a study in material computation and agency new school of architecture + design

FABRIc
{e}bot-lutions // symb(i)ot
a study in material computation and agency

IAM

Design Re_Search

newschool of architecture + design

Chris topher Schulte Kevin Sullivan Michael Flynn Richard Ciavetta Robert Merchel

special thanks
Gabriel Morales-olivares

introduction computation methodology material agency digital analog prototypes final prototype environment simulation installation

10 11

12 34 42

48 80 82 86

introduction

This five-week workshop explored the ideas of material computation and agency. The first objective of the workshop was to identify and analyze the constraints and potentials of the chosen material, fabric elasticity, hardening potential (form resistance), and those of the agents (robots) performance, self-awareness (sensors), scale potential, range of motion. In order to merge the information of both parties, creating a feedback loop between the robots and the material was necessary. By attaching various sensors proximity (ping), location (xbee) and a webcam to the SERBs, surface features could be detected and responded to, thus allowing the material to be affected on a local level in real time. Using Processing, these specific responses could be coded into the robots, i.e. a decision-making protocol was created (agency). Imparting the robots with agency thus allowed the system to adapt to the material at a local level and work towards desired spatial outcomes enclosure, light filtration, density, etc. As such, the system exhibits a high potential for variation and adaptability.

10

COMPUTATION

Computations relation to design is the processing of information, behaviors and reactions between each individual element to result in a spatial environment. Computation used in this research was used to realize complex formations from a small number of nite rule sets. This algorithmic approach denes operations for the solution of given input parameters. Therefore the solution to a given problem is deterministic based on the simple rules, yet highly complex and adaptive to the starting parameters.

The methodology of this research was conducted via material computational agency. Analogue testing of the material properties were analyzed for their computational possibilities, while the computational simulations were run to test the material properties. Working inside this feedback loop allowed discovery of material potential and computational complexities. Computational studies we used as abstractions to test the materials in spatial qualities. These theoretical computational analyzes allowed for the fabrication through analogue testing of the design of our nal formations.

METHODOLOGY

11

12

material

After testing out different materials, we chose fabric and plaster as our primary materials. The fabric is a papyrus rayon spandex knit consisting of 95% rayon and 5% spandex . The plaster is a general purpose plaster of paris, which was mixed at a ratio of 4:1 with water before being applied to the fabric. In order to understand the material constraints and opportunities, it was necessary to perform several tests on the fabric. We began by carrying out simple operations such as pushing and pulling at one and multiple points, twisting, cutting and stitching. These tests gave us specific data about the ability of the fabric. For example, on a 22 square inch frame, a square piece of fabric pinned at all 4 corners deflects roughly 2.5 under gravity, roughly 7.5 under maximum tension and roughly 5 under the weight of 5 ounces of our plaster mixture. The following pages present a series of these material tests accompanied by their digital counterparts.

13

14

Pinch_2_Ridges

SETUP: OPERATIONS: RULES:

22 square inch frame; 20 square inch fabric, pinned at 4 corners pinch (zip tie) Scan for main ridge within local neighborhood. Zip tie 2 bisecting holes along ridge.

M
CONCLUSIONS The action taken in this test added texture and variation to the material based on a simple rulebased procedure. The result shows a surface manilupation but not a spatial result.

growth / texture / tension

15

16

Cut_Max_Tension

SETUP: OPERATIONS: RULES:

22 square inch frame; 20 square inch fabric, pinned at 4 corners cut, pull Zip tie first encountered point along main ridge. Zip tie next 2 closest holes along trajectory of resulting ridge.

M
CONCLUSIONS The action taken in this test added depth, texture and variation to the material based on a simple rule-based procedure. This operation was one of the primary taken on the final prototype. At scale this operation allows for depth in the surface through voids.

depth / porosity

17

18

Pull_Ridge

SETUP: OPERATIONS: RULES:

22 square inch frame; 20 square inch fabric, pinned at 4 corners pull Pull midpoint of ridge between pinches to closest control point.

