Mundos Procedurales. Idea general. datos. rica y variada. Gustavo Patow Geometry and Graphics Group (GGG) Universitat de Girona (UdG( UdG)

Transcription

1 Jornadas sobre estrategias de generación de entornos colaborativos virtuales Mundos Procedurales Gustavo Patow Geometry and Graphics Group (GGG) Universitat de Girona (UdG( UdG) 1 Idea general Importante amplificación de bases de datos Una entrada limitada produce una salida rica y variada Si puede generarse sobre la marcha El artista no tiene que diseñarlo No hay necesidad de almacenarlo/transmitirlo 2 1

2 Tecnologias Relacionadas (General) Houdini 3 Tecnologias Relacionadas (Específicas) 4 2

3 Tecnologias Relacionadas (con los temas de estas jornadas) 5 Contenidos Capítulo 1: Introducción Capítulo 2: Paisajes Procedurales Planetas Terreno Fractales Perlin Noise Hypertextures Erosión Capítulo 3: Jardín n virtual Botánica Virtual L-systems Ecosistemas Capítulo 4: Vida Virtual Capítulo 5: Modelado Urbano Edificios Ciudades ½clase ½clase 6 3

4 Planets 7 Creating Spherical Worlds From Siggraph 2007 Technical Sketch By Kate Compton, James Grieve, Ed Goldman, Ocean Quigley, Christian Stratton, Eric Todd, Andrew Willmott 8 4

5 Need to have lots Planet Constraints many more than we can manually author Need to be fast to generate We can t store all these planets Would like to transmit them at some point Need to support terraforming Player modification of planet to support life planets need to be spherical 9 Parameterization: Cube Maps Chose cube maps as the best compromise Faces are grids Familiar from previous games Distortion at corners But not too bad, much better than pole distortion Face wrapping is tractable Pick right face mappings -> simple permutation rules Projective projection Lines map to great circles on sphere: very useful! Colour Map 10 5

6 Generating Height Fields Brush system that operates on the sphere Brushes are 2D textured rectangles Several different brush operations Conditionally raise or lower terrain 11 Texturing Derive Control Map from height field Filter: water level, gradient, curvature Combine according to tech artist formula Blends source textures to form base colour Blends detail maps on the fly Planets have type, atmosphere, temperature Control colour ramps, and atmosphere/fogging 12 6

7 Pipeline (Swarm) 13 Terraforming 14 7

8 Authoring: sketches

9 17 Results 18 9

10 Landscapes 19 Perlin Noises in 1-D

11 Perlin Noises in 2-D2 21 Weighted Sums noise: -Worn metal -Water wave (gradient) Sin( x + Sum[1/f * noise ] ): -Turbulent flows -Fire -Marble Sum[1/f * noise]: -Rock -Mountains -Clouds Sum[1/f * noise ]: -Turbulent flows -Fire -Marble -Clouds K. Perlin 22 11

12 Fractal Noise Terrain F.K. Musgrave 23 Adding Water Use an elevation threshold (z < z water ) F.K. Musgrave 24 12

13 Terrain Example F.K. Musgrave 25 Terrain Example F.K. Musgrave 26 13

14 Terrain Example F.K. Musgrave 27 Hydraulic Erosion Simulation Based on Fast Fast Hydraulic Erosion Simulation and Visualization on GPU, Xing Mei, Philippe Decaudin, Bao-Gang Hu 28 14

15 Introduction Hydraulic Erosion Process Water flow takes away (dissolves) the soil and relocates (deposits) it somewhere else Effect Changes terrain appearance Creates interesting geo-morphological structures Gullies Eroded landscape on Maui Island Erosion: Motivation An hydraulic erosion simulation method Effective produces the important features in the erosion process Gullies, water catchments, deposited sediment Efficient Well mapped to GPU Interactive frame rates for large size terrains

16 5 steps in one cycle of the simulation Rainfall & River Source water Erosion on GPU Method Overview Water Movement (1) Water Increment (2) Water Movement (flow simulation) (3) Erosion or Deposition (4) Sediment Transportation (5) Evaporation Evaporation Erosion Deposition Sediment Transportation Catchment Terrain Deposited Sediment Results Deposited sediment in a basin The bottom of the basin is flattened by the deposited sediment

17 Results River flow in a drained channel Part of the original river bank get eroded Results Mountain scene eroded by rainfall

