A Lightweight Coordinate-Conditioned Diffusion Approach for 6G C-V2X Radio Environment Maps
Transmitter vehicles that broadcast 6G Cellular Vehicle-to-Everything (C-V2X)-based messages, e.g., Basic Safety Messages (BSMs), are prone to be impacted by PHY issues due to the lack of dynamic high-fidelity Radio Environment Map (REM) with dynamic location variation. This paper explores a lightweight diffusion-based generative approach, the Coordinate-Conditioned Denoising Diffusion Probabilistic Model (CCDDPM), that leverages the signal intensity-based 6G V2X Radio Environment Map (REM) from limited historical transmitter vehicles in a specific region, to predict the REMs for a transmitter vehicle with arbitrary coordinates across the same region. The transmitter vehicle coordinate is encoded as a smooth Gaussian prior and fused with the Gaussian noise through a lightweight two-channel conditional U-Net architecture. We demonstrate that the predicted REM closely matches the statistics and structure of ground-truth REM while exhibiting the improved stability and over other widely applied generative AI approaches. The resulting predictor enables rapid and scenario-consistent REM with arbitrary transmitter coordinates, which thereby supports more efficient 6G C-V2X communications where transmitter vehicles are less likely to suffer from the PHY issues.