Parallelized Physical Optics computations for Scattering Center Models in radio channel simulations

Scattering Center Models (SCMs) are an approach to efficiently characterize electromagnetic scattering of complex shaped objects in terms of a distribution of equivalent sources. SCMs have been widely used in the area of radar target modeling, particularly for object characterization and classification. Recently, the SCM approach has been adapted to model the electromagnetic properties of vehicles embedded in a comprehensive environment for the simulation of automotive radar channels. In this paper the application of SCMs for deterministic simulations of communication channels in automotive environments is presented. The main advantage of this concept is a drastic reduction of complexity for the ray tracing simulations. Instead of using a full geometrical description of the scene, the most complex objects are replaced by an equivalent SCM and a strongly simplified surface model. The SCM of an object is determined in advance, independent from the surrounding environment. In order to achieve a precise parameterization of individual Scattering Centers (SC), the Physical Optics (PO) approximation can be applied. Unfortunately, in contrast to a parameterization based on the Geometrical Optics (GO) contribution only, this approach requires field computations for a large number of observation directions. To handle this computational challenge, a parallelization concept is presented, which is based on multi-core architectures and on the utilization of Graphics Processing Units.

I. Introduction

SC representations are a well-known technique for modeling the scattering characteristics of complex objects. They are used in the fields of target identification, efficient Radar Cross Section (RCS) simulation or data compression, see e.g., [1], [2], [3]. Generally, these representations are based on the observation that for very high frequencies, i.e., the wavelength is much smaller than typical object dimensions, the scattering mechanisms can be attributed to distinct geometrical features of the object and are therefore well localized. This can be used for a sparse scattering representation.