18 Results Final example: the combination of the rainfall and the river source Virtual Botany 36 18

19 L-Systems: Tree structures 37 Tree structures 38 19

20 Tree structures 39 Tree structures 40 20

21 L-Systems for Plants L-Systems can capture a large array of plant species Designing rules for a specific species can be challenging 41 PovTree 42 21

22 Ecosystems Oliver Deussen, Pat Hanrahan, Bernd Lintermann, Radomir Mech, Matt Pharr, and Przemyslaw Prusinkiewicz. Realistic modeling and rendering of plant ecosystems. Proceedings of SIGGRAPH Sources Real Data Manually painted height fields Fractal generation Terrain erosion Ecosystems 44 22

23 Explicitly Average densities Interactively created Plant populaitons Procedurally Simulating interactions between plant populations 45 Terrain-based ecosystem 46 23

24 River Scene 47 Evolving Virtual Creatures Karl Sims ( SIGGRAPH '94) 48 24

25 Creature Representation Genotype Directed graph Phenotype Hierarchy of 3D parts 49 Creature Control Brain: A directed graph of neurons Effectors: Applied at Joints as Forces or Torques 50 25

26 Creature Evolution 1. Create genotypes: 1. Randomly generated 2. Existing genotypes from previous evolution 3. Designed by hand 2. Create an initial population and calculate survival ratio 3. Evaluate fitness and kill the weaks 4. Reproduce the better ones 5. Evolve, and go to step 3 51 Snake like Results - Swimming Padding Symmetrical flippers 52 26

27 Lizard-like Results - Walking Pushers/ Pullers Hoppers 53 Results - Jumping They use the same strategies found for walking and swimming 54 27

28 Build your own creatures! brevecreatures screensaver Procedural Flocking (Boids( Boids) Simulate the movement of a flock of birds in 3-space3 Separation: move to avoid crowding local neighbors Alignment: steer towards average heading of neighbors 56 28

29 Procedural Flocking (Boids( Boids) Simulate the movement of a flock of birds in 3-space3 Cohesion: steer towards average position of neighbors Limited Senses: only neighbors in forward- facing arc are observable

30 Procedural Urban Modeling 59 What do you need to do this? 60 30

31 And this? 61 A lot of work! Assassins Creed 240+ persons Assassins Creed II 450+ persons Grand Theft Auto IV 150+ persons 62 31

32 What is urban modeling? Creating digital models of real or virtual cities Cities are large collections of complex architectural structures 63 Why should you learn about it? Urban models are important! 64 32

33 Why should you learn about it? Urban models are important! Entertainment 65 Why should you learn about it? Urban models are important! Entertainment Mapping and visualization 66 33

34 Why should you learn about it? Urban models are important! Entertainment Mapping and visualization Urban planning time 67 What is the main challenge? Solving the content problem As computing and display capabilities continually improve, audience expects ever higher quality digital content Traditional tools are insufficient for increasing demand and few tools are available for efficient large-scale urban modeling 68 34

35 What are the current approaches? Geometric content creation: Procedural Methods: tools to program the geometry of buildings, parcels, roads, facades 69 Procedural Modeling of Cities Parish and Müller - SIGGRAPH

36 L-Systems for Cities Start with a single street Branch & extend w/ L-L System goals/constraints: Goals control street direction, spacing Contraints allow for parks, bridges, road loops Real Manhattan 71 System Pipeline 72 36

37 Module 1: Streetmap Creation Input: Image maps, parameters for rules Output: A street graph for interactive editing 73 Module 2: Division into Lots Input: Street graph, area usage map Output: Polygon set of allotments for buildings 74 37

38 Module 3: Building Generation Input: Lot polygons,, age map and zone plan Output: Building strings with additional info 75 Module 4: Geometry and Facades Input: Strings and building type Output: City geometry and facade texture (procedural shader) 76 38

39 Modeling of Urban Spaces 77 Modeling of Urban Spaces The traditional urban modeling pipeline Input Major Minor Blocks Lots Buildings Mass Facades Aerial Views 78 39

40 Modeling of Urban Spaces The traditional urban modeling pipeline Input Major Minor Blocks Lots Buildings Mass Facades Aerial Views Input 79 Modeling of Urban Spaces The traditional urban modeling pipeline Input Major Minor Blocks Lots Buildings Mass Facades Aerial Views Input Target architectural design 80 40

