Interactive Design of Urban Spaces using Geometrical and Behavioral Modeling

1,870 views

Published on

Presentation slides.

Carlos A. Vanegas, Daniel G. Aliaga, Bedrich Benes, Paul Waddell, “Interactive Design of Urban Spaces using Geometrical and Behavioral Modeling”, ACM Transactions on Graphics (Proceedings SIGGRAPH Asia), 28(5), 2009.

0 Comments
1 Like
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total views
1,870
On SlideShare
0
From Embeds
0
Number of Embeds
8
Actions
Shares
0
Downloads
62
Comments
0
Likes
1
Embeds 0
No embeds

No notes for slide

Interactive Design of Urban Spaces using Geometrical and Behavioral Modeling

  1. 1. 1<br />
  2. 2. Interactive Design of Urban Spaces using Geometrical and Behavioral Modeling<br />Carlos Vanegas1 Daniel Aliaga1<br />Bedřich Beneš2Paul Waddell3<br />1Computer Science, Purdue University, USA<br />2Computer Graphics Technology, Purdue University, USA<br />3City and Regional Planning, UC Berkeley, USA<br />2<br />
  3. 3. Objective and Applications<br />Objective<br /><ul><li>Fast generation of 3D urban models that reflect the behavior of real-world cities</li></ul>Applications<br /><ul><li> Urban planning
  4. 4. Urban visualization
  5. 5. Policy evaluation and education
  6. 6. Content generation for games and movies</li></ul>3<br />
  7. 7. Examples (1)<br />Designer increases height of buildings in downtown<br />Number of jobs increases<br />Population increases<br />New housing appears in accessible areas<br />4<br />
  8. 8. Examples (2)<br />Designer inserts a highway into the model<br />Accessibility increases in formerly remote areas<br />Population redistributes<br />Location of jobs, houses, and buildings changes<br />5<br />
  9. 9. Observation<br /> Reduction of design time Production of plausible urban models<br />Behavioral Modeling<br />Geometrical Modeling<br />6<br />
  10. 10. System Overview<br />Geometrical Modeling <br /><ul><li> Roads
  11. 11. Blocks
  12. 12. Parcels
  13. 13. Buildings </li></ul>3D Model<br /> Behavioral Modeling<br /><ul><li> Population
  14. 14. Jobs
  15. 15. Accessibility
  16. 16. Land Value</li></ul>User<br />7<br />
  17. 17. System Overview<br /><ul><li>Input: Interactive design interface to change and constrain values of simulation variables</li></ul>Geometrical Modeling <br /><ul><li> Roads
  18. 18. Blocks
  19. 19. Parcels
  20. 20. Buildings </li></ul>3D Model<br />User<br /> Behavioral Modeling<br /><ul><li> Population
  21. 21. Jobs
  22. 22. Accessibility
  23. 23. Land Value</li></ul>8<br />
  24. 24. System Overview<br /><ul><li>Process: Interactions between the variables are continually simulated to bring the system back into equilibrium</li></ul>Geometrical Modeling <br /><ul><li> Roads
  25. 25. Blocks
  26. 26. Parcels
  27. 27. Buildings </li></ul>3D Model<br /> Behavioral Modeling<br /><ul><li> Population
  28. 28. Jobs
  29. 29. Accessibility
  30. 30. Land Value</li></ul>User<br />9<br />
  31. 31. System Overview<br /><ul><li>Output: Procedurally generated 3D urban model (based on the state of the system)</li></ul>Geometrical Modeling <br /><ul><li> Roads
  32. 32. Blocks
  33. 33. Parcels
  34. 34. Buildings </li></ul>3D Model<br /> Behavioral Modeling<br /><ul><li> Population
  35. 35. Jobs
  36. 36. Accessibility
  37. 37. Land Value</li></ul>User<br />10<br />
  38. 38. Related Work<br /><ul><li>Procedural Modeling of Cities, Buildings, Facades, Streets
  39. 39. Müller et al., 2001, 2006
  40. 40. Wonka et al., 2003
  41. 41. Chen et al., 2008
  42. 42. Aliaga et al., 2008</li></ul>11<br />Geometrical Modeling <br /> Behavioral Modeling<br />
  43. 43. Related Work<br /><ul><li>Large crowd modeling
  44. 44. Sung et al., 2004
  45. 45. Treuille et al., 2006
  46. 46. Flocking and animal behavior modeling
  47. 47. Reynolds., 1987
  48. 48. Tu and Terzopoulos., 1994 </li></ul>12<br />in Graphics<br />Geometrical Modeling <br /> Behavioral Modeling<br />
