Emergence by Rebecca Bradley

EMERGENCE SEMINAR 2013/14 Mafe Chaparro

Rebecca Bradley

Anja Hein

Lei Zheng

03. INTRODUCTION 04. SEQUENCE 1 Breeding Strategy Generation 2 Generation 3 Mutation Strategy Evaluation 08. SEQUENCE 2 Gene Pool Body Plan Killing Strategy Generations 4 Generation 5 Generation 6 AnalysisSEQUENCE 3 Shibam, Yemen Assessment Design Strategy Generation 1-3 Analysis SEQUENCE 4 Adjustments Fitness Criteria Case 1 Case 2 Results Case 3 Analysis CONCLUSION

EMERGENCE SEMINAR 2013/14

INTRODUCTION The following experiments will explore the potential for applying the biological patterns of genetic behaviour to the computational development of an urban block. Throughout the study, the evolution of a primitive shape will be observed through four sequences, each consisting of several generations to produce variation, and every generation will be composed of a population of individuals. The individuals will possess a genetic makeup of four genes, each gene a physical instruction that communicates a change in the individualâ&#x20AC;&#x2122;s form. Each population of individuals will undergo fitness rankings based on specific criteria, elimination of individuals based on a killing strategy, and the employment of a breeding strategy in order to genetically produce the next generation. The power and performance of the gene pool will be observed and measured through the introduction of a genetic mutation. Ultimately, an evolutionary goal will be pursued through the application of these genetic techniques to a tower block, to explore the urban, architectural, and spatial potentials of a system that generates random variation from a set of rules. This experiment aims to simulate the intelligence that occurs in natural selection and apply the advantages of genetic variation to the production of strongly fit individuals in a highly diverse population.

EMERGENCE SEMINAR 2013/14

SEQUENCE 1 In order to produce the first generation of sequence 1, a set of instructions were followed as “genes” in order to develop the primitive - a pyramid - into ten new individuals. There were four genomes: scale, copy, mirror, and rotate, applied at varying predefined intensities. An object could be scaled by a factor of 2 or 3, for instance, or rotated at varying degrees, and other genes which executed the mirror command required specification of parameters like axis. Scaled individuals were consistently scaled on the normal to the base BODY PLAN of the pyramid.

NERATION 4

MITIVE

primitive are: mirroring about the Z axis on the individual’s base, copy the unit 1 unit in the X direction, rotating the new element by 45 degrees about the XY plane, and scaling (or stretching) the element 1 unit along the Z axis.

Certain genes contributed to an individual’s likelihood GE to have a higher fitness ranking. For instance, scaling SEG up and mirroring in the Z direction produced individuals with higher fitness rankings, while scaling down FITNESSmir mo or rotating 90 degrees had the tendency to produce 2 rotP SHADOW individuals with lower fitness rankings. sc

ON THE G

BREEDING STRATEGY A breeding method was chosen in which the first two instructions of the least fit (shortest) individuals would breed with the last two instructions of the most fit (tallest) individuals. The killing strategy targeted the KILLING elimination of individuals with exposure of the base in relation to the point. [figure 1.0]

3 As such, each of the ten individuals in Generation 1 is composed of four of these genes. To illustrate the standard for genetic composition, the example of the Breeding Strategy first and most fit individual in the generation will be deconstructed (G01.01). The four genes applied to the

NES POOL

NTS 1 AXIS

which individuals will survive

45°

90°

in what order genes will be exchanged

Element

BY MEASURING T PROJECTED ON T

GROWTH

INTENSITY

45°

90°

Z Z

Generation 1 - breeding

01 02 03 04 05 06 07 08 09 10

NTS 3 AXIS XY

MUTATION

INTENSITY 1

Generation 2

45°

90°

01 02 03 04 05 06 07 08 09 10 11

Z figure 1 1.0 :Breeding 2 strategy3: The most fit (or tallest) individuals then had their genetic sequencing combined with the least fit (or shortest) individuals. 4

EMERGENCE SEMINAR 2013/14 AXIS

mir

Genes 1-4

AXIS

SEG

BREEDING

01 02 03 04 05 06 07 08 09 10

FITNESS CRIT

01 02 03 04 05 06 07 08 09 10

rot

These individuals were then ranked from most to sc least fit based on the first fitness criteria, height. The SEG normal distribution of the first generation indicates that there was a good amount of variation, showing mir mo a typical number of average fitnesses, with outliers or rot “monsters” on both ends of the heightSHADOW spectrum. PROJEC sc

which individuals will breed

mir

01 02 03 04 05 06 07 08 09 10

INTENSITY

SEG

rot

Generation 1

NTS 2

NTS 4

INTENSITY

Z XY

EVOLUTIO

0.4 0.35 0.3 0.25 0.2 Gen 1

0.15 0.1 0.05 0 0

Component height figure 1.1 : Standard Deviation for Generation 1 (S1G1)

figure 1.2 : Individuals generated from S1G1 5

EMERGENCE SEMINAR 2013/14

GENERATION 2 The genes from generation 1 were recombined using the following breeding strategy in order to produce generation 2. Selected individuals as shown were combined using two methods of breeding. Two genes from each individual were combined in both methods. In the first method, the first two genes of the first individual are combined with the last two genes of the second individual. In the second method, the middle two genes of each individual are combined. A number of individuals from generation 1 were also chosen to survive directly into the second generation. The individuals of generation 2 were ranked once again according to the same fitness criteria. Overall fitness in the second generation improved, and there was sufficient variation in the individuals. Generation 2 had less variation than Generation 1, but higher overall fitness of the individuals. GENERATION 3 To produce the third and final generation of Sequence 1, the killing strategy for generation 2 was revised, now targeting individuals with the least “stability” with regards to their perceived ability to stand up as a 3 dimensional object. For instance, an individual that is a rotated primitive pyramid about the XY plane (standing on its point) would not balance in reality and was consider least fit, as compared to an object standing on a flat base. This not only allowed for more variation in the population but an increase in varying fitness criterias in thinking forward towards an evolutionary goal. By ranking the individuals based on their height, the genome became complex enough to achieve interesting variation by preserving other physical qualities of the individuals. Genetic instructions affecting various aspects of the individual’s appearances were carried

through each generation. The execution of four instructions on a simple primitive, and the observed effects of selection and breeding provided a fundamental understanding necessary to move forward into sequence 2. The goal for the first generation of the following sequence would favour the least surface area of shadow on the ground, having minimal impact on the site around it but still offering the same volume. The fittest individuals in next generation might be those which exhibit mutation for less shadow, but are taller structures. With respect to the evolutionary goal, this would dictate future killing strategies. MUTATION STRATEGY A mutation was to be introduced to the gene pool. Being as the mutation strategy was directly related to achieving the evolutionary goal, the effects of each mutation type on the population were predicted in order to strategically choose a mutation. Mutation as Deletion could remove a gene, leading to a greater surface area inside the objects. This would have little to no effect on the evolutionary goal. Mutation as Duplication could be beneficial by repeating a step that minimises the structure’s footprint. This, along with inversion, were taken into consideration. Inversion could be applied in the sense of making an instruction work backwards or counteract itself, for instance, instead of doubling in size it would be cut in half. Mutation as Insertion might not be useful, essentially just rearranging the instructions. This would only be useful if the order of instructions is relevant, but since its application would happen rather randomly, this can’t be as controlled in the way other mutations could. Ultimately duplication was chosen as the mutation type, and it was applied to only one gene, at the end of the sequence of four instructions.

