Channel Modeling Based on 3D Scenario Information for V2I Communications

In this paper, we propose a novel 3D scenario information based channel modeling method with deep learning for Vehicular-to-infrastructure (V2I) communications. Specifically, the vehicular scanning sensors are utilized to capture the point cloud information of 3D scenario directly, and then the channel characteristics are generated and mapped with the 3D scenario information by deep learning network. Benefited from the above modeling framework, the complexity of environment reconstruction and geometric calculation can be reduced greatly. The simulation results show that the prediction precision of reflection point by deep learning has significant impact on the modeling accuracy. By some efficient scenario information extraction operations, the proposed method can realize more than 95% channel modeling accuracy with much lower complexity.