An Improved Ray Tracing Algorithm for High-efficiency Prediction of Ultra-large-scale Complex Electromagnetic Environments

Ray tracing is a widely used computational method for solving ray paths in channel modeling, radiation prediction, and spectrum management. This paper introduces an ultra-large-scale ray tracing model that uses an improved ray tracing algorithm. The model supports the input of real-world terrain data and building data to construct large-scale real 3D scenes that cover a variety of scenarios. The model can provide accurate ray propagation paths and electromagnetic field distributions. To optimize the effectiveness of the model from both accuracy and resource consumption perspectives, the computation of the ray path-solving is delegated to the GPU, which possesses powerful parallel computing capabilities. Additionally, to adapt to the vast amount of data in ultra-large-scale scenarios and alleviate hardware burdens, this study employs an improved Douglas-Peuker algorithm applied to 2D grid to simplify the terrain data. To apply the DP algorithm which originally designed for line elements to the terrain grid, this study proposes a method to subdivide the regular modeling of the terrain data. And connections between neighbor nodes are also established by applying it to four directions of the gird: rows, columns, and two diagonals. The model also addresses the inadequacy of the uniform diffraction theory in handling various scenarios. It analyzes situations where the theory cannot adapt and proposes a diffraction coefficient calculation method that includes the effects of vertex diffraction, thereby improving calculation accuracy. In the previous diffraction models, the edge and vertex diffraction are usually treated as a whole diffraction field because both types of diffraction occur simultaneously. However, the angles of the outgoing rays of edge and diffraction are quite different which brings error when tracking the ray path. In this paper, to reduce this error, a more precise method for calculating the diffraction coefficient is given by separating the vertex diffraction from uniform diffraction calculation model. Compared to traditional semi-empirical models and CPU-based models, this model demonstrates stronger adaptability to real environments and provides more accurate predictions for radio wave propagation.