0.4 0.35 0.3

0.25

0.2

0.15

Gen 3

0.15

0.1

Gen 2

0.05

0.1 0.05 0

0 0

Component height

figure 1.3 : Normal distribution graph for S1G2 6

EMERGENCE SEMINAR 2013/14

-0.05

Component height

figure 1.4: Normal distribution graph for S1G3

Generation 1

Do Not Continue

G1.01

G1.04

G1.10

G1.07

G1.08

G1.09 +

G1.02

G1.05

G1.03

G1.06

Continued from G1

Generation 2

G2.02

G2.03

G2.04

G2.05

G2.06

G2.07

G2.08

G1.04

G2.04

G1.10

G2.03

G1.09

G2.10

G1.02

G2.11

G2.10

G2.11

G2.09

Generation 1+2

G2.01

G2.02

Generation 3

Carried on from Past Generations

G3.01

G3.02

G3.03

G3.04

G3.05

G3.06

G3.07

G3.08

G3.10

G3.11

G3.09

figure 1.5. :Breeding strategy of generation 1-3

0.4 0.35 0.3 0.25 0.2

Gen 1 Gen 2

0.15

Gen 3 0.1 0.05 0 0 -0.05

Componet height figure 1.6 : Normal distribution graph for Sequence 1, generation 1-3 7

EMERGENCE SEMINAR 2013/14

GENERATION 4

SEQUENCE 2 Sequence 2 consists of Generation 4 to 6. Generation over these segments randomly. 4 (G4) population is randomly generated by the gene BODY PLAN 1 pool; Generation 5 (G5) and Generation 6PRIMITIVE (G6) we created a new population by performing a crossover Mutation breeding strategy, mutation and killing strategies ac- Strategy cording to our fitness criteria to improve the individuals of each iteration.

Size o

GENE POOL 3 4 The gene pool used in this sequence consists of all of Mutation creates variation GENERATION 4 figure 2.1 : Body plan for S2 the genes used in Sequence 1 but with a series of inUnfavourable mutations selected against GENES F tensities. In this case the gene pool will consist of POOL : KILLING STRATEGY PRIMITIVE mirror, move, rotate and scale (Fig 2.1), eachSEGMENTS gene 1has Each of the ten are made out of four genes; SH Reproduction andindividuals mutation occur a variation of axis and intensity in order to achieve theAXIS INTENSITY each of these genes will be arranged randomly as well Favourable mutations morewill likelyaffect. to survive desire fitness criteria. as the segments they After obtaining the mirror XY 1 GENERATION 4 move Z 1 2 3 discard the ones in which any of its segindividuals we And reproduce Goal 45° 90° To be able to create individuals that has therotate less im- Z ments was disconnected to theGENES rest POOL of the body as our B 1 2 3 P pact on their surroundings, for this context:scale the one Z killing strategy (Fig 2. 3). FITNESS CRITERIA that projects less shadow on the ground. The intensiBODY PLAN PRIMITIVE 1 2 2 SHADOW PROJECTED Mutation SEGMENTS ties established are emphasize in vertical growth thisStrategy ON THE GROUND PredictingAXIS of mutation INTENSITY is why the genes move and scale are being pursued in the effect Size of mutation to be applied after 4 gene inst mirror 1 No e ect XY Axis dependant Great e ect the z direction. BODY PLAN

SEGMENTS 1

AXIS

INTENSITY

mirror

move

rotate

45°

90°

scale

SEGMENTS 2

AXIS

move

Z 1 2 Mutation creates variation Z 45° 90°

rotate

INTENSITY

mirror

move

rotate

45°

90°

scale

BODY PLAN mutations selected against KILLING STRATEGY scale ZUnfavourable 1 2 We draw our body plan for the primitive used in SeReproduction and mutation occur Gene mutated Shadow size Vertical stability SEGMENTS 3 3 of Favourable 4mutations more likely to survive quence 1; the body plan consists on the subdivision INTENSITY the pyramid in the vertical and horizontal axis, result-AXISAnd reproduce Goal: introduce a mutation that may become d mirror XY 1 ing in four segments. The genes are going to Rotate operate SEGMENTS 3

AXIS

INTENSITY

mirror

move

rotate

45°

90°

scale

SEGMENTS 4

GENES POOL

move

SEGMENTS 1

rotate Z 45° 90° Predicting the effect of mutation scale Z 1 2 3 No e ect Axis dependant Great e ect

AXIS XY

move

rotate

45°

90°

scale

SEGMENTS 2

move

rotate

45°

90°

scale

AXIS

Scale

mirror

move

rotate

45°

90°

scale

rotate

45°

90°

scale

Vertical stability

INTENSITY

BY MEASURING THE SHADOW AREA PROJECTED ON THE GROUND.

Rotate move

rotate

45°

90°

GROWTH STRATEGY

Move

INTENSITY

move

Shadow size

Scale

SEGMENTS 3

mirror

Insertion Gene mutated Rotate Vertical stability Mirror Scale Move

Duplication Inversion Duplication Gene mutated Shadow size Vertical stability Gene mutated Shadow size Gene Shadow size Vertical Rotate mutated Rotate stability Mirror Mirror Rotate Scale Scale Mirror Move Move Scale Move

Inversion Vertical stability Gene mutated Rotate Mirror Scale Move

Move

INTENSITY

mirror

move

rotate

45°

90°

scale

figure 2.2 : Gen pool for S2 8

EMERGENCE SEMINAR 2013/14

MUTATIONShadow STRATEGY size Vertical stability

Deletion Gene mutated Shadow size Vertical stability Rotate Deletion Insertion Gene mutated Shadow size Vertical stability Gene mutated Shadow size Mirror Rotate Rotate Scale Mirror Mirror Scale Scale Move

Move

SEGMENTS 4

SHADOW PROJECTED BREEDING STRATEGY

Gene mutated

Move

INTENSITY

mirror

AXIS

INTENSITY

mirror

SEGMENTS 4

scale Mirror

AXIS

Mirror

INTENSITY

mirror

AXIS

FITNESS CRITERIA AXIS

figure 2.4 : Mutation setups and strategies

Deleti

Deletion - remove an instruction that causes gre

Duplic

Duplication - useful to repeat a scale down

Inserti Insertion - not useful, rearranges an instruction? Shadow size

Vertical stability

Inversion - could usefully reverse an instruction EVOLUTIONARY GOAL Invers

GENERATION 4 PRIMITIVE

BODY PLAN

MOST FITTED

LESS SHADOW PROJECTED

GENES POOL

FITNESS CRITERIA

SEGMENTS 1

SHADOW PROJECTED

FITNESS CRITERIA AXIS

mirror

INTENSITY

move

rotate

45°

90°

scale

SHADOW PROJECTED ON THE GROUND

LEAST FITTED

SEGMENTS 2 AXIS

INTENSITY

mirror

move

rotate

45°

90°

scale

KILLING STRATEGY

BY MEASURING THE SHADOW AREA PROJECTED ON THE GROUND.

21 JUNE 12 PM

MOST SHADOW PROJECTED

SEGMENTS 3

AXIS

INTENSITY

mirror

move

rotate

45°

90°

scale

21 DECEMBER 12 PM

SEGMENTS 4

FITNESS CRITERIA AXIS

INTENSITY

mirror

move

rotate

45°

90°

scale

SHADOW PROJECTED BREEDING STRATEGY

BY MEASURING THE SHADOW AREA PROJECTED ON THE GROUND.

GROWTH STRATEGY

INDIVIDUAL THAT CONTAINS ANY SEGMENT THAT IS SPREAD OUT MORE THAN ONE UNIT FROM THE WHOLE. EMERGENCE SEMINAR SEQUENCE 2

-BEST INDIVIDUAL OF GENERATION WILL BREED WITH THE 4 WORST. 21 JUNE

12WILL PM BREED RANDOMLY -THE REMAINING INDIVIDUALS WITH EACH OTHER.

