Ray Tracing for Sensor Simulation using Parallel Functional Programming
Examensarbete för masterexamen
von Nordenskjöld, Ari
In the modern automotive industry, cars are tested in virtual environments in order to save time and money. Sensors are important components of modern cars which also need to be tested together with the rest of the car and its software. Ray tracing is an effective method for simulating how light interacts with an environment, and can be used for simulating the sensors used in modern cars. We present an implementation of a physically based ray tracer that can flexibly simulate both vision and LIDAR sensors. This is enabled by the presence of certain features like MIS and spectral path tracing. The ray tracer is written in the parallel functional programming language Futhark, and runs in parallel on the GPU. For validation, we provide a framework of relative measurements which can be used to compare the ray tracer to other implementations, or interpreted as an error to minimise for systematically improving the accuracy of the implementation.
physically based ray tracing , parallel functional programming , sensor simulation , LIDAR , camera , Futhark