M
CONCLUSIONS The action taken in this test added deminsion to the surface through the use of attractor points connecting to the surface. This method was used on the final prototype.

depth / growth / volume / variation

19

20

Cut_Pull_Opposite

SETUP: OPERATIONS: RULES:

22 square inch frame; 20 square inch fabric, pinned at 4 corners cut, pull If 2 points are 3.5 apart, cut between them. Pull midpoint of interior edge of cut to closest lower control point. Pull midpoint of outer edge of cut to closest upper control point.

M
CONCLUSIONS The action taken in this test added openings and depth to the surface system. Adding the voids to the system created a spatial quality to the enclosing surface structure.

depth / porosity

21

22

Point_Growth

SETUP: OPERATIONS: RULES:

22 square inch frame; 20 square inch fabric, pinned at 4 corners pinch (zip tie) Zip tie first encountered point along main ridge. Zip tie next 2 closest holes along trajectory of resulting ridge.

M
CONCLUSIONS The action taken in this test utilized vector projections from the agents to create areas of density in the surface. While not achiving the larger spatial qualities of form, the operation was used in the final prototype for surface complexity.

growth / branching / texture / variation

23

24

Pinch_Perpendicular_Pull

SETUP: OPERATIONS: RULES:

22 square inch frame; 20 square inch fabric, pinned at 4 corners pinch (zip tie), pull Pinch 2 holes perpendicular to ridge. Pull to closest control point.

M
CONCLUSIONS The action taken in this test added complexity to the surface while creating a spatial condition. The use of the control point helped the surface remain in tension while creating volume..

depth / growth / variation

25

26

Gather_Holes

SETUP: OPERATIONS: RULES:

22 square inch frame; 20 square inch fabric, pinned at 4 corners pinch (zip tie) If point falls on a ridge, zip tie with nearest hole.

M
CONCLUSIONS The action taken in this test utilized local conditionals from the agents to create areas of density in the surface. While not achiving the larger spatial qualities of form, the operation was used for surface complexity.

tension / texture / variation / volume

27

28

Cut_Pinch_Pull_2

SETUP: OPERATIONS: RULES:

22 square inch frame; 20 square inch fabric, pinned at 4 corners pinch (zip tie), cut, pull If points are 3.5 apart, cut between them. Pinch midpoint of adjacent flaps. Pull midpoint of outer flaps to closest ground control point. Pull pinch to closest upper control point.

M
CONCLUSIONS The action taken in this test added depth, texture and variation to the material based on a simple rule-based procedure. This operation was one of the primary taken on the final prototype.

depth / growth / porosity

29

30

Cut_Pull

SETUP: OPERATIONS: RULES:

22 square inch frame; 20 square inch fabric, pinned at 4 corners cut, pull If 2 points are 3.5 apart, cut between them. Pull midpoints of cut to closest control point.

M
CONCLUSIONS The action taken in this test reduced the surface tension of the matieral properties. This function highlighted the need for some agents to be dependant on the actions of other agents.

depth / porosity / variation

31

32

The main application for testing of the design took place in simulation through the use of a personal computer. Code was launched through the open source compiling application Processing and its visual environment. Processing is a Java based computer language and design environment for designer and artists. This platform was selected because of its ease of entry into the computer programming environment for many non-computer scientists. Having the base of Java allowing for some higher functionality of the Java language to be implemented as needed for greater control. Using a higher level language like Processing and its open source environment allows for the use of many public libraries and functions to be implemented. These functions get the project to prototyping faster, thus allowing for more system testing and less programming.

agency

33

34

A
Agent Based Modeling (ABM) starts from a foundation in complexity. Complexsystemshavealargenumberofelements,self-organization, emergent patterns, deterministic results and they occur nonlinearly. ABM starts with these parameters and layers in functions of intelligence. The early utilization of these modeling systems sought to understand the complex phenomenon of nature. At the core and typical agent based model has only three base components: agents, rules and an environment. The agents carry their own parameters and are grouped into like-minded sets or sets or sets. The rules are the programmed parameters of control inside the system. Each action creates the functionality of the output based on these connections. The environment is where and when the system is occurring.