TO ACHIEVE THE LEAST SHADOW IMPACT ON THE SITE

MUTATION STRATEGY

Z XY

-EMPHASIZE GROWTH IN VERTICAL - WE RESTRICT GENE MIRROR SO IT DOESNT HAVE INTENSITYES. -SCALE ONLY ON THE Z DIRECTION

21 DECEMBER 12 PM

20 % OF THE POPULATION OF EACH GENERATION WILL MUTATED.

EVOLUTIONARY GOAL

A FORM THAT HAS THE LEAST IMPACT ON ITS SURROUNDING

EMERGENCEfigure SEMINAR SEQUENCE 22 2.3 : Design Strategy for Sequence EMERGENCE SEMINAR SEQUENCE 2 9

EMERGENCE SEMINAR 2013/14

PRIMITIVE

G4.01

G4.02

G4.03

G4.04

G4.05

GENERATION 4 For this generation we discard only one individual and then we rank the remaining ones from the less shadow projected (most fitted) to the most shadow projected (less fitted). For this generation we have a population in which there are only few individuals that fit our fitness criteria as shows the standard deviation graph (Fig 2. 4). INDIVIDUALS G4.01

2,3882356

G4.10

1,8878672

0,3016768

G4.02

4,3004656

G4.06

2,0114474

0,3715717

G4.03

2,4010268

G4.09

2,3126739

0,5307893

G4.04

2,6011158

G4.01

2,3882356

0,5612443

2,4010268

0,5658134

2,6011158

0,6107741

G4.05

4.05

LESS SHADOW

G4.06

3,0994272 2,0114474

G4.07

G4.03 G4.04

G4.08

G4.09

G4.07

2,722564

G4.07

2,722564

0,6109171

G4.08

2,9017235

G4.07

2,9017235

0,5734214

G4.09

2,3126739

G4.08

3,0994272

0,4896152

G4.10

1,8878672

G4.02

4,3004656

0,0257193

MOST SHADOW

2,6626547

0,6502563

G4.10

table 2.1 : Results from Generation 4

0,7 0,6 0,5 0,4 0,3

Gen 4

0,2 0,1 0 0

Shadow area (mm^2) figure 2.7 : Normal distribution graph for Generation 4

EMERGENCE SEMINAR SEQUENCE 10

EMERGENCE SEMINAR 2013/14

figure 2.5 : Individuals in Generation 4

GENERATION 4

BREEDING STRATEGY RANKING

INDIVIDUALS

G4.01

TIVE

G4.02

G4.03

G4.04 VIDUALS

G4.05

GENES

LEAST SHADOW

XY1

Z45

XY1

Z90

XY1

G4.01

XY1

2,3882356 3 4

4,3004656

G4.06 03

2,4010268

Z90

N45

XY1

2,6011158

3,0994272 Z2 Z90

XY1

2,0114474 3 2

2,722564

G4.08 08

XY1 Z3 2,9017235

2,3126739

1,8878672

G4.09

G4.10

G4.02

G4.03

LESS SHADOW G4.10

G4.03

XY1

Z45

XY1

Z90

3 Z2

G4.07

G4.08

G4.09

G4.10

figure 2.6 : Shadows of individuals in Generation 4

GENERATION 4 0,3016768

2,0114474

0,3715717

2,3126739

0,5307893

G4.01

2,3882356

0,5612443

G4.03

2,4010268

0,5658134

2,6011158

0,6107741

2,722564

0,6109171

2,9017235

0,5734214

3,0994272

0,4896152

4,3004656

0,0257193

2,6626547

0,6502563

G4.09

G4.04

G4.07

G4.08

G4.08 G4.02 MOST SHADOW

G4.05

BREEDING STRATEGY

0,7 1,8878672

G4.07 Z90

G4.06

G4.01

Z90

G4.05

G4.09

G4.04

G4.06

G4.07

G4.10

MOST SHADOW

0,6 0,5

RANKING INDIVIDUALS G4.01

GENES

LEAST SHADOW

XY1

Z45

0,4

0,3 G4.02

XY1

Z90

XY1

G4.10

G4.06

Gen 4

0,2

G4.03 0,1 0 G4.04 0

XY1

Z90

G4.09

G4.01 5

Shadow area (mm^2)

G4.05

G4.06

G4.07

G4.08

G4.09

G4.10

XY1

N45

XY1

Z90

XY1

Z90

XY1

Z45

XY1

Z90

G4.03

EMERGENCE SEMINAR SEQUENCE SEQUENCE 2 2 EMERGENCE SEMINAR G4.04

3 Z3

G4.07

4 Z1

G4.08

Z90

G4.05

3 Z2

MOST SHADOW

figure 2.8 : Ranking for Generation 4

EMERGENCE SEMINAR 2013/14

GENERATION 5 In order to improve the population for the G5, we use the ranking chart (Fig 2.5) to evaluate our breeding strategy: crossover of 50/50 , in which we select the most fitted and breed it with the four less fitted, each pair will generate 2 individuals. We picked the genes of the best individual to pass them to the next generation to create a population most fitted. For this

BREEDING STRATEGY

GENERATION 5

GENE POOL

random mix of genes on random segments of the body plan mutation duplication of genes only on segment 1 and 2

RANKING

iteration 20% of the population is mutated by duplication where pair of genes are copied twice into the new individual, this in order to introduce a mutation that may become dominant with a new environmental condition (southern sun). In G5 2 individuals are discarded by the killing strategy and the 8 remaining tent to project slightly less shadow on the ground. (Fig 2.6)

INDIVIDUALS

LEAST SHADOW Z45

Z90

XY1

Z90

2 XY1

1 XY1

3 Z90

2 XY1

G4.10

G4.06

G4.10

G4.09

G4.10

G4.01

G4.10

G4.03

Z90

G4.05

G4.08

G4.07

---->

GENES

G5.01

---->

G5.02

---->

G5.04

---->

G5.05

G4.10

G4.04

G4.07

---->

G5.06

G4.10

G4.07

G4.04

---->

G5.07

G4.10

G4.08

G4.04

---->

G5.08

G4.06

G4.05

Z90

G4.03

---->

G5.09

MOST SHADOW

G4.09

G4.01

---->

Z45

XY1

G5.03

G5.10

MUTATED

Z90

XY1

Z45

XY1

Z45

XY1

Z45

XY1

Z90

XY1

Z45

Z90

PZB Z2

NXY45 Z45

NZ1 XY1

NXY45 Z3

PXY0 XY1

NZ45 Z45

NXY0 Z45

NZ45 Z1

XY1

Z45

figure 2.9 : Breeding strategy for Generation 5

EMERGENCE SEMINAR SEQUE

ENERATION 5

figure 2.10 : Individuals in Generation 5 BREEDING STRATEGY

GENERATION 5

RANKING INDIVIDUALSG5.01 GENES G5.01

G5.02

LEAST SHADOW G5.04 G5.03

-the genes of the highest individual in G5 breed with the shortest of G4 -G1.03 and G1.06 go to next generation

