基于Vulkan的电大复杂目标电磁散射SBR计算方法与GPU加速技术

      Vulkan-based SBR method for electromagnetic scattering of electrically large and complex targets with GPU acceleration techniques

      • 摘要: 本文提出了一种基于 Vulkan 架构的弹跳射线(shooting and bouncing rays, SBR)加速计算方法,用于复杂电大目标雷达散射截面的快速计算。本文设计了高效的 Vulkan 计算着色器,充分利用GPU硬件光追,显著提升了SBR法中光线求交的计算速度;引入了双命令缓冲机制,使得 CPU 与 GPU 能够高效协同工作,从而加速多角度扫描任务的执行;在虚拟孔径面上划分互不干扰的子任务,进一步提升了多 GPU 并行的利用效率。实验结果表明,所提出方法在计算复杂电大目标雷达散射截面相较于 RLGO 方法实现了 40倍以上的加速;双命令缓冲机制提升了约 42%的多角度扫描速度;双 GPU计算并行效率超过 90%。

         

        Abstract: This paper proposes a Vulkan-based acceleration method for Shooting and Bouncing Rays (SBR) to obtain fast computation of the radar cross section (RCS) of complex electrically large targets. An efficient Vulkan compute shader is designed to fully leverage GPU hardware ray tracing, which an significantly improve the ray intersection computation speed for the SBR method. A dual-command-buffer mechanism is introduced to enhance CPU-GPU collaboration for accelerating multi-angle scanning tasks. Additionally, non-interfering sub-tasks are divided on the virtual aperture surface, further improving multi-GPU parallel utilization efficiency. Experimental results demonstrate that the proposed method achieves over 40x acceleration in RCS computation for complex electrically large targets compared to the RLGO method. The dual-command-buffer mechanism improves multi-angle scanning speed by approximately 42%, while dual-GPU computation achieves over 90% parallel efficiency.

         

      /

      返回文章
      返回