modeling (MCM) algorithm, tri-modal input, and RadioNeXt network. Simulations and measurements verify that the proposed method outperforms transfer learning (TL)-based approaches, especially in hybrid-density scenarios, with the RadioNeXt network surpassing ResNet50 and Deep Vision Transformer (DeepViT).