G5.05

G5.06

G5.07

G5.08

G5.09

Z45

XY1

G5.08

G4.10

G5.08

G4.10

G5.08

XY1

Z90

G5.10

G4.06

G5.10

G4.06

G5.10

G5.03

G4.09

G5.03

G4.09

G5.03

figure42.11 :3Shadows of individuals in Generation 5 2 4

G5.02 12

BREEDING FOR GENERATION 6

4 2 1 2 EMERGENCE SEMINAR 2013/14

G5.03 INDIVIDUALS

Z45

XY1

LESS SHADOW

0,7

G5.10

INDIVIDUALS

GENES

G5.01

G5.02

G5.03

G5.04

G5.05

G5.06

G5.07

G5.08

G5.09 G5.01

G5.10

.03

.04

.05

.06

.07

.08

.09

.10

Z45

XY1

G5.01

Z90

Z45

XY1

Z45

XY1

Z45

XY1

Z90

XY1

Z45

Z90

PZB Z2

NXY45 Z45

NZ1 XY1

NXY45 Z3

G5.02 XY1

-G1.03 and G1.06 go to next generation

LEAST SHADOW

XY1

G5.08

G4.10

G5.08

G5.10

G4.06

G5.10

G5.03

G4.09

G5.03

G5.05

G4.01

G5.05

G5.09

G4.03

G5.09

G5.02

G4.04

G5.02

G5.07

G4.07

G5.07

G5.01

G4.08

G5.01

2 XY1

MOST SHADOW

G5.03

G5.04

Z1 G5.02 Z2

Z45

G5.03 Z1

G5.05 G5.04

G5.06

G5.07

G5.05

G5.06

G5.08 G5.07

G5.09

G5.08

G5.09

figure 2.13 : Ranking for Generation 5

DIVIDUALS

.02

-the genes of the highest individual in G5 breed with the shortest of G4

RANKING

GENERATION 5

.01

BREEDING FOR GENERATION 6

BREEDING STRATEGY

GENERATION 5

LESS SHADOW INDIVIDUALS 3,0892771

G5.08

LESS SHADOW

G5.01

3,0892771

G5.08

G5.02

2,4755252

G5.10

G5.03

1,8970751

G5.06

G5.04

3,2456053

G5.03

G5.05

2,2506906

G5.05

G5.06

1,6837789

G5.09

G5.07

3,040625

G5.02

G5.08

1,4374517

G5.07

1,5969922 G5.10

G5.04 1,5969922

2,4755252 1,8970751 3,2456053 2,2506906 1,6837789 3,040625

1,4374517

2,3479185 G5.09

G5.10 G5.06 G5.03 G5.05 G5.09 G5.02 G5.07

G5.01 2,3479185

G5.01 G5.04

LESS SHADOWLESS SHADOW

0,7 1,4374517 1,5969922 1,6837789 1,8970751 2,2506906 2,3479185 2,4755252 3,040625

0,2420344

1,4374517

0,2420344

1,5969922

0,3347989

1,6837789

0,3885737

1,8970751

0,5159988

2,2506906

0,6378123

2,3479185

0,6389621

2,4755252

0,6172331

0,3347989 0,3885737 0,5159988 0,6378123 0,6389621 0,6172331

0,3198089 3,040625 0,3198089 3,0892771 3,0892771 0,2908026 0,2908026 3,2456053 3,2456053 0,2055844 0,2055844 2,306494 2,306494 2,306494 2,306494 0,6229811 0,6229811 0,6229811 0,6229811

0,7

0,6

0,5

0,4

0,3

0,2

0,1

0 0

table 2.1 : Results from Generation 5

00,5

0,51

1 1,5

1,5 2

2 2,5

2,5 3

3,5 3,5

area (mm^2) shadowshadow area (mm^2)

figure 2.12 : Normal distribution graph for Generation 5

EMERGENCE 13SEMINA EMERGENCE SEMINAR EMERGENCE SEMINAR 2013/14

GENERATION 6 We evaluate the ranking chart of G4 and G5 in order to establish a breeding strategy that would improve the population of the G6. The best individual of each iteration will breed with the three worst ones of the other (Fig 2.7) and will produce one individual per pair. A ranking between the second best of each generation is made in order to pick the best one of both iterations

to breed with the remaining individuals of the other generation. For this iteration 20% of the population is mutated by duplication. One individual is discarded and the other nine are ranked in order to evaluate the improvement on the population according to the fitness criteria and to compare them with the other generations of the sequence 2.

BREEDING FOR GENERATION 6

-the genes of the highest individual in G5 breed with the shortest of G4 -G1.03 and G1.06 go to next generation