The simulation environment is a virtual physics based modeling arena. We dene the size, scale and power of our virtual world by controlling each of its parameters. Our tests showed that to simulate our Spandex-Rayon fabric, the fabric had a 4 fall due to gravity when laid at over a 24 rig. This data was then used to nd and set the power of gravity in the model. Gravity is the only physical boundary used in this simulation. An area was setup for the simulations to run inside. This started as a square for use as a general bounds for each action and reaction. Later for installation testing, the bounds included the site conditions.

d i g i t a l

35

36

//SeperatewithAgentBtype----------------------------------------------------void seperateAA(oat magnitude) { Vec3D steer = new Vec3D(); int count = 0; for (int i = 0; i < arrayOfAgentB.size(); i++) { AgentB other = (AgentB)arrayOfAgentB.get(i); oat distance = loc.distanceTo(other.loc); if (distance > 0 && distance < 30) { Vec3D diff = loc.sub(other.loc); // Strengh of the seperation factor then how mechical the reaction is diff.normalizeTo( 1.0 / distance); steer.addSelf(diff); count++; } // if } // for if (count > 0) { steer.scaleSelf( 1.0 / count); } // if steer.scaleSelf(magnitude); acc.addSelf(steer); } // void seperate Agent B type

void createFabric() { for (int i = 0; i< ROWS; i++) { for (int j = 0; j< COLS; j++) { VerletParticle vp = new VerletParticle((-DIMX/2)+ (10*i), (-DIMX/2)+(10*j), (DIMZ/2)); physics.addParticle(vp); arrayOfVP.add(vp); if (j ==0 || j==COLS-1 || i ==0 || i == ROWS-1) { vp.lock(); } if (j >= 0 && j%ROWS!=0) { VerletParticle vp2 = (VerletParticle) arrayOfVP.get(index-1); VerletSpring sp = new VerletSpring(vp, vp2, 10, 0.5); physics.addSpring(sp); } if (i >= 0 && i%COLS!=0) { VerletParticle vp2 = (VerletParticle) physics.particles.get(indexCOLS); VerletSpring sp = new VerletSpring(vp, vp2, 10, 0.5); physics.addSpring(sp); } } } }

The primary reason for utilizing an Agent Based Modeling system is the reactions and interactions of the system components. Interactivity is classied at the social reactions between the agents and the dynamic reactions between their behaviors. Interactions between agents occur at the local level meaning that each interaction happens only within the agents local neighborhood. The range of an agents neighborhood can be set or changing and could be different than the range at which an agent can see objects. This local level interactivity is the dening characteristic in creating a decentralized system for modeling How agents interact in the environment is through a movement action called Flocking. This is based on Shiffmans processing examples, and he derived its actions from the movement of birds in nature. Each agent works to attract towards one another to form the ock. Once near each other there is a lesser repulsion force to prevent contact. And a nal force of alignment to steer all the agents towards a similar goal. These locations change and are updated every cycle.

Our simulation utilizes the Toxiclibs VerletPhysics libraries for processing. These are methods and classes that have prewritten functionality for modeling real world applications. The simulated fabric is represented as a grid of particles and springs. The particle is a body that can be inuenced using vector analysis. This body has no intelligence or agency. Utilization of the verlet particle allows for an accurate simulation. The springs are the glue that holds the surface together. The amount of tension was selected based on the real world tests. Because the fabric is not a ridged body this method allowed for a exible simulation.

