GPU Accelerated EM Modeling Tool for Electrically Large Objects

Abstract

Fast modeling of electromagnetic (EM) scattering from electrically large and complex objects is essential to a wide range of applications, such as remote sensing and rapid prototyping of man-made structures. With emerging technologies at higher frequency bands such as terahertz and millimeter wave, such modeling capability will be in increasing demand. Full wave methods like method of moments (MoM) and multilevel fast multipole method (MLFMM) remain useful tools to understand fundamental physical phenomena but quickly become intractable as the electrical size of the object increases. This is due to the prohibitive requirements of computational resources which are often inaccessible to many. Therefore, these methods are not suitable for realistic analysis and fast simulation of EM scattering from electrically large objects. On the contrary, hybrid high frequency methods trade some accuracy to enable efficient and effective simulation on affordable hardware. Collectively, physical optics (PO) method, shooting and bouncing rays (SBR), and physical theory of diffraction (PTD) approximately account for the EM scattering from the objects. The adoption of graphic processing unit (GPU) further enables fast generation of scattering data of various configurations, articulations, and obscuration. In this talk, we will discuss algorithm mapping and GPU implementation of hybrid high frequency methods.