G4.10

G5.08

G4.10

G5.08

G4.06

G5.10

G4.06

G5.10

G4.09

G5.03

G4.09

G5.03

G4.01

G5.05

G4.01

G5.05

G4.03

G5.09

G4.03

G5.09

G4.04

G5.02

G4.04

G5.02

G4.07

G5.07

G4.07

G5.07

G4.08

G5.01

G4.08

G5.01

figure 2.14 : Breeding strategy for Generation 6

RENDER GENEREATION 6 EMERGENCE SEMINAR LEI

ENERATION 6

SEQUENCE 2

figure 2.16 : Individuals in Generation 6 GENERATION 6 RANKING

GENERATION 6 INDIVIDUALS G6.01 INDIVIDUALS

G6.01 G6.02 G6.01

RANKING

GENES Z2

GENES4 Z2

XY1

Z90

LESS SHADOW

G6.02

Z2 Z2

XY1 Z45

Z2 Z45

Z2 Z2

Z2 Z45

XY3

Z45 Z2

Z90 Z2

G6.03

G6.04

1 1 2 figure 2.17 :4 Shadows of individuals in Generation 6

G6.02 G6.03

G6.03 G6.04

XY1 Z45

Z1 XY3

G6.01

G6.05 G6.06 2,238478

Z45

XY1 XY1

Z2Z1

11 Z45

XY1

Z45

XY1

1 1

2 3

3 4

XY1

Z45 Z2

Z3 Z45

Z90 Z3

XY1

Z45

XY1

2 3

LESS SHADOW

G6.07

G6.05

G6.06

G6.09

G6.10

G6.04 G6.05 EMERGENCE SEMINAR 2013/14

INDIVIDUALS

G6.07

G6.04

G6.02

0,757931 0,146213 Z2

0,6 0,5

G6.07

G6.08

G6.09

G6.10

GENERATION 6

INDIVIDUALS INDIVIDUALS

G6.01 G6.02

G6.02 G6.03

G6.03 G6.04

G6.04 G6.05

G6.05 G6.06

G6.06 G6.07

G6.07 G6.08

G6.08 G6.09

G6.09 G6.10 G6.01

RANKING

GENES

G6.01

GENERATION 6

GENES4

XY1

Z90

XY1 Z45

Z90 Z2

Z2 Z2

Z2 Z45

G6.09 Z2 Z2

Z2 Z45

XY3

Z45 Z2

XY1 Z45

Z1 XY3

Z45

XY1 XY1

Z2Z1

11 Z45

XY1

Z45

XY1

1 1

2 3

3 4

XY1

Z45 Z2

Z3 Z45

Z90 Z3

XY1

Z45

XY1

2 3

Z45

Z1 Z3

Z1 Z90

Z45 Z2

XY1

4 1

Z1 XY1

Z45 Z1

XY1 Z1

Z45

4 3

XY1

Z1 Z90

Z1Z1

Z45 XY1

Z90

Z1 Z1

Z90 XY1

Z1Z3

3 2

G6.02 3

G6.01

2,238478

G6.02

1,586494

G6.03

2,599203

G6.04

1,550984

1,586494

2,599203

1,550984

3,274084 G6.05

G6.07 G6.09 G6.10 G6.04

3,274084G6.02 3,351701G6.08

LESS SHADOW

G6.07 G6.09 G6.10 G6.04

G6.02

G6.01

G6.05

G6.03 Z1

G6.04

G6.08

G6.04

G6.04Z90 1

Z90

G6.05 Z1

LESS SHADOW

INDIVIDUALS

G6.10

G6.03

Z90 Z1 Z2 G6.01 Z1 G6.02 table 2.3 : Results from Generation 6

2,238478

LESS SHADOW G6.07

G6.10

DIVIDUALS

RANKING

GENERATION 6

G6.06

G6.06 MOST G6.07 SHADOW G6.08

G6.05 Z1

G6.06

0,757931 0,146213

1,257979 0,316765

1,5407 0,414089

1,550984 0,417182

0,5

0,4

1,5864941,586494 0,42751 0,42751 2,0341512,034151 0,4932010,493201

0,3

G6.08

0,3

0,2

0,1

G6.02

3,351701 G6.06

0,757931 G6.07

0,757931G6.01

G6.01

2,2384782,238478 0,4754790,475479

2,034151 G6.08

2,034151G6.03

G6.03

2,5992032,599203 0,3814190,381419

G6.09 1,257979

1,257979G6.05

G6.05

3,2740843,274084 0,1480030,148003

G6.10 1,5407

1,5407G6.06

G6.06

3,3517013,351701 0,1269390,126939

LESS SHADOW 2,0191712,019171 LESS SHADOW 0,8087450,808745

figure 2.15 : Ranking for Generation 6

G6.09

0,6

0,5

1,550984 0,417182

G6.08

0,6

0,757931 0,146213

G6.07

G6.09

G Ge

0 0

00,5

0,5 1

1 1,5 1,5 2 2 2,52,5 shadow area (mm^2)

3 3

3,5 3,5

shadow area (mm^2)

figure 2.18 : Normal distribution graph for Generation 6

EMERGENCE SEMINAR

EMERGENCE SEMINAR EMERGENCE SEMINA EMERGENCE SEMINA 15

EMERGENCE SEMINAR 2013/14

ON 4 GENERATION 4

ANALYSIS The iterations of sequence 2 consisted on the experimentation on how to improve a population according to the fitness criteria. The gene pool of the generations had a specific intensity for each gene and emphasis in a specific axis; we specify these variables predicting an outcome that would create most fitted individuals according to our fitness criteria. The killing strategy (disconnection of segments to the rest of the body) was defined by the amount of shadow a single segment would produce when placed at a certain distance of the whole body, this helped us to select our individuals in order to discard the less fitted

RATION 4 GENERATION 4

G4.01

G4.02 G4.01

G4.01

G4.02 G4.01

G4.03 G4.02 G4.03 G4.02

ones before the ranking. The mutation strategy (20% of individuals) will assure the genetic diversity of the population for breeding purposes. The populations of generation 5 and 6 were slightly improved by the breeding strategy of crossover in which the genes of the most fitted individuals breed with the less fitted resulting in a population in which the overall of the individuals will project the less shadow on the ground. The evolution of this sequence towards a most fitted population can be achieved after running a certain (n) iterations in which the individuals are improved generations after generation and applying the breeding, killing and mutation strategies.

G4.04 G4.03

G4.05 G4.04

G4.06 G4.05

G4.07 G4.06

G4.08 G4.07

G4.09 G4.08

G4.08 G4.07 G4.09 G4.10

G G4.10

ON 5 GENERATION 5

RATION 5 GENERATION 5

GENERATION 4 GENERATION G5.014 G5.02 G5.02G5.03 G5.01 G5.01 G5.02 G5.03 G5.01

G5.02

G5.03

G4.01

RATION 6 6 ON 6GENERATION

G5.03 G5.04 G5.04

G5.04

G4.02 G4.01

G5.05

G4.03 G4.02

G5.04 G5.06 G5.07 G5.05 G5.06G5.05G5.06 G5.07 G5.08

G5.05G5.06

G5.07

G4.04 G4.03

G5.08

G4.05 G4.04

G5.09

G4.06 G4.05

G5.07 G5.08 G5.09

G4.07 G4.06

G5.10

G5.08

G5.10

G4.08 G4.07

G4 G4.0

GENERATION 6

GENERATION 5 G6.01 5 GENERATION

G6.02

G6.02 G6.01

G6.01

G6.03

G6.03 G6.02

G6.04

G6.05

G6.05 G6.03

G6.06

G6.06 G6.04

G6.07

G6.07 G6.05

figure analysis result for GenG5.02 4, 5, and 6 G5.03 G5.04 0,7 2.19 : Shadow G5.01 G5.03 G5.04 G6.05 G5.05 G6.01 G6.02 G5.01 G5.02 G6.03 G6.04

G6.08

G6.06G6.09

G6.08

G6.09

G6.10

G6.07G6.10

G5.05G5.06 G5.08 G6.06 G5.06 G5.07 G6.07 G5.07

G5.08 G6.08 G5.09

0,6 0,7

GENERATION 6 GENERATION 6 0,5

0,6

0,4

G6.01

0,5

0,3

G6.01

0,2 0,4

0,7

0,1

0,6

0,5

0,3 0,2

G6.02

G6.03

G6.04

G6.05 Gen 4

G6.05

G6.06

G6.06 Gen 5

G6.07

0,2

G6.08

Gen 4 Gen 5 1

Gen 6

EMERGENCE SEMINAR SEQUENCE EMERGENCE SEMINAR SEQUEN2

shadow area (mm^2)

Gen 4

0,3

G6.07

Gen 6

0,4

0,1

G6.02

Gen 5

0,1

Gen 6

EMERGE

shadow area (mm^2)

figure 2.20 : Normal distribution graph for Gen 4,5, and 6 16

0 EMERGENCE SEMINAR 2013/14

shadow area (mm^2)

EMERGEN EM

EMERGENCE SEMINAR 2013/14

SEQUENCE 3 Sequence Three- Testing the Evolution of Urban Blocks The Tower Blocks of Shibam, Yemen Based on a vertical construction city planning system, Shibam, is a 16th century condensed city composed of mud brick architecture that lies upon a hill in the middle of a flood plain in Yemen. The focus of our evaluation of this city is the verticality of the urbanism, as some of the clay buildings reach up to sixteen stories; These taller buildings surround the most public areas to provide for shadow. A network of roads then connect the public areas, widening as they approach the public spaces and narrowing as they distance themselves. As a result, the height of the buildings begin to correspond with the width of the streets. Assessment of the City In the evaluation of Shibam, the city was divided into four main grids of 60m by 60m with a main road division in between with varying widths or 2m and 12m. Each of those grids were subdivided into twenty-five plots, each 12m by 12m. The sun hits the city at a southern direction with an average sun angle of 70 degrees (averaged from January to December). Design Strategy The goal for the urban block was a vertical-oriented city with maximum sun-ground exposure and minimum sun-building exposure. As a result there were three main fitness criteria: Volume, sun-ground exposure, and sun-building exposure (all equally weighted). The success of this strategy would be how the urban blocks are affected by the sun while still maintaining a high verticality. To achieve this goal, a killing strategy was implemented to eliminate any individual with a volume less than 90,000m^3. This strategy ensures a high volume outcome in all generations. The number of genomes introduced into Octopus, a Genetic plug-in for Grasshopper, was three: copy, move, and scale. Each genome were restricted to a specific range. ‘Copy’ is the number of floors (510) that existed in each tower within the plot. ‘Move’ relocated the floors within its own plots between -1 and 1 in both x and y direction so that the blocks would have some variation. ‘Scale’ then shrank or enlarged the floors to allow for greater variation between each block. With a 10% mutation and a 80%

EMERGENCE SEMINAR 2013/14

crossover rate, we then ran 3 generations using each of the three genomes, our fitness criteria, and killing strategy. Generations In generation one there is a large physical variation in the individuals. The least fittest are individuals with the least volume in relation with low sun-ground exposure and high sun-building exposure: the individuals with narrow towers. The most fittest individuals are those with the most volume in relation to high sun-ground exposure and low sun-building exposure: The individuals with dense towers of varying heights. The individuals with dense blocks of the same height ranked in the middle of the fittest scale. The four of the lowest individuals have a proportionally larger sun-building exposure in relation to the volume and sun-ground exposure. In generation two the narrow towers begin to disappear and denser towers emerge: some individuals with varying heights and some with almost identical heights (G2.05). Similar to generation one, the weakest is that with the smallest volume and the highest sun-ground exposure. Unlike generation one, there is mainly one individual that is supremely weak as oppose the four in the previous generation. In generation three some narrow towers re-emerges. There are three supremely weak individuals, with G3.03 as the weakest due to its disproportional high sun-building exposure. Throughout the three generations there were two main urban blocks that were emerging: wide-compact and narrow-scattered. However, as more generations were ran, one can begin to see a pattern emerge from the urban blocks. By generation twenty, it was clear that the urban blocks begin adjust their height according to the sun vector with most individuals being wide-compact as opposed to narrow-scattered. Unfortunately, the narrow-type individuals were still apparent in generation twenty. As see n by the fitness ranking, the narrow-type urban blocks continuously ranked on the lower spectrum in generation 1-3. In order to eliminate this typology modification to the fitness criteria and killing strategy were introduced in sequence four.