  49. 49. Related Work<br /><ul><li>Cellular automata, Agent-based
  50. 50. Microsimulation
  51. 51. Uses discrete-choice models
  52. 52. Uses agents that make decisions to locate and move within urban space
  53. 53. e.g., UrbanSim [Waddell et al., 2002]</li></ul>13<br />in Urban Simulation<br />Geometrical Modeling <br /> Behavioral Modeling<br />
  54. 54. Related Work<br />14<br />Integration<br />Geometrical Modeling <br /><ul><li> Interactive geometric simulation of cities over time
  55. 55. Weber et al., 2009
  56. 56. Visualization of Simulated Cities
  57. 57. Vanegas et al., 2009</li></ul> Behavioral Modeling<br />Do not focus on producing interactive design/editing time and ensuring plausible 3D urban models<br />
  58. 58. Contents<br /><ul><li>Introduction
  59. 59. Urban Design as a Dynamical System
  60. 60. Behavioral Modeling
  61. 61. Geometrical Modeling
  62. 62. Results</li></ul>15<br />
  63. 63. Urban Design as a Dynamical System<br /><ul><li> System
  64. 64. Consists of N variables defined over a spatial domain
  65. 65. Each variable sampled over a 2D spatial grid G of sizeWx H
  66. 66. vk(i,j) denotes the value ofk-thvariable at grid cell (i,j)</li></ul>j<br />Grid<br />i<br />v0(i,j), v1(i,j), … , vN (i,j)<br />16<br />H<br />W<br />
  67. 67. Urban Design as a Dynamical System<br /><ul><li>Minimal set of design variables used in our system</li></ul>17<br /><ul><li> Population count
  68. 68. Job count
  69. 69. Accessibility
  70. 70. Land value
  71. 71. Road length
  72. 72. Building volume
  73. 73. Average tortuosity
  74. 74. Terrain elevation</li></li></ul><li>Urban Design as a Dynamical System<br /><ul><li> Variable modeling
  75. 75. The change in each variable vk(i,j)is represented as a differential equation
  76. 76. Iftheuserchanges a variable, thesystemiterativelyupdatesallother variables in ordertoreturnto a state of equilibrium</li></ul>18<br />
  77. 77. Urban Design as a Dynamical System<br /><ul><li> Iterative System (a classical formulation)</li></ul>In urban design, the term is difficult to express symbolically due to widespread dependencies<br />19<br />Instead, we propose algorithms for computing vk(i,j) or <br />
  78. 78. Urban Design as a Dynamical System<br /><ul><li> Actual dependencies between variables are:</li></ul>20<br /><ul><li> The system contains algorithms for modeling variables:</li></li></ul><li>Urban Design as a Dynamical System<br /><ul><li>Variable Modeling: The system considers these dependencies and for each variable:
  79. 79. Simulates its change as a function of other variables (behavioral variables)</li></ul>OR<br /><ul><li> Calculates its target value vk(i,j)as function of other variables, and procedurally generates the geometry that matches the target values (geometrical variables)</li></ul>21<br />
  80. 80. System Overview<br />Geometrical Modeling <br /><ul><li> Roads
  81. 81. Blocks
  82. 82. Parcels
  83. 83. Buildings </li></ul>3D Model<br /> Behavioral Modeling<br /><ul><li> Population
  84. 84. Jobs
  85. 85. Accessibility
  86. 86. Land Value</li></ul>User<br />22<br />
  87. 87. Behavioral Modeling<br />3D Model<br />Geometrical Modeling <br /><ul><li> Roads
  88. 88. Blocks
  89. 89. Parcels
  90. 90. Buildings </li></ul> Behavioral Modeling<br /><ul><li> Population
  91. 91. Jobs
  92. 92. Accessibility
  93. 93. Land Value</li></ul>User<br />23<br />
  94. 94. Behavioral Modeling<br /><ul><li>Variables
  95. 95. Population count
  96. 96. Job count
  97. 97. Accessibility
  98. 98. Land value</li></ul>24<br />
  99. 99. Behavioral Modeling<br /><ul><li>Population and Jobs count
  100. 100. Operations:
  101. 101. Remove a fraction of population/jobs from their current location (Mobility algorithm)
  102. 102. Locate population/jobs to predicted locations (Location choice algorithm)
  103. 103. Each unit of population/jobs is referred to as an agent</li></ul>25<br />
  104. 104. Behavioral Modeling<br /><ul><li>Operation: Remove agents (Mobility Algorithm)