41 Modeling of Urban Spaces The traditional urban modeling pipeline Input Major Minor Blocks Lots Buildings Mass Facades Aerial Views Input Target architectural design Example 3D model/gis data/imagery 81 Modeling of Urban Spaces The traditional urban modeling pipeline Input Major Minor Blocks Lots Buildings Mass Facades Aerial Views Input Target architectural design Example 3D model/gis data/imagery Socioeconomic data/elevation data 82 41

42 Modeling of Urban Spaces The traditional urban modeling pipeline Input Major Minor Blocks Lots Buildings Mass Facades Aerial Views Input Target architectural design Example 3D model/gis data/imagery Socioeconomic data/elevation data Tensor field 83 Modeling of Urban Spaces The traditional urban modeling pipeline Input Major Minor Blocks Lots Buildings Mass Facades Aerial Views Input Target architectural design Example 3D model/gis data/imagery Socioeconomic data/elevation data Tensor field A simple description of the goal intention 84 42

43 Modeling of Urban Spaces The traditional urban modeling pipeline Input Major Minor Blocks Lots Buildings Mass Facades Aerial Views Road generation 85 Modeling of Urban Spaces The traditional urban modeling pipeline Input Major Minor Blocks Lots Buildings Mass Facades Aerial Views Road generation Extended L systems 86 43

44 Modeling of Urban Spaces The traditional urban modeling pipeline Input Major Minor Blocks Lots Buildings Mass Facades Aerial Views Road generation Extended L systems Hyperstreamlines 87 Modeling of Urban Spaces The traditional urban modeling pipeline Input Major Minor Blocks Lots Buildings Mass Facades Aerial Views Road generation Extended L systems Hyperstreamlines Directed random walks 88 44

45 Modeling of Urban Spaces The traditional urban modeling pipeline Input Major Minor Blocks Lots Buildings Mass Facades Aerial Views Road generation Extended L systems Hyperstreamlines Directed random walks Grow seeds/traffic simulation 89 Modeling of Urban Spaces The traditional urban modeling pipeline Input Major Minor Blocks Lots Buildings Mass Facades Aerial Views Road generation Extended L systems Hyperstreamlines Directed random walks Grow seeds/traffic simulation 90 45

46 Modeling of Urban Spaces The traditional urban modeling pipeline Input Major Minor Blocks Lots Buildings Mass Facades Aerial Views Road generation Extended L systems Hyperstreamlines Directed random walks Grow seeds/traffic simulation Patterns 91 Modeling of Urban Spaces The traditional urban modeling pipeline Input Major Minor Blocks Lots Buildings Mass Facades Aerial Views Blocks and Lots 92 46

47 Modeling of Urban Spaces The traditional urban modeling pipeline Input Major Minor Blocks Lots Buildings Mass Facades Aerial Views Blocks and Lots Recursive block subdivision 93 Modeling of Urban Spaces The traditional urban modeling pipeline Input Major Minor Blocks Lots Buildings Mass Facades Aerial Views Blocks and Lots Recursive block subdivision Voronoi diagram based subdivision 94 47

48 Modeling of Urban Spaces The traditional urban modeling pipeline Input Major Minor Blocks Lots Buildings Mass Facades Aerial Views Blocks and Lots Recursive block subdivision Voronoi diagram based subdivision Paris like blocks 95 Modeling of Urban Spaces The traditional urban modeling pipeline Input Major Minor Blocks Lots Buildings Mass Facades Aerial Views Buildings 96 48

49 Modeling of Urban Spaces The traditional urban modeling pipeline Input Major Minor Blocks Lots Buildings Mass Facades Aerial Views Buildings Shape/split grammars Mass/façade modeling 97 Modeling of Urban Spaces The traditional urban modeling pipeline Input Major Minor Blocks Lots Buildings Mass Facades Aerial Views Buildings Shape/split grammars Mass/façade modeling Build by numbers 98 49

50 Modeling of Urban Spaces The traditional urban modeling pipeline Input Major Minor Blocks Lots Buildings Mass Facades Aerial Views Buildings Shape/split grammars Mass/façade modeling Build by numbers Image based synthesis 99 Modeling of Urban Spaces The traditional urban modeling pipeline Input Major Minor Blocks Lots Buildings Mass Facades Aerial Views Buildings Shape/split grammars Mass/façade modeling Build by numbers Image based synthesis User friendly Graph editing