IAGRAM

MAXIMUM VOLUME

VOLUME (m^3)

FITNESS CRITERIA

MOST FITTED

MAXIMUM GROUND EXPOSURE MINIMUM BUILDING EXPOSURE

SUN-GROUND EXPOSURE SUN-BUILDING EXPOSURE

MINIMUM VOLUME

LEAST FITTED MINIMUM GROUND EXPOSURE MAXIMUM BUILDING EXPOSURE

KILLING STRATEGY

ANY INDIVIDUAL WITH A VOLUME LESS THAN 90,000(m^3) DIES

BREEDING STRATEGY

RANDOM (CROSSOVER 80%)

GROWTH STRATEGY

TO ACHIEVE THE MAX VOLUME AND SUN-GROUND EXPOSURE ALONG WITH MIN SUN SURFACE BUILDING EXPOSURE FOR THE BUILDING BLOCKS

MUTATION STRATEGY

10% OF THE POPULATION OF EACH GENERATION WILL BE MUTATED

EVOLUTIONARY GOAL

A FORM OF A CITY BLOCK WHERE THE MAXIMUM VOLUME AND SUN-GROUND EXPOSURE IS ACHIEVED WHILE MINIMIZING THE SUN-SURFACE BUILDING EXPOSURE

EMERGENCE SEMINAR 2013/14

SEQUENCE 3

EMERGENCE SEMINAR 2013/14

SEQUENCE 3

EMERGENCE SEMINAR 2013/14

SEQUENCE 4 Based on the analysis generated from sequence 3, a few adjustments on the genetic algorithm were done in order to do further generations.

building blocks, the maximum number of floors was increased to 30(case one and case two) and 50 (case three) in sequence 4. (Figure 4.1)

ADJUSTMENTS FITNESS CRITERIA The gene pool and its intensity was manipulated. The fitness criteria remains the same as sequence Rotation of each cell was introduced. This results in three in sequence 4 to achieve a better comparison a more varied set of sun exposure data. The rotation to the result achieved from sequence 3. The fitness range for each cell is between 0 degree and 30 degree, criteria are maximum building volume, maximum sunbased on the centre of each cell. In sequence three, ground exposure and minimum sun-building surface the scale gene was only applied to the base cell (first exposure, all equally weighted. (Figure 4.2) level cells) with intensity of 0.1 to 1. Maximum building volume was still one of the fitness criteria. AccordIn Octopus, a Genetic plug-in for Grasshopper, the geing to the result from sequence three, the scale range netic algorithm works is designed to reach the most result of the base cell was from 0.68 to 1. Based on balanced point base on the three criteria set in the the result from sequence three, the scale intensity for program. Additionally, a preference criteria is set: the the base cells was changed to 0.7 to 1 in sequence program still weights all criteria equally for the next HEIGHT 3m four to reduce the calculation load of the algorithm. simulated generation. Instead of changing the weight Also, scale gene was applied to all the other cells with of the criteria, a decision was made to change to killBLOCK 120m x 120m intensity of 0.8 to 1.3 to increase the variety of the ing strategy. The designed killing strategy will kill the SUBDIVISION 12 x 12and generate a more desired result. The simulation result. outlier values 60 new killing strategy will be discussed in each case inm m STREET WIDTHS 2m - 12m 60 In sequence 3, the number of floors was constrained dividually. from 5 to 10 floors. In order to allow for higher density

URBAN BLOCK

ROTATION 0â&#x20AC;&#x2122;-30â&#x20AC;&#x2122;

COPY UNITS 5 - 30 FLOORS 5 - 50 FLOORS

MOVE x (-1.0 - 1.0) y (-1.0 - 1.0)

GENOME/PARAMETERS figure 4.1 : Gene pool and intensity for Sequence 4

EMERGENCE SEMINAR 2013/14

SCALE 0 .8- 1.3

URBAN BLOCK

HEIGHT

BLOCK

120m x 120m

SUBDIVISION

12 x 12

STREET WIDTHS 2m - 12m figure 4.4 : urban block dimensions

URBAN BLOCK

figure 4.2 : Fitness criteria for S4

EMERGENCE SEMINAR SEQUENCE 4

GENOME/PARAMETERS

figure 4.3 : Killing strategy comparison from S3 and S4

GENOME/PARAMETERS 23

EMERGENCE SEMINAR 2013/14 EMERGENCE SEMINAR S

CASE 1 In case 1, a testrun was tested first, to give a general range of data (volume, ground exposure and surface exposure). The killing strategy for generation 1 and generation 2 was designed according to the result achieved from the testrun, which is programmed to kill all the blocks which have a volume below 900,000 m^3. Using this killing strategy, another 20 blocks for

generation one and two were generated. During the analysis of the result from generation 1 and 2, the ground exposure value was found to be around 70 to 90 m^2. (Fig 4.3)

GENERATION 1

GENERATION 2 table 4.1 : Results for Generation 1 in Case 1

table 4.2 : Results for Generation 2 in Case 1

EMERGENCE SEMINAR 2013/14

EMERGENCE SEMINAR

CASE 1

5-30 STOREY KILL VOLUME < 900,000 (M^3)

N TEST RUN testrun

Generation 1

EMERGENCE SEMINAR SEQUENC

EMERGENCE SEMINAR SEQUENC Generation 2 figure 4.5 : Individuals of testrun, Gen1,Gen2 in Case 1

25 EMERGENCE SEMINAR SEQUENCE

EMERGENCE SEMINAR 2013/14

gen9 gen29

100

150

GE (m^2) CASE 2 other two criteria, the improvement of ground expo0,0007 Base on the resulting data generated from case 1, a sure will be very hard to achieve. 0,0006 new killing strategy was designed for gen1 case 2. Blocks 250 with either volume below 900,000m^3 or ground0,0005 exThe normal distribution graph for surface exposure is gen2 posure area below 100 m^2 will be killed. Twenty-nine also the same as expected. (figure 4.10) The curves 0,0004 generations were generated from this strategy, and are shifting from generation one to twenty-nine to the 1354460 gen3 200 area. The convergence 199,2 0,0003 subsequently analysed. As assumed, the ten blocks lower surface exposure graphs gen9 187,4 from twenty-ninth generation converged to one typolfor each criteria were plotted with the corresponding 0,0002 169,9 gen29 for each ogy. Normal distribution graphs were plotted standard deviation range. 0,0001 criteria. From the normal distribution graphs, (figure 150 0 4.6) the curves in the volume graph were shifting to 150 200 250 300 26000 28000 30000 32000 34 the bigger volume side, which is the same as whatAverage the 24000 Volumes GE (m^2) algorithm is supposed to achieve. SE (m^2) 100

