Increasing the quality of real-time rendering in driving simulation by means of programmable graphics hardware [Elektronische Ressource] = Qualitätssteigerung der Echtzeitvisualisierung in der Fahrsimulation mittels programmierbarer Graphik-Hardware / vorgelegt von Franz Lankes

Increasing the Quality of Real-Time Rendering in
Driving Simulation by Means of Programmable
Graphics Hardware
Qualitätssteigerung der Echtzeitvisualisierung in
der Fahrsimulation mittels programmierbarer
Graphik-Hardware
Der Technischen Fakultät der
Universität Erlangen–Nürnberg
zur Erlangung des Grades
DOKTOR–INGENIEUR
vorgelegt von
Dipl.-Inf. Franz Lankes
Erlangen - 2010Als Dissertation genehmigt von
der Technischen Fakultät
der Universität Erlangen–Nürnberg
Tag der Einreichung: 21.09.2009
Tag der Promotion: 09.11.2009
Dekan: Prof. Dr.-Ing. Reinhard German
Berichterstatter: Prof. Dr.-Ing. Marc Stamminger
Prof. Dr.-Ing. Stefan Mülleri
Abstract
Driving simulation is an essential research and development tool in car engineering and
other sciences world wide. It allows to study driver behavior as well as to analyze and
evaluate new car devices at early development stages, saving costs and time.
To obtain research results which are transferable to reality, subjects must be provided
with a highly convincing simulation environment. The driving experience should feel
as real as possible so that subjects behave as they would do when driving a real car.
Since human vision is significant for real driving, highly realistic image generation must
be a fundamental part of driving simulation.
In the past, the simulation of human vision has been widely neglected in driving simu-
lation. This thesis addresses this shortcoming and focuses on real-time rendering tech-
niques to drastically improve the image quality in high-end driving simulation soft-
ware. Advanced rendering techniques developed at the departement of driving sim-
ulation of BMW Group Research and Technology are presented that are valuable for
high-quality image generation. The techniques described include the generation and
rendering of terrains, the reproduction of road surfaces, the visualization of cars, the
simulation of light and shadows and more.
Aside from analyzing existing methodologies, novel rendering techniques that address
the aforementioned visualization issues in driving simulation have been developed. All
introduced algorithms are real-time capable and are well suited for the implementation
on programmable graphics hardware.
The presented rendering techniques improve the image generation quality of driving
simulation software dramatically. This leads the simulation of vision much closer to the
final goal of photo realism.ii Chapter 0. Abstractiii
Contents
Abstract i
Contents iii
List of Figures vii
I Introduction and Problem Statement 1
1 3
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Project Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
II Background 9
2 Driving Simulation 11
2.1 Driving Simulation at BMW Group Research and Technology . . . . . . . 11
2.2 Simulation Software SPIDER . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2.1 Modular Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2.2 Module Communication . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2.3 SGI’s OpenGL Performer and Custom Enhancements . . . . . . . . 18
2.2.4 SPIDER’s Visual Quality . . . . . . . . . . . . . . . . . . . . . . . . . 19
3 Fundamentals and Background of Image Generation 23
3.1 The Rendering Pipeline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.1.1 The Application Stage . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.1.2 The Geometry Stage . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.1.3 The Rasterization Stage . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.2 The Graphics Processing Unit . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.2.1 The Vertex Shader . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.2.2 The Geometry Shader . . . . . . . . . . . . . . . . . . . . . . . . . . 26iv Contents
3.2.3 The Pixel Shader . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2.4 The Output Merger . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
III Study on Driving Simulation 29
4 Simulator Survey 31
4.1 Motivation and Objective of the Study . . . . . . . . . . . . . . . . . . . . . 31
