Fast Physical Optics Integrated Based on Vulkan Ray Tracing
Xirun Yin,Ruoming Zhang,Ce Ding,Yanpeng Jia,Hai Lin
DOI: https://doi.org/10.1109/piers62282.2024.10618696
2024-01-01
Abstract:High-frequency asymptotic methods, such as Physical Optics, handle the scattering analysis of electrically large targets effectively, due to reducing storage and CPU time. But the occlusion relationship analysis of a complex 3D mesh still brings an obvious computation overhead. In the PO method, ray casting is a basic method for judging the visibility of faces. Ray tracing naturally conforms to human visual characteristics, simply emitting a ray from an object to the source to see if it is obstructed by any object in the scene to determine if it is in shadow. For a long time before, the implementation was mainly based on the CPU. This process is time-consuming and once the correlation between rays is not strong, the coherence of memory will deteriorate. With the development of modern GPUs, ray tracing computation cores have been integrated into the GPU hardware to achieve real-time ray tracing. We utilized Vulkan Ray Tracing interface for electromagnetic scattering calculation. Different from implementations by CUDA and compute shader, ours directly employs hardware computing units designed for ray tracing. Our method greatly accelerates simulation speed for PO under the far field condition, especially in the case of large-scale calculation. We will introduce the principle of PO algorithms and the framework of Vu lkan Ray Tracing hardware interface first. And then we illustrate the design and architecture of our simulation system based on ray tracing hardware. Finally, by testing benchmark problems, and comparing results and time performance with other commercial software and our CPU-based self-developed program, we will show that the method can handle scattering analysis of electrically large bodies in high efficiency.