00 99,2 0

GE (m^2)

0,0009 900.000 1.100.000 1.300.000 1.500.000 1.700.000 0,0008 V (m^3)

The ground exposure graph (Figure 4.8) showed the 32000 ground exposure value is in the range of 100 to 250 m^2. There is no significant curve shifting tendency from generation 1 to 29. This may be caused by the 31000 designed killing strategy. The blocks with ground ex185,4 posure below 100 m^2 were all killed. To balance the 169,9 30000 Gen 1 Gen 2 Gen 3 Gen 9 Gen 29 141,8 Average GE GENERATION 1

SE (m^2)

700.000

0,005

0 Gen 1

29000

Gen 2

GENERATION 2

Gen 3

2838

28000 27000 26000 Gen 3

Gen 9

Gen 29

GENERATION 3

Gen 1

Gen 2

Gen 3

Gen 9

Gen 29

GENERATION 29

table 4.3 : Results for Generation 1,2,3 & 29 in Case 2

EMERGENCE SEMINAR SEQ 26

EMERGENCE SEMINAR 2013/14

0,000003

gen3

0,000002

gen9

0,000001

gen29

0 500.000

700.000

900.000 1.100.000 1.300.000 1.500.000 1.700.000

CASE 2CASE 2

V (m^3)

1600000 1400000

0,025 1000000

0,025

000007

0,000006

000006

0,000005

gen1

0,000004

gen2

gen1

gen3

gen2

gen9

0,000002

gen3

gen29

0,000001

gen9

0 500.000

700.000

900.000 1.100.000 1.300.000 1.500.000 1.700.000 V (m^3)

gen29

0 4.6 : Volume graph 1.700.000 500.000figure 700.000 900.000 normal 1.100.000distribution 1.300.000 1.500.000 1600000

1200000

V (m^3)

gen1

Gen 2

Gen 3

Gen 9

Gen 29

gen29

50 2 Gen

1003 Gen

1509 Gen

GE (m^2)

20029 Gen

250

300

250

GENERATION 200 2501

100

150

gen9 gen29

GE (m^2) 200

187,4

250

GE (m^2)

150

gen2 gen9

GE (m^2)

0,0008

187,4

141,8 Average GE

185,4

169,9

141,8 Average GE

0,0006

gen1

0,0005 0

Gen 1

Gen 2

Gen 3

Gen 9

gen2

Gen 29

gen9

GENERATION 2 0,0001 0

Gen 1 24000

185,4

169,9

Average GE

Average GE 27 gen1

Gen 1

Gen 2

Gen 3

Gen 9

gen3 gen9

GENERATION 2

gen3

0,00030 0

Gen 126000 Gen 228000 Gen 30000 3 0,0002 24000

Gen32000 9 Gen34000 29

gen9

SE (m^2)

figure 4.9 : Ground exposure convergence graph 28000

GENERATION gen29 3 Gen 3 28000

32000

SE (m^2)

31000 29000

Average SE 28380,6

27000 29000

Average SE 28380,6 Gen 1

Gen 2

Gen 3

Gen 9

Gen 29

27000

GENERATION 29

Gen 9 30000

Gen 29 32000

Gen 1

34000

Gen 2

Gen 3

Gen 9

Gen 29

SE (m^2)

figure 4.10 : surface exposure normal distribution graph

32000 GENERATION 2

GENERATION 3

figure 4.11 : Surface exposure convergence graph

GENERATION 29

31000 30000

SE (m^2)

verage GE

27 29000 28000

gen29

GENERATION 3 34000

30000

26000 Gen 2 26000

gen1

GENERATION 3gen29 gen2

0,0004 0,0001

0,0001

gen2

Gen 29

0,0005 0,0002

26000 28000

Average SE 28380,6

141,8

gen3

0,0002

199,2

187,4

30 141,8

30000 28000

50 0,0003

gen29

185,4

30000

0,0004

199,2

187,4

32000

199,2

0,0009 0,0007

gen1 100 gen3

150

0,0001

GENERATION 1

0,0005 0

185,4

169,9

0,0002 gen3

GE (m^2)

31000

199,2

gen2 0,0003

gen29 24

0,0008

0,0004

0,0005

gen1 0,0004

figure Volume convergence 150 4.7 : 200 250 300graph

100

0 32000 GENERATION 2 24000 26000

300

0,0006 100

0,0006 50 0,0003

gen3

100

GE (m^2)

25029 Gen

0,0009

gen3

gen2

figure 4.8 : Ground exposure normal distribution graph

200 e Volumes

2009 Gen

0,0008 50 0,0007

0,0007

gen9

GENERATION 1

gen29

150 3 Gen

100

gen2

gen1

200000 0,005

100 2 Gen

0,0009 150

Average Volumes

Gen 1

0 1 Gen

50 1 Gen

150

200

Average Volumes

0,015 gen9

0,005

200

0,0006

gen9

169,9

400000 0,01 gen3

250

0,02 gen2

0,01

1354460

200000

gen1 600000 0,015

gen29

400000

800000 0,02

250

600000

0,025

0,005

1354460

800000

1000000 0,025

gen9

SE (m^2)

1000000

gen3

0,005 200000 0,01

GE (m^2)

1400000 V (m^3)

1200000

gen2

0,015 0,01 400000

V (m^3)

1400000

1600000

gen1

0,015 600000

GE (m^2)

000001

0,0007 Average Volumes

0,02

SE (m^2)

000002

0,0008

0,02 800000

0,000003

000003

0,0009

V (m^3)

0,000008 0,000007

000004

1200000

000008

000005

1354460

SE (m^2)

EMERGENCE SEMINAR 2013/14

EMERGENCE SEMINAR SEQUENCE 4

RESULTS The results confirms the conclusion from the normal distribution graph. The triangle radar graph for generation one, two, three and twenty-nine shows from one generation to the next, the generated blocks became more balanced of the three criteria. Most of the outlier values were deleted in the twenty-ninth generation radar graph. (Figure 4.12)

EMERGENCE SEMINAR SEQUENCE 4

figure 4.12 : Octopus result and triangle radar graph for Case 2

EMERGENCE SEMINAR 2013/14

G1 SE 2

KILL SUN GROUND EXPOSURE < 100

5-30 STOREY KILL VOLUME < 900,000 (M^3) KILL SUN GROUND EXPOSURE < 100

EMERGENCE SEMINAR SEQUEN

G29

EMERGENCE SEMINAR SEQUENCE

figure 4.13 : Individuals of Gen1,Gen2, Gen3, Gen29 in Case 2

EMERGENCE SEMINAR 2013/14

G1.07

G2.03

G1.03 -

G2.05 -

G1.04 CASE 3 G2.02

G2.01 G3.03

G2.07 G3.05

In case 3 of sequence 4, while keeping the same killG1.01 ing strategy from case two, the number of floor levels G2.04 was raised to 50. Nine generation were generated. + designed genetic algorithm was too heavy to run The G2.06 more than ten generations. The same result analysis G2.03 strategy as case two was applied to case three. As assumed, the normal distribution graph for both volume G2.05 and surface sun exposure area were optimised. Same as case 2, no significant improvement based on fitness G2.01 criteria of ground exposure was achieved. The trianG3.03 gle radar graphs showedGN.01 the three criteria becoming GN.02 G2.07 more and more optimised from generation 1 to 9. G3.05

