1-
AGENT BEHAVIOUR | AADRL 2020
CONTENTS CHAPTER I | BRIEF & RESEARCH 1.1 Introduction & Workflow 1.2 Research on Springs & Networks
01 02
CHAPTER II | HOMOGENEOUS SYSTEM | CUBE 2.1 Spring Properties 2.2 Particle Properties
03 04
CHAPTER III | CUBE | INTRODUCING PATTERNS 3.1 Different Connectivity 3.2 Intoducing Attractors
CHAPTER IV | CUBE | TWO POPULATIONS 4.1 Spring Properties 4.2 Particle Properties
CHAPTER V | TWO POPULATIONS | FURTHER EXPLORATION 4.1 Spring Properties 4.2 Particle Properties
05 09 15
17 55
AGENT BEHAVIOUR | AADRL 2020
-4
CHAPTER I BRIEF AND RESEARCH
1.1 INTRODUCTION & WORKFLOW
The workshop II, Agent Behavior, part of the Design Research Laboratory (DRL) program at Architectural Association School of Architecture in London is held by the tutors Mustafa EI-Sayed and Aleksandar Bursac. The work presented in this booklet was produced by Amin Yassin(Israel), Daphne Drayiou(Greece), Panayiota Tsaparikou(Greece). Agent behaviour workshop aimed to develop emergent behaviors across a population by changing partcle attributes. We used Computer Animation Physics and more specifically the Maya Particle Engine in order to control the physical properties of the particles. Including Grasshopper in our workflow allowed us to control and variate particle and spring attributes. Our exploration focused firstly on the particle and spring behavior, the interaction between these two, and how these behaviors are scaled across populations. Following that we study the interaction between two spring particle populations and the potential to scale up this exploration in two larger populations which form a strong relationship of influence. In our research we focus specifically on spring particle systems with static set-outs and are not influenced by external forces of their environment. Rhino Grasshopper, Autodesk Maya and Adobe Premiere Pro were the software tools used to simulate geometry and evaluate the behavior of the connections.
AGENT BEHAVIOUR | AADRL 2020
-6
WHAT IS A SYSTEM?
NETWORK STUDY OPEN
><
CLOSED
STAR
HIERARCHICAL
MESH
radial
tree
grid
RING
TOPOLOGY
WORKFLOW DIAGRAM 7-
[INPUT GEOMETRY] GEOMETRIC MANIFOLD WHICH ACTS AS A SET OUT GEOMETRY
AGENT BEHAVIOUR | AADRL 2020
RHINO GRASSHOPPER
DEFINING SPRING PARTICLE SYSTEM BEHAVIOR
SPRING PROPERTIES
PARTICLE PROPERTIES
rest length
A particle system is usually defined as a highly chaotic system. However, introducing springs between the particles makes them highly constrained. The linkages provide a level of constraint and control, we can variate the level of control them by changing the attributes of springs. Stiffness is defined as the extent to which a spring resists deformation in response to an applied force. Higher stiffness values give us a loose bonding beween the particles while lower stiffness values offer a tight bonding between the particles. Spring 'rest length' is the length of the spring and its ability to contract or extract. While, spring 'connectivity' deals with the number of connections forged between the different particles which influence the particle movements. Particle 'stickiness' is the particle's ability to stick to one another. Particle 'bounce' is their ability to bounce when they touch each other. Particle 'repulsion' is the extent to which particles want to repel other particles.