51 Example: Manhattan Example: Look-a-like Venice

52 Example: Procedural Pompeii 103 Example: Petronas Towers

53 skylineengine 105 skylineengine Engine: : Traditional Pipeline The traditional pipeline is not enough for a complete complex city. Input Major Minor Blocks Lots Buildings Mass Facades Aerial Views As an example, we can have parks at different levels in the pipeline. This results in an Acyclic Directed Graph (DAG)

54 Configuration Files Image Maps skylineengine Pipeline Parcs Traditional Pipeline Districts Major Minor Blocks Lots Buildings Mass Facades Aerial Views Street Content Road Maps buildingengine Landmarks Bridges 107 skylineengine Engine: Input In an XML file: For synthetic cities: Avenues pattern or district map (different patterns) Land use map Heights map For real cities: City map (image) OpenStreetMap

55 skylineengine Engine: Districts Districts with different patterns 109 skylineengine Engine: Major streets Grid Pattern Eixample Barcelona Radial Pattern Karlsruhe (Alemanya)

56 skylineengine Engine: Major streets Organic Organic Pattern (voronoise( voronoise) We use a Voronoi variant (Llyod( Llyod s algorithm) Density can come both from a map or from an analytic function São Paolo 111 skylineengine Engine: Major streets Image processing Margin selection Binarization If we have a district map: We do the same processing to get the avenues between districts

57 skylineengine Engine: Major streets If we have a land-use map Show were we cannot build buildings 113 skylineengine Engine: Minor streets Number of minor roads proportional to the block area Create a grid with this number of divisions Rotate grid Do an intersection between both

58 skylineengine Engine: Buildings Height map 115 skylineengine Engine: Results - Pipeline

59 skylineengine Engine: Results - Manhattan-like 117 skylineengine Engine: Results - Girona Synthetic Buildings

60 Modeling of Buildings 119 Buildings: CGA Shape Production process: Rule-driven modification & replacement of shapes Iteratively evolve a design by creating more and more details Sequential application (like Chomsky grammars)

61 Modeling of Facades Approach: Split grammars Used instead of L-systemsL L-systems simulate growth in open spaces (better for plants and road networks) Buildings have stricter spatial constraints and their structure does not reflect a growth process 121 Modeling of Facades Goal: from a target design, create a desired abstraction as a set of grammar rules

62 CreditanstanltI1 Modeling of Facades Goal: from a target design, create a desired abstraction as a set of grammar rules 123 Modeling of Facades Facade Subdiv( Y,3.5,0.3,1r) { firstfloor ledge floors} Floors Repeat( Y,3){floor}

63 Modeling of Facades floor Repeat( X,tile_width,tile_width){ Tile } = 125 Modeling of Facades Image-based Procedural Modeling of Facades Müller, Zeng, Wonka,, Van Gool SIGGRAPH

64 Modeling of Facades Input: Rectified R photograph Output: 3D model and shape grammar rules of the facade 127 User-friendly graph editing for procedural buildings

65 User-friendly graph editing The regular procedure for Procedural Building Modeling is to generate a text file with the rules, 129 User-friendly graph editing F L L a a C F L Initial Shape C Legend Shape Terminal Shape Component Split Split Repeated Splits

66 User-friendly graph editing and then apply them iteratively,, generating the building 131 User-friendly graph editing resulting in a large tree even for small buildings L F a a a L a a

67 User-friendly graph editing that can be really large! 133 User-friendly graph editing The current trend is a shift in the modeling paradigm based on two realizations

68 User-friendly graph editing First, if the ruleset IS a graph of operations 135 User-friendly graph editing why not directly use it? F L L a a C F L C F L L a

69 User-friendly graph editing And second 137 User-friendly graph editing instead of taking every element, transforming and replacing it by its product element

70 User-friendly graph editing which would result in a large production tree 139 User-friendly graph editing which would result in a large production tree

71 User-friendly graph editing why don t t we take the whole set of elements 141 User-friendly graph editing and stream them through each operation? INSERT

72 User-friendly graph editing Combined, they form an extremely powerful system! 143 Modeling of buildings

73 Example: Raccolet House 145 Muchas gracias! Raccolet House Girona Manhattan

74 Jornadas sobre estrategias de generación de entornos colaborativos virtuales Mañana ana: Modelado procedural mediante programación visual Gustavo Patow Geometry and Graphics Group (GGG) Universitat de Girona (UdG( UdG)