GENERATION 1

+ G3.07

G3.02 G3.01 G3.04 G9.03

GN.03

G3.06 G9.10 + G9.02

G3.02

G9.06

G3.01

G9.09

G3.04

GN.02

+ G3.07

G3.06 +

GN.03

GN.04

GN.05

GENERATION 2

G9.08 G9.01 G9.03

GN.04

GN.05

GENERATION 2

GN.06

GN.07

GN.08

GENERATION 3

G9.04 G9.07 +

GN.05

GN.06

GN.07

GENERATION 3

table 4.4 : Results for Generation 1,2,3 & 9 in Case 3

EMERGENCE SEMINAR 2013/14

GN.08

GN.09

GN.10

GENERATION 9

EMERGENCE SEMINAR SEQ

0,0000015

gen3

0,000001

gen9

0,0000005 0 500.000

CASE 3

900.000

1.300.000

CASE 3

1.700.000

V (m^3) 1,6000E+06

0,0000025 0,000002

0,012

0,0000035

gen1

0,000002

gen2

0,0000015

gen2gen9 gen3

0 500.000

6,0000E+05 0,006 0,008

900.000

1.300.000

gen9

1.700.000

0,002

1,6000E+06

V (m^3) 1,3288E+06

1,6000E+06 1,2000E+06

V (m^3)

1,2000E+06 8,0000E+05

V (m^3)

4,0000E+05 008 gen2

gen2

Gen 1

Gen 2

Gen 3

Gen 9

0,004

GENERATION 1

Gen 2

100

Gen 3

150

200

250

Gen 9 300

350

150

200

GENERATION 250 300 3501

200,0

400

gen2 gen3 50,0 gen9

0,0

189,9

50,0

Gen 9

0,0003 0,0002

GENERATION 2 0,0001

gen2 gen3 gen1 gen9

gen9

GENERATION 3

35000 40000 45000 50000 55000 60000 SE (m^2) 41985,1

41985,1 Average SE

10000,0

25000,0

15000,0

gen3

10000,0

GENERATION 3gen9

Average SE

5000,0 0,0 Gen 1

Gen 2

Gen 3

Gen 9

GENERATION 9

5000,0 0,0

Gen 1

Gen 2

Gen 3

Gen 9

SE (m^2)

figure 4.18 : Surface exposure normal distribution graph GENERATION GENERATION 3 50000,02

figure 4.19 : Surface exposure convergence graph GENERATION 9

45000,0 41985,1

35000,0

SE (m^2)

verage GE

30000,0 25000,0 20000,0 15000,0 10000,0

gen3

30000,0

gen2

20000 25000 35000 45000 55000 60000 Gen 1 Gen30000 2 Gen40000 3 Gen50000 9

40000,0

gen2

15000,0

gen1

Gen 3

20000,0 35000,0

0,0

Average SE

gen1

figure 4.17 : Ground exposure convergence graph

20000,0

Average GE

GENERATION 3

25000,0 40000,0

Average GE

189,9

Average GE Gen 2

Average GE

45000,0 30000,0

0,0007 50,0 0,0006

Gen 1

50000,0 35000,0

162,4

0,0008

162,4

SE (m^2)

190,3

162,4

40000,0

189,9

190,3

162,4

100,0

0,0004

190,3

45000,0

150,0

0,0005

figure 200 4.15 250 : Volume 300 convergence 350 400 graph

0,0

SE (m^2)

GE (m^2)

gen1

GE (m^2)

250,0

400

GENERATION 1

GENERATION 2 20000 25000 30000 50000,0

400

190,3

Gen 9

GE (m^2)

150,0

GE (m^2)

100,0

150

350

0,0003 0,0005 Gen 1 Gen 2 Gen 3 Gen 9 gen1 gen3 50,0 0,0002 0,0004 gen2 gen9 0,0001 0,0003 gen3 0 0,0 0,0002 20000 25000 30000 35000 40000 45000 50000 55000 60000 Gen 1 Gen 2 Gen 3 Gen 9 gen9 SE (m^2) 0,0001

figure 4.16 : Ground exposure normal distribution graph

150,0

100

300

Gen 3

GENERATION 2

GE (m^2)

Average V 200,0

250

GE (m^2)

0,0007 0,0005 100,0 0,0004 0,0006

gen1

0,0004

0, gen10,0003 0,00020, gen9 gen2 0,00010, gen3 00, gen9 20 0,

0,0008 0,0006

Average V

0,0082,0000E+05

50 250,0 100

250,0

200

Gen 2

150,0 0,0007

6,0000E+05 0,01 gen1

150

0,0008 100,0

0,014,0000E+05

100

200,0

Average V

012 8,0000E+05

Gen 1

250,0

1,3288E+06

0,0126,0000E+05 1,0000E+06

200,0

GE (m^2)

1,4000E+06 1,0000E+06

Gen 1

1,4000E+06

0,0060,0000E+00

0 0,0000E+00

0,0005

gen3

0,002 2,0000E+05

0 figure 4.14 : Volume normal distribution graph 500.000 900.000 1.300.000 1.700.000

002

gen2

0,004

V (m^3)

gen1

4,0000E+05 0,004 gen3 0,006

0,000001

006 2,0000E+05 0,002 gen9 004 0,0000E+00 0

Average V

gen1

0,0000015 0,0000005

gen3

0,0006

8,0000E+05

0,008 0,01

0,0000025

0,0000005

0,0007

0,0120,01

0,000003

0,000001

0,0008

1,0000E+06

V (m^3)

0,000003

1,3288E+06

1,2000E+06

GE (m^2)

0,0000035

1,4000E+06

SE (m^2)

0,000004

EMERGENCE SEMINAR SEQUENCE 4

EMERGENCE SEMINAR 2013/14

ANALYSIS The sun vector set in this generic algorithm is a single vector with angle of 70 degrees. Since the number of floors was already increased to 50, based on the three criteria, the number of floors was expected to follow the same gradient as the sun vector. The result from nine generations was not enough to observe this expected result. Because of the heavy calculation, another simple algorithm was designed to test this result. Twenty-five columns were set with variable height range. The analysis is only on maximum ground exposure and minimum sun surface exposure. Two hundred generations were run, which confirm the gradient assumption. figure 4.20 : Simple experiment to get expected result

figure 4.21 : Simple experiment in Octopus for 100 generations

EMERGENCE SEMINAR 2013/14

KILL SUN GROUND EXPOSURE < 100

NDIVIDUALS figure 4.22 : Individuals of Gen1,Gen2, Gen3, Gen9 in Case 3 EMERGENCE SEMINAR SEQUENCE 4

V = 1160700 M^3 GE = 37 SE = 26331 FAR = 26 PLOT =77%

V = 1601500 M^3 GE = 70 SE = 26158 FAR = 38 PLOT = 87%

V = 1697600 M^3 GE = 56 SE = 26293 FAR = 39 PLOT =85%

V = 1114400 M^3 GE = 149 SE = 30930 RAN= 3 FAR = 25 PLOT = 76%

V = 1239700 M^3 GE = 152 SE = 28786 RAN= 6 FAR = 23 PLOT = 88%

V = 996.001M^3 GE = 164 SE = 29321 RAN= 3 FAR = 30 PLOT = 92%

V = 1584400 M^3 GE = 109 SE = 38184 RAN= 4 FAR = 37 PLOT = 80%

V = 11948M^3 GE = 196 SE = 9831 RAN= 4 FAR = 28 PLOT = 82%

V = 10731M^3 GE = 117 SE = 21393 RAN= 7 FAR = 19 PLOT = 79%

V = 600000 M^3 GE = 70 SE = 17658 PLOT = 65%

figure 4.23 : 10 picked individuals from S4

EMERGENCE SEMINAR SEQUENCE 4

EMERGENCE SEMINAR 2013/14

CONCLUSION kjfkjalskdjlaskjdalksdjlaskjdlask

EMERGENCE SEMINAR 2013/14