摘要 : In this letter, a fast algorithm based on domain decomposition and machine learning is proposed for radio wave propagation over complex terrain. In this approach, equivalence surfaces (ESs) with the same size are arranged based on... 展开 In this letter, a fast algorithm based on domain decomposition and machine learning is proposed for radio wave propagation over complex terrain. In this approach, equivalence surfaces (ESs) with the same size are arranged based on the distribution of obstacles; thus, the impact of each obstacle on the radio wave propagation can be expressed by the input and output equivalent electromagnetic currents (EEMCs) on the ESs. Due to the presence of dielectric ground, the full-wave calculation process involves the discrete complex image method in the Poggio–Miller–Chang–Harrington–Wu equation. Deep learning is introduced to complete the complex process of determining the output EEMCs from input EEMCs on ESs to accelerate the calculation. Considering scattering field coupling between multiple terrains, we iterate the scattering fields generated by the interaction between ESs until convergence occurs. The simulation results show that the proposed algorithm is in good agreement with the method of moments approach, and its computational speed is significantly improved, demonstrating that the proposed method has high efficiency and good generalization ability. 收起