4.2 Test Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.2.1 Subjects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.2.2 Experiment Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.2.3 The Trial . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.2.4 Questionnaire Design . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.3 Trial Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
IV Research Domains 37
5 Procedural Generation of Road Sceneries 39
5.1 Procedural of Landscapes . . . . . . . . . . . . . . . . . . . . . 39
5.1.1 Procedural Modeling With Noise Functions . . . . . . . . . . . . . . 40
5.1.2 Terrain Elevation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.1.3 Texture Creation and Texturing . . . . . . . . . . . . . . . . . . . . . 41
5.1.4 Water Bodies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.2 Vegetation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.2.1 Generation of Vegetation . . . . . . . . . . . . . . . . . . . . . . . . 44
5.2.2 Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
5.3 Generation of Roads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
5.3.1 Route Definition and Road Generation . . . . . . . . . . . . . . . . 46
5.3.2 Procedural Bridges and Tunnels . . . . . . . . . . . . . . . . . . . . 46
5.4 Procedural Generation of Buildings . . . . . . . . . . . . . . . . . . . . . . . 48
5.4.1 The Process . . . . . . . . . . . . . . . . . . . . . . . . . 48
5.4.2 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
6 Real-Time Rendering of Road Sceneries 53
6.1 Lighting Simulation for Terrains . . . . . . . . . . . . . . . . . . . . . . . . 53
6.1.1 Outdoor Lighting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
6.1.2 Light Scattering . . . . . . . . . . . . . . . . . . . . . . . . 54
6.2 Rendering of Clouds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
6.3 of Vegetation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
6.3.1 Rendering Prairies . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
6.3.2 Fast Lighting of Vegetation . . . . . . . . . . . . . . . . . . . . . . . 63
6.4 Rendering Waters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64Contents v
6.5 Dynamic Range . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
6.5.1 Generating High Dynamic Range Content . . . . . . . . . . . . . . 66
6.5.2 Tone Mapping and Automatic Exposure . . . . . . . . . . . . . . . 67
6.5.3 Hierarchical Blooming . . . . . . . . . . . . . . . . . . . . . . . . . . 68
7 Visualization of Roads 71
7.1 Visualizing the Roughness of Asphalt . . . . . . . . . . . . . . . . . . . . . 71
7.2 Simulating the Varying Reflectivity of Road Surfaces . . . . . . . . . . . . . 72
7.3 Visualizing Roads with Distinctive Height Variances . . . . . . . . . . . . . 74
7.4 Spreading Details on Roads . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
8 Visualization of Vehicles 79
8.1 Model Acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
8.2 Rendering High Quality Car Paint . . . . . . . . . . . . . . . . . . . . . . . 81
8.3 Precalculated Real-Time Environment Reflection . . . . . . . . . . . . . . . 82
8.3.1 Considerations and Caching Strategy . . . . . . . . . . . . . . . . . 84
8.3.2 Enhancements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
8.4 Motion Blur for Wheels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
8.5 Simulation of Lights and Indicators . . . . . . . . . . . . . . . . . . . . . . . 88
9 Simulating Light and Shadows 89
9.1 Rendering of . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
9.2 Shadows and Lighting for Static Objects . . . . . . . . . . . . . . . . . . . . 90
9.2.1 HDR Light Mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
9.2.2 Parallel Projected Light Maps . . . . . . . . . . . . . . . . . . . . . . 92
9.3 Shadows and Lighting for Dynamic Objects . . . . . . . . . . . . . . . . . . 94
9.3.1 Extensions on Shadow Volumes . . . . . . . . . . . . . . . . . . . . 94
9.3.2 Improving Performance on Assarsson’s Algorithm . . . . . . . . . 97
9.3.3 on Shadow Maps . . . . . . . . . . . . . . . . . . . . . . 97
9.4 Dynamic Ambient Occlusion and Ambient Lighting . . . . . . . . . . . . . 99
9.4.1 Considerations on Sample Points . . . . . . . . . . . . . . . . . . . . 101
9.4.2 on Directions . . . . . . . . . . . . . . . . . 102
9.4.3 Ambient Occlusion in Real-Time . . . . . . . . . . . . . . . . . . . . 103
9.4.4 Lighting Usi