Repulsion example

Fabric Creation

d i g i t a l

37

38

void cutSurface() { for (int i =3; i<infArrayB.size(); i++) { VerletParticle vp = new VerletParticle(0, 0, 0); VerletParticle a = (VerletParticle)( infArrayB.get(i)); VerletParticle b = (VerletParticle)( infArrayB.get(i-1)); VerletParticle c = (VerletParticle)( infArrayB.get(i-2)); oat distance = a.distanceTo(b); if (distance > 10 && distance < 25) { physics.removeSpring(physics.getSpring(a, b)); physics.removeSpring(physics.getSpring(b, c)); b.z = b.z - 20; b.lock();

void cutPara() { for (int i =3; i<infArrayB.size(); i++) { VerletParticle vp = new VerletParticle(0, 0, 0); VerletParticle a = (VerletParticle)( infArrayB.get(i)); VerletParticle b = (VerletParticle)( infArrayB.get(i-1)); VerletParticle c = (VerletParticle)( infArrayB.get(i-2)); VerletParticle d = physics.particles.get(infArrayB.indexOf(i) + ROWS); VerletParticle e = physics.particles.get(infArrayB.indexOf(i-1) + ROWS); VerletParticle f = physics.particles.get(infArrayB.indexOf(i-2) + ROWS); oat distance = a.distanceTo(b); if (distance > 1 && distance < 25) { physics.removeSpring(physics.getSpring(a, b)); physics.removeSpring(physics.getSpring(b, c)); physics.removeSpring(physics.getSpring(d, e)); physics.removeSpring(physics.getSpring(e, f));

One agent rule was to cut the surface when the nodes of the surface reached a threshold of tension. This tension can be simulated as a distance between points on the fabric grid. The agent would search its range of inuence of the surrounding 9 particles and make the comparision for determination if the situation would require a cut or if to move on.

b.z = b.z - 20; b.lock();

This rule sought an area of the surface that had a high level of tension and provided two parallel cuts along the grid of the surface. A small weight is added to the cut line to simulate the now loose fabric falling from the surface.

Agent Rule - Cut Surface

Two Parallel Cuts

d i g i t a l

39

40

void stichRunning() { for (int i =2; i<infArrayC.size(); i++) { VerletParticle a = (VerletParticle)( infArrayC.get(i)); VerletParticle b = physics.particles.get(physics.particles. indexOf(a) + ROWS); oat distance = a.distanceTo(b); if ( a.isLocked() == true || b.isLocked() == true ) { break; } if (distance > 10 && distance < 25) { a.x = b.x; a.y = b.y; a.z = b.z; a.lock(); b.lock();

void stichGather() { for (int i =2; i<infArrayC.size(); i++) { VerletParticle gatherVP = (VerletParticle)( infArrayC.get(0)); VerletParticle a = (VerletParticle)( infArrayC.get(i)); VerletParticle b = physics.particles.get(physics.particles. indexOf(a) + ROWS); if (b.isLocked() == true ) { break; } b.x = gatherVP.x; b.y = gatherVP.y; b.z = gatherVP.z; a.lock(); b.lock();

The gather command uses the agents range of inuence to collapse the surrounding points to the nearest point to the agent on the surface. The running stitch command seeks two grid points and welds them together. This can be set to gather excess material, tighen surfaces or add context. This can be applied to tighten the surface, or through the use of surface embellishments to heighten the nal resolution.

Running Stich

Gather Points

d i g i t a l

41

42

Servo-Robot

Arduino Uno

Ping Sensor

Arduino is a hardware component allowing for computation to take place within our system. It is capable of syncing with Processing in so that the robot can become an autonomous agent capable of taking decisions as it moves around the environment. With the addition of extra components, the robot has the capability of sensing conditions, which serve to activate its decision-making protocol. The following images represent some of the tools that our system implemented.

