Groningen Twister

1
Groningen Twister

• Fabian ScheurerFederal Institute of Technology
  Zurich

2
The project

• Stadsbalkon Groningen
• Public square in Groningen by the train station
• Two layers
• semi-underground, for parking 3000 bicycles
• public square on top, only accessible by
  pedestrians
• No level surfaces in the design
• Miscellaneous incisions and holes
• Architecture KCAP (Kees Christianse), Rotterdam
• Engineering Arup, Amsterdam

3
Demands for supporting structure

• The engineers
• Regular grid of columns!
• 30 cm column diameter!
• 6 m column spacing!
• The architects
• A bit of freedom!
• Varying column profiles!
• Varying column spacing!
• The design
• Holes and incisions in the pedestria level limit
  column placement!
• No level surfaces all columns are somehow
  tilted!
• Bicycle- and pedestrian paths not to be
  obstructed!

4
Brain-twisting task

• Place an adequate number of columns...
• with 3 different profiles, and therefore...
• with 3 different load bearing capacities, and...
• with a maximum tilt of 10 from the vertical
• So that...
• the number of columns in minimized
• the bearing load of the pedestrian level is
  evenly distributed
• the bicycle- and walkways are not obstructed

5
Idea Make the columns mobile!

• Physical model of the mobile column
• Two masses at the top and bottom end
• Moveable within the respective planes defined by
  floor and ceiling
• Linked by an elastic column
• Top mass surrounded by a bumper to keep
  appropriate distance to other columns
• Let them get organised by themselves!

6
Programming Java3D

• Java 3D API
• On MS Windows, SGI and Sun
• Test simple gravity-simulation
• 100 columns
• Attracting each other
• Good performance
• OK, we can handle it.

7
First implementation

• Three different column types
• 9 parameters for each type
• Bicycle stands to attract columns

8
Additional constraints

• Outline of pedestrian level
• boundary for column tops
• Holes in pedestrian level
• not allowed for column tops
• Bikestands and roundabout
• attractors for colum-bases
• Areas without cellar
• No columns necessary

9
Current version

• 3 column types
• 9 parameters each
• 3 attractor types
• Columns (among themselves)
• Bike stands
• Slab (with holes and walls)
• 7-8 parameters each
• Preset viewpoints
• Export functions
• Graphic (SVG)
• Text (table of column pos.)

10
Up to now rapid prototyping

• Straight forward programming
• Not very elaborate class model
• Extension (new constraints) is difficult
• Brute force computing
• No optimization in collision detection
• Necessary computing time grows quadratic with
  number of columns
• Limits are reached quickly

11
New constraints

• Translucent panels
• Linear bearings
• Expansion joints

12
To do optimization

• Introduce lean and consistent class model
• Easy extension of existing software
• construction kit for similar problems
• Optimize algorithms
• Fast update of neighborhood-relations
• Faster Collision detection
• Reach min. 5 frames per second on an average
  workstation

13
To do new concepts

• Columns grow (change type 1-2-3)
• if there is enough space
• Columns split
• If they cant grow any further (type 3)
• Columns shrink
• If they lack space
• Columns die
• If they cant shrink any further (type 1)

14
A colony of columns

• Growth within habitat
• Define different areas in the habitat which
  enable/restrict growth
• Columns grow to appropriate size
• Genetic programming
• Encode all column parameters into virtual DNA
• Define fitness criteria
• Plant 100 genetically different columns
• Select fittest DNA of this generation
• Mate with 100 random columns
• Start next generation

15
Thank you for your attention!

• Fabian Scheurer
• T 41 (0)1 633 40 25
• F 41 (0)1 633 10 50
• scheurer_at_arch.ethz.ch

• Federal Institute of Technology Zurich
• Professorship for CAAD
• Prof. Ludger Hovestadt
• ETH Hönggerberg HIL E.15
• CH-8093 Zürich