CCFPMF: A Channel Model Migration Approach for C-V2X Channel Modeling

As an important part of the intelligent transportation system, the Internet of Vehicles (IoV) uses wireless communication technology to achieve information exchange and sharing between vehicles, pedestrians, infrastructure, and cloud networks. The reliability of V2X communication is influenced by real-time wireless channel features. Hence, V2X channel model becomes a hot spot, attracting academic and industry attention. However, V2X channel models are constructed according to on-road measurement data, which should cover various V2X scenarios, the implement is complex and great workload. But here's a novel thought: shall we migrate existed channel model into V2X model? In this paper, we try to establish a Cross-frequency Channel Feature Parameter Migration Framework (CCFPMF) to generate V2X channel model, The framework integrates MMD-AAE with CycleGAN to capture common channel features in different scenario domains at a specified frequency and trains the model to learn channel scenario migration from 2.6 GHz to 5.9 GHz. The proposed CCFPMF is trained and validated using the CQU-V2XSD dataset and tested on the WAIR-D public dataset. The generated data is very close to the measured data in LCR, AFD and RMS-DS fitting, and shows high consistency in KL divergence, JS divergence and Wasserstein distance. These results confirm that our transfer model generates reliable and realistic data in various detection similarity metrics.