Laser Pointer

Cutting Tool

Robotic Arm

attachments

a n a l o g

43

44

tests
a n a l o g

45

46

application
a n a l o g

47

48

After testing the basic material properties, it was time to begin performing analog versions of our agent-based fabrication process. To do so, we assumed the role of the robot and performed its given task. Each test began by establishing specific operational rules by which each robot would operate. These rules had to be local and based on specific material conditions, and they would therefore yield a high resolution result that displayed both local variation and adaptation. Our initial tests suffered from improper rulesets that relied on more global rather than local rules and functioned in a linear manner instead of as continuous feedback loops, i.e. rule 1 was performed, then rule 2, then rule 3, etc. Instead, rule 1, 2 and 3 should have all occurred simultaneously and in concert with each other, thus affecting each other and the local condition of the material as they collide and intersect. This resulted in surfaces, which were low resolution and lacked sufficient variation and adaptation.

prototypes

49

50

51

52

TEST: SETUP: OPERATIONS RULES: POST-PROCESSING: CONCLUSIONS:

Pushing 24 inch x 8 inch frame; 36 x 24 inch, fabric piece. Push Press up on fabric at various locations and heights. Hardened with a mixture of Urea Resin Powder, water and berglass shreds (2:1:1 water-urea-berglass). Heat gun @ approx. 6 inches for 5 minutes. Results starting to showcase some repitition of the rule. Actions are not inuencing ofther actions.

TEST: SETUP: OPERATIONS RULES: POST-PROCESSING: CONCLUSIONS:

Cutting 10 foot frame; 10 foot x 48 inch, fabric piece. Cut/Stich Cut a U shaped slot in fabric at random locations. Stitch inside of cut to nearby location. Unnished felt Very large scale protoype showcase the potential application of the project. The rule was less informative because it developed no diminsionality at scale.

f a b r i c

f a b r i c

53

54

TEST: SETUP: OPERATIONS RULES: POST-PROCESSING: CONCLUSIONS:

Twisting 24 inch x 8 inch frame; 36 x 24 inch, fabric piece. Push/Twist Press in center of fabric. Wrap fabric around apex 180 degrees Hardened with berglass resin. Dry overnight. Single operation test showcases the potential for depth from a at surface. Twisting action provides new formations on the fabric. Resin shown to be a hindrance in application and potential.

TEST: SETUP: OPERATIONS RULES: POST-PROCESSING: CONCLUSIONS:

Pulling 24 inch, square frame; 26 inch, square fabric piece. Pulling From four locations add weight to fabric. Hardened with breglass resin. Dry overnight. Limited results. Fabric shows an ease to the deformation, but lacks any complexity without different weight measures applied. Resin shown to be a hindrance in application and potential.

g l a s s

g l a s s

55

56

TEST: SETUP: OPERATIONS RULES: POST-PROCESSING: CONCLUSIONS:

Pushing 24 inch x 8 inch frame; 36 x 24 inch, fabric piece. Push Press up in rst third of fabric. Press down in second third of fabric Hardened with breglass resin. Dry overnight. Limited results. Showcased fabric ability to work with opposing forces. Starting to show some dimensionality. Resin shown to be a hindrance in application and potential.

TEST: SETUP: OPERATIONS RULES: POST-PROCESSING: CONCLUSIONS:

Pushing 24 inch x 8 inch frame; 36 x 24 inch, fabric piece. Push Press up in rst third of fabric. Then strech to the side. Hardened with breglass resin. Dry overnight. Results were interesting because of the global arch formation that developed. The lateral movement started to showcase extra formations. Resin shown to be a hindrance in application and potential.

g l a s s

g l a s s

57

58

TEST: SETUP: OPERATIONS RULES: CONCLUSIONS Early prototypes showed the potential of the fabric in terms of spatial conditions. Because of the harder agent used there was limited material testing. Results proved concept could yield a full scale result. POST-PROCESSING: CONCLUSIONS:

Smocking I 24 inch, square frame; 22 inch, square fabric piece. Push / Pull / Stitch Establish 2 points @ distance of 4 inches. Resin perpendicular path between points. Pull down and stitch points together. Find end midpoints of resined path; push up. Hardened with Urea Resin Powder & water mixture (2:1 water-urea). Heat gun @ approx. 6 inches for 5 minutes. This test demonstrated the hardening capacity of the Urea resin on the fabric and was the impetus for further smocking pattern tests. This resin mixture provided minimal stiffness / form-resistance in comparison to the berglass resin and the plaster mixtures.