  105. 105. Agents de-located from grid are moved to queue Q
  106. 106. Agents added by user are moved to queue Q</li></ul>Grid<br />User addspopulation/jobs<br />26<br />
  107. 107. Behavioral Modeling<br /><ul><li>Operation: Locate agents (Location Choice Algorithm)
  108. 108. Agents from queue are placed back in the grid</li></ul>Grid<br />27<br />
  109. 109. Behavioral Modeling<br /><ul><li>Operation: Locate agents (Location Choice Algorithm)
  110. 110. Uses weighted attractiveness measure as a probability
  111. 111. The probability qstthat an agent es will locate at the grid cell (it, jt) is given byat :accessibility at the grid celllt : land value at the grid cellTs : total count of the agent throughout the entire grid</li></ul>28<br />
  112. 112. Behavioral Modeling<br /><ul><li> Accessibility
  113. 113. Measure of access that a grid cell has to jobs and to the rest of the population
  114. 114. Intuitively
  115. 115. Decreases with increasing terrain slope
  116. 116. Increases with higher road connectivity and nearby population/jobs</li></ul>29<br />
  117. 117. Behavioral Modeling<br /><ul><li> Accessibility
  118. 118. Represented by logistic function
  119. 119. u1: proximity to highways, arterials and streets
  120. 120. u2: local slope of the terrain
  121. 121. u3: Distance-normalized measure of activity level at (i,j) </li></ul>30<br />
  122. 122. Geometrical Modeling<br />Geometrical Modeling <br /><ul><li> Roads
  123. 123. Blocks
  124. 124. Parcels
  125. 125. Buildings </li></ul>3D Model<br /> Behavioral Modeling<br /><ul><li> Population
  126. 126. Jobs
  127. 127. Accessibility
  128. 128. Land Value</li></ul>User<br />31<br />
  129. 129. Geometrical Modeling<br /><ul><li>Variables
  130. 130. Road length
  131. 131. Average tortuosity
  132. 132. Building volume
  133. 133. Terrain elevation at grid cell (user-controlled)</li></ul>32<br />
  134. 134. Geometrical Modeling<br /><ul><li> Road length
  135. 135. Total length of roads in grid cell
  136. 136. Two types: Arterials and Streets
  137. 137. Average Tortuosity
  138. 138. Ratio between road segment length and distance between segment endpoints</li></ul>33<br />
  139. 139. Geometrical Modeling<br /><ul><li>Arterials and Streets: Seeds
  140. 140. To connect the main population clusters a set of seeds is generated considering the population/jobs distribution and the location of highways</li></ul>34<br />
  141. 141. Geometrical Modeling<br /><ul><li>Arterials and Streets: Expansion of Arterials
  142. 142. Each seed is used as an intersection of the arterial roads network and used to generate arterial segments</li></ul>35<br />
  143. 143. Geometrical Modeling<br /><ul><li>Arterials and Streets: Expansion of Streets
  144. 144. Street seeds are generated along arterial road segments and used to create streets</li></ul>36<br />
  145. 145. Geometrical Modeling<br /><ul><li>Arterials and Streets: Expansion Algorithm
  146. 146. Using the seeds positioned according to population/jobs and/or along arterials, we generate road segments using a breadth-first expansion method
  147. 147. These seeds are placed into a queue</li></ul>37<br />
  148. 148. Geometrical Modeling<br /><ul><li>Arterials and Streets: Expansion Algorithm</li></ul>38<br />
  149. 149. Geometrical Modeling<br /><ul><li>Arterials and Streets: Expansion Algorithm</li></ul>39<br />
  150. 150. Geometrical Modeling<br /><ul><li>Arterials and Streets: Expansion Algorithm</li></ul>40<br />
  151. 151. Geometrical Modeling<br /><ul><li>Arterials and Streets: Expansion Algorithm</li></ul>41<br />
  152. 152. Geometrical Modeling<br /><ul><li>Arterials and Streets: Expansion Algorithm</li></ul>42<br />
  153. 153. Geometrical Modeling<br /><ul><li>Arterials and Streets: Expansion Algorithm</li></ul>43<br />
  154. 154. Geometrical Modeling<br /><ul><li>Arterials and Streets: Expansion Algorithm</li></ul>44<br />
