Optical simulation for illumination using GPGPU ray tracing

Abstract

High-speed ray tracing for illumination optics using GPGPU was investigated. Optical simulation for illumination optics requires many rays tracing for precise simulation. Especially, optics for automotive LED lighting have small textures on the exit surface of the lens to diverge part of the light for satisfying specific illumination pattern which is required in the regulation. Many ray tracing requires much simulation times and it increases development cost. Recently, parallel computing using CPU and GPU has been used for accelerating computing speed and reported its merit in computer sciences. In this research, the ray tracing consists of two parts which are intersection searching and refraction calculation was done in parallel using CUDA, GPGPU API provided by NVIDIA. Interpolation calculations such as linear interpolation, Nagata triangular patch interpolation, and Nagata quadrilateral patch interpolation were used in intersection searching calculation. The results indicate that there is a possibility to accelerate ray tracing speed by using GPU. As a representative example, GPU ray tracing was about twice faster than the commercial software. In addition, error differences depend on the interpolation types for intersection calculation were observed. Moreover, the results indicate calculation error differences between single precision float calculation and double precision float calculation. In conclusion, even there are several issues such as errors from interpolation and calculation precision, accelerated ray tracing using GPU was achieved.