s m o c k

59

60

TEST: SETUP: OPERATIONS RULES:

Smocking II 70 inch, square frame; 65 inch, square fabric piece. Push / Pull / Stitch Resin parallel strips along length of fabric. Establish pairs of points along edges of resin. Pull down and stitch corresponding points together. Hardened with Urea Resin Powder & water mixture (2:1 water-urea). Heat gun @ approx. 6 inches for 5 minutes. This test further demonstrated the hardening capacity of the Urea resin and was the rst attempt at a larger scale (6x6). The multiple passes of resin provided minimal extra stiffness / form-resistance in comparison to previous tests. The resulting form lacked sufcient three-dimensionality, variation and local adaptation.

TEST: SETUP: OPERATIONS RULES: POST-PROCESSING: CONCLUSIONS:

Smocking III 24 inch, square frame; 26 inch, square fabric piece. Pull/Stitch Drop 6 tubes of exible plastic onto the surface to dene stitching points. Attach ends together using thread. Using exible plastic secure stitch point. This test showcased the used of a simi-ridiged material to articially create depth. While this test only showcased its atness, the question was raised by the team about how Plaster can be diverted into different areas.

POST-PROCESSING: CONCLUSIONS:

s m o c k

s m o c k

61

62

TEST: SETUP: OPERATIONS RULES:

Smocking IV 24 inch, square frame; 26 inch, square fabric piece. Pull/Stitch Create a 3 stitch along a ridge. Alternate side of ridge. CONCLUSIONS Hardened with Urea Resin Powder & water mixture (2:1 water-urea). Powered foam added. Heat gun @ approx. 6 inches for 5 minutes. The alternating pattern created strong channels under the stitch. Test had a limited amount of dimensional quality. Created a strong central ridge, but focused on the global attributes. This series of tests showcase the differnt pattern effects on the system. Utilization of a singular pattern for creating of form shows potential. All tests lack a general spatial quality of the system.

POST-PROCESSING: CONCLUSIONS:

s m o c k

63

64

TEST: SETUP: OPERATIONS RULES: POST-PROCESSING: CONCLUSIONS:

Stitching I 24 inch, square frame; 26 inch, square fabric piece. Fabric started with hole. Cutting/Stitching Cut 8 line along midpoint from attachment point to hole. Stitch centroid of new plane to centroid of opposite new plane. No post-processing. Discoveries about the volumetric possibilities for passing fabric over and under itself. Due to fabrication constraints method was dreamed unusable. Starting with a hole in the fabric did not change the material potential.

TEST: SETUP: OPERATIONS RULES:

Stitching II 24 inch, square frame; 26 inch, square fabric piece. Stitching From start point agents moved 10 in a random vector. Agents were repelled by any fabric ridge. Midpoints of the path of travel were stitched 5 oz of plaster was poured at each stitch point. Left to harden overnight The adaptation of the plaster was the most interesting part of the prototype. Some plaster joined with other ows creating more downward force on the material.

POST-PROCESSING: CONCLUSIONS:

s p a c e

s p a c e

65

66

TEST: SETUP: OPERATIONS RULES: POST-PROCESSING: CONCLUSIONS:

Stitching III 24 inch, square frame; 36 inch x 48 inch piece of fabric Stitching Create a 3 stitch along right side of the centroid vector. Create a 5 stitch along left side of the centroid vector. 1 lb of plaster added to the left and right side of the centroid vector. Left to harden overnight Because of the weight of the plaster the most interesting deformations occurred at the edges of the fabric. Areas began to experience both very high and very low levels of tension. CONCLUSIONS Pattern function with an added layer of depth works at the small scale. Questions from the team arise about how to scale the smaller surface deformations into full scall prototypes.

s p a c e

67

68

TEST: SETUP: OPERATIONS RULES: Establish 3 random points @ min. distance of 4 inches.