  155. 155. Geometrical Modeling<br /><ul><li>Arterials and Streets: Expansion Algorithm</li></ul>45<br />
  156. 156. Geometrical Modeling<br /><ul><li>Arterials and Streets: Expansion Algorithm</li></ul>46<br />
  157. 157. Geometrical Modeling<br /><ul><li>Arterials and Streets: Expansion Algorithm</li></ul>47<br />
  158. 158. Geometrical Modeling<br />Grid<br />Radial<br />Tortuosity<br />48<br />
  159. 159. Geometrical Modeling<br /><ul><li> Building volume
  160. 160. Total volume of all the buildings in grid cell
  161. 161. Computed as a function of population and jobs
  162. 162. To generate a building geometry that matches the volume, we first generate parcels and building footprints</li></ul>49<br />
  163. 163. Geometrical Modeling<br /><ul><li>Parcels
  164. 164. Blocks are extracted from the road network and partitioned into parcels
  165. 165. The number of parcels in the block is proportional to the product of the area of the block and the count of population/jobs in the grid cells inside the block</li></ul>50<br />
  166. 166. Geometrical Modeling<br /><ul><li>Parcels</li></ul>Close up view<br />51<br />
  167. 167. Geometrical Modeling<br /><ul><li> Buildings
  168. 168. Procedurally generated inside each parcel based on the socioeconomic information of the area
  169. 169. Process:
  170. 170. Calculate geometry of the building footprint
  171. 171. Calculate building height
  172. 172. Use procedural rules to generate 3D geometry that matches these attributes</li></ul>52<br />
  173. 173. Results<br />Geometrical Modeling <br /><ul><li> Roads
  174. 174. Blocks
  175. 175. Parcels
  176. 176. Buildings </li></ul>3D Model<br /> Behavioral Modeling<br /><ul><li> Population
  177. 177. Jobs
  178. 178. Accessibility
  179. 179. Land Value</li></ul>User<br />53<br />
  180. 180. Results: System Specs<br /><ul><li>Dimensions:
  181. 181. Grid-cell size 0.1x0.1 km
  182. 182. Grid size up to 50x50 km in our experiments
  183. 183. Time:
  184. 184. Update time step during editing:
  185. 185. <0.3 seconds for small grid (5x5 km)
  186. 186. <4 seconds for large grids (15x15 km)
  187. 187. Total Design Time:
  188. 188. <5 minutes for any of our examples</li></ul>54<br />
  189. 189. Results: Stability<br />55<br />Terrain and Roads<br />Equilibrium<br />Equilibrium<br />
  190. 190. Results: Terrain Editing<br />56<br />Example 1<br />Example 2<br />Lake<br />Mountains<br />Initial<br />
  191. 191. Results: Completion and Validation (1)<br />57<br />Town center<br />Population<br />Parcels<br />Parks<br />Jobs<br />Buildings<br />Terrain<br />Input<br />Output<br />
  192. 192. Results: Completion and Validation (2)<br />58<br />Real City<br />Synthetic City<br />
  193. 193. Conclusions<br /><ul><li> We have presented an interactive system to design and edit 3D urban models
  194. 194. Our key inspiration is to close the loop between behavioral modeling and geometrical modeling producing a single dynamical system that assists a designer in creating urban models</li></ul>59<br />
  195. 195. Conclusions<br /><ul><li> Limitations
  196. 196. Stochastic component that does not allow behaviors and geometries to be exactly repeatable
  197. 197. Over-constraining the system can lead to unfeasible urban models</li></ul>60<br />
  198. 198. Future Work<br /><ul><li> Including additional elements of behavioral modeling, such as a more sophisticated accessibility/land value model
  199. 199. Adding methods to generate more complex geometric structures using socioeconomic data (e.g., additional building details such as facades)</li></ul>61<br />
  200. 200. Acknowledgements<br /><ul><li> Purdue Research Foundation for partially funding this work
  201. 201. Aaron Link for his help with exporting models
  202. 202. Students from the Computer Graphics and Visualization Lab at Purdue for their feedback
  203. 203. Reviewers for their valuable comments and suggestions</li></ul>62<br />
  204. 204. Thank you!<br />63<br />

×