Application of Relational Databases to the Acceleration of Ray Tracing in High Frequency Asymptotic Techniques

Abstract

A new ray-tracing acceleration technique is presented for electromagnetic simulation problems using the Uniform Theory of Diffraction and meshes of planar facets. The innovation involves using relational databases to accurately store spatial information, enabling spatial indexing through space partitioning with R-trees. This technique effectively reduces the computational cost of several critical phases, including the shadowing test. Additionally, there are multiple advantages to utilizing this technology, such as automated memory and disk management along with a query planner that organizes the instructions automatically. Direct rays, multiple reflections, multiple transmissions, simple diffraction, and combinations of these effects have been implemented in PostgreSQL and its spatial library PostGIS. Compared to traditional techniques that employ Angular Z-Buffer acceleration and store information solely in RAM using a low-level language, this approach decreases memory usage by more than 90% in complex scenarios. It also shows a decrease in execution time by more than half when the scenario is sufficiently complex.