Push & Twist 24 inch, square frame; 22 inch, square fabric piece. Push / Twist Pushupatvarying heights. Twist each point 720 counter-clockwise.

TEST: SETUP: OPERATIONS RULES:

Subdivide I 24 inch, square frame; 26 inch, square fabric piece. Push / Pull / Stitch / Fold Subdivide surface along diagonal. Pull corners perpendicular to subdivision to ground. Cut 10 inches perpendicular to ridge. Pull one midpoint of cut to ground, other push up. Repeat. Hardened with Plaster of Paris & water mixture (2:1 plaster-water). Heat gun @ approx. 6 inches for 5 minutes. Dry overnight. The rule set for this prototype was global in nature, not local. The resulting form did yield enclosure and spatiality; however, no emergence or local adaptation occurred from the rule set.

POST-PROCESSING: CONCLUSIONS:

Hardened with a mixture of Urea Resin Powder, water and berglass shreds (2:1:1 water-urea-berglass). Heat gun @ approx. 6 inches for 5 minutes. This test further demonstrated hardening capacity of the Urea resin on the fabric as well as the form resistance of the fabric under twisting pressure. Adding berglass shreds to the hardening agent did not provide as much added stiffness as expected. After releasing the fabric (post-hardening), it sunk and did not retain its form.

POST-PROCESSING: CONCLUSIONS:

v o l u m e

v o l u m e

TEST: SETUP: OPERATIONS RULES:

Subdivide II 24 inch, square frame; 26 inch, square fabric piece. Cut/Pull Surface is subdivided based on natural ridges. An 8in cut is placed along the surface vector, creating two new subdivs. Process has 3 generations Hardened with Plaster of Paris & water mixture (2:1 plaster-water). Dry overnight. The rule set for this prototype was global in nature, not local. Although the subdivision rule shows an ability to change the initial formations dramatically.

TEST: SETUP: OPERATIONS RULES:

Subdivision III 70 inch, square frame; 65 inch, square fabric piece. Cut/ Stitch Starting with a triangular folded ridge in the fabric, the edge between each triangle was bisected. The edge was then stitched to its neighbor. Unnished fabric Testing the spatial potential of a subdivision at the larger scale showcase the potential for volumes to be created. The rules we found to be to global in focus, not leading to higher levels of complexity.

POST-PROCESSING: CONCLUSIONS:

POST-PROCESSING: CONCLUSIONS:

v o l u m e

v o l u m e

71

72

TEST: SETUP: OPERATIONS RULES: CONCLUSIONS Singular actions are taken at a system size scale to start to generate a spatial potential. Starting with a larger agengcy move allows for more complex surfaces from fewer generations of actions. POST-PROCESSING: CONCLUSIONS:

Multi-Agent 10 foot frame; 10 foot x 48 inch, fabric piece. Cut/Stich Cut a U shaped slot in fabric at random locations. Stitch inside of cut to nearby location. Unnished felt Very large scale protoype showcase the potential application of the project. The rule was less informative because it developed no diminsionality at scale.

c o m p l e x

73

74

TEST: SETUP: OPERATIONS RULES:

Multi-Agent 10 foot frame; 10 foot x 48 inch, fabric piece. Cut/Stich Cut a U shaped slot in fabric at random locations. Stitch inside of cut to nearby location. Unnished felt Very large scale protoype showcase the potential application of the project. The rule was less informative because it developed no diminsionality at scale.

TEST: SETUP: OPERATIONS RULES:

Multi-Agent 10 foot frame; 10 foot x 48 inch, fabric piece. Cut/Stich Cut a U shaped slot in fabric at random locations. Stitch inside of cut to nearby location. Unnished felt Very large scale protoype showcase the potential application of the project. The rule was less informative because it developed no diminsionality at scale.

POST-PROCESSING: CONCLUSIONS:

POST-PROCESSING: CONCLUSIONS:

c o m p l e x

c o m p l e x

75

76

TEST: SETUP: OPERATIONS RULES:

Multi-Agent 10 foot frame; 10 foot x 48 inch, fabric piece. Cut/Stich Cut a U shaped slot in fabric at random locations. Stitch inside of cut to nearby location. Unnished felt Very large scale protoype showcase the potential application of the project. The rule was less informative because it developed no diminsionality at scale.

TEST: SETUP: OPERATIONS RULES:

Multi-Agent 6 foot square frame; 96 x 48 inch, fabric piece. Cut/ Stitch/Pull Cuts are formed with a recursive pattern seeking a maximin length of 36 and a minimum length of 18. The midpoint of the cut is pulled to the oor. Stitching is required to reach no farther than 2 grid point from a cut and must be 8 grid point in length. Hardened with Plaster of Paris & water mixture (2:1 plaster-water). Dry overnight. Results show an interesting form. The rules that applied did not happen with agency in mind. Final form showcase the linear application of each rule without emergence.

POST-PROCESSING: CONCLUSIONS:

POST-PROCESSING: CONCLUSIONS:

c o m p l e x

c o m p l e x

77

78

Perchita dem si bearchi system at the lautate largest alignim scale to-date and llaborio corernam, consedi dolupdemonstrated its ability to adapt to a tatur amuch ad maio. Itatethan siti alique core si space different the orthogonal quate rerspel enihil in pa dolorem aut frames that had previously characterized lNam acesed magnam eseque dolo every prototype. moluptas doluptatiore rem et voloriosandi ut andit, ofciis accust la et qui The second-floor student loft was used for dolupta volorrum que the installation, andquunderrovid it featured a ceiling inciam volormaio. Mil mincimo mincthat ranged from 5 feet high on one side turiti ut autem ab intium dia qui dolto 15 feet at its peak. Four openings two oratent vellis pratisit fugiasinvel interior, quate one window and a sky light as
well as an oft-used ping pong table acted peliquiatum as primary influences quam onderum the formation seque of the fabric. porro corrorum quo blantia quam quo

The final installation showcased our

maxime nossimil moluptatem fugias The installation essentially acts epeas a sam consequi quaestem inullor spatial while leaving intact rum ea experience nonsed mo vendae nonsecthe previously existing functionality tat voluptur? Iquatem quis derum ut and even as introducing moments of ut veliciis de es es new dolupta tempor function. rescium net laceatum enienem rehent

final prototype
79

80

TEST: SETUP: OPERATIONS RULES:

Final Installation The Loft (RIGHT); 2 6x24 bolts of standard fabric. Cut, Stitch (zip-tie), Pull The rules used in this prototype were dictated by both environmental conditions (openings, ping pong table) and previous prototype information. An initialization of 3 corner pinned points yielded the beginning material condition, which was then altered according to the rules shown in the Material section (pages 14 - 31). LED-Arduino sound modulation; glowing ink. The nal prototype/installation demonstrated the capabilities of the system uncovered up to this point. The rulesets used in its fabrication produced a semi-high-resolution structure that was able to adapt to its environment. Ultimately, it lacked sufcient variation in relation to the specic conditions of the environment.
sky light exterior window

flip!

POST-PROCESSING: CONCLUSIONS:

interior openings

pong table

ENTE

rule set

R THE LOFT

environment

81

82

fast!

simulation

83

84

FASTER!

F
85

86

...

F
87

88

...

F
89

90

...

F
91

92

...

F
93

94

...

F
95

96

...

F
97

98

...

F
99

100

...

F
101

102

...

F
103

104

...

F
105

106

...

F
107

108

...

F
109

110

...

post-processing

111

112

...

F
113

114

...

F
115

116

...

F
117

118

...

F
119

120

...

F
121

122

...

F
123

124

... :)

F
125

126

{e}bot-lutions

//

symb(i)ot

IAM

Design Re_Search
127

newschool of architecture + design