contraction
extraction
connections
instability
stability
stiffness
soft
AGENT BEHAVIOUR | AADRL 2020
rigid
-8
RESEARCH
[ INPUT GEOMETRY ]
[ OUTPUT ATTRIBUTES ]
GRASSHOPPER
MAYA MEL
Spring Attributes
[ SIMULATION ] MAYA PHYSICS ENGINE
Rest length Connectivity Stiffness
Key-Framing
Spring – Particle system
Particle Attributes
9-
Mass Radius Collide Width Scale Color Stickiness Bounce
AGENT BEHAVIOUR | AADRL 2020
AGENT BEHAVIOUR | AADRL 2020
- 10
CHAPTER II HOMOGENEOUS SYSTEM BASIC BEHAVIORS
11 -
AGENT BEHAVIOUR | AADRL 2020
REST LENGTH - STIFFNESS
Stiffness
2.1. SPRING PROPERTIES
0.00
0.50
AGENT BEHAVIOUR | AADRL 2020
1.00
1.50
Rest Length
- 12
2.1. SPRING PROPERTIES
Rest Length
REST LENGTH - STIFFNESS
0.50
0.50
1.50
5.00
Stiffness
1.50
13 -
AGENT BEHAVIOUR | AADRL 2020
MASS - RADIUS - BOUNCE - COLLIDE WIDTH SCALE
Rest Length
2.1. PARTICLE PROPERTIES
0.50
1.00
5.00
10.00
Mass
1.50
AGENT BEHAVIOUR | AADRL 2020
- 14
2.1. PARTICLE PROPERTIES
Rest Length
MASS - RADIUS - BOUNCE - COLLIDE WIDTH SCALE
0.50
0.10
0.20
0.30
Radius
1.50
15 -
AGENT BEHAVIOUR | AADRL 2020
MASS- RADIUS - BOUNCE - COLLIDE WIDTH SCALE
Rest Length
2.1. PARTICLE PROPERTIES
0.10
1.00
AGENT BEHAVIOUR | AADRL 2020
2.00
2.20
Bounce
- 16
2.1. PARTICLE PROPERTIES
Rest Length
MASS - RADIUS - BOUNCE - COLLIDE WIDTH SCALE
0.10
1.00
1.50
2.00
Collide Width Scale
0.50
17 -
AGENT BEHAVIOUR | AADRL 2020
AGENT BEHAVIOUR | AADRL 2020
- 18
CHAPTER III INTRODUCING PATTERNS
19 -
AGENT BEHAVIOUR | AADRL 2020
SPRING LENGTH (CONTRACTION/EXTRACTION)
CONNECTIVITY
FOCAL POINT (ATTRACTOR)
AGENT BEHAVIOUR | AADRL 2020
- 20
Rest Length
REST LENGTH - CONNECTIVITY - MASS - RADIUS
Inner point attractor
21 -
Outer point attractor
Outer curve attractor
AGENT BEHAVIOUR | AADRL 2020
Inner attractor
REST LENGTH - CONNECTIVITY - MASS - RADIUS
1.00
1.50
Rest Length
Outer attractor
0.50
AGENT BEHAVIOUR | AADRL 2020
- 22
Inner attractor
REST LENGTH - CONNECTIVITY - MASS - RADIUS
1.00
1.50
Rest Length
Outer attractor
0.50
23 -
AGENT BEHAVIOUR | AADRL 2020
AGENT BEHAVIOUR | AADRL 2020
- 24
CHAPTER IV CUBE | TWO POPULATIONS
25 -
AGENT BEHAVIOUR | AADRL 2020
4.1 CUBE TWO POPULATION: DIAGRAMS In order to achieve a 'behavior' within the population, various attributes within the particlespring set-outs were testeded. These include spring attributes (such as spring rest-length, spring stiffness, connectivity) and particle attributes (such as particle stickiness, particle bounce, and particle repulsion force).
SPRING LEGNTH (REST LEGNTH)
CONNECTIVITY
PARTICLE STICKINESS
PARTICLE REPULSION
AGENT BEHAVIOUR | AADRL 2020
- 26
4.1 CUBE TWO POPULATION In this stage, we tested an interaction between two-particle populations. The particles are divided into "active" particles (blue) and "passive" particles (red) in a ratio of 1/10 (the number of the blue agents is 48 and the number of the red agents is 489). The movement in the arrangement is triggered by an increase in the blue particles spring length only, while the springs connecting the red particles remains unchanged in all tests. The spring length property, which can either contract or extract, is used in order to create the interaction between the two populations.
FRONT VIEW
Population A
TOP VIEW
Population B
agent numbers
agent numbers
489
48
SPRING SIMULATION (FRONT VIEW) 27 -
AGENT BEHAVIOUR | AADRL 2020
4.2 SPRING ATTRIBUTES: SPRING STIFFNESS In the initial tryouts, the interaction between the two populations was tested with different spring 'stiffness' values. Stiffness is defined as the extent to which a spring resists deformation in response to an applied force (the more flexible an object is, the less stiff it is). In each test, we changed the stiffness value (of the red springs only) which resulted in different red particle distribution caused by the movement of the blue, active particles. While most tests had a random or failed result, high spring stiffness values combined with bigger repulsion force from the particles themselves, created a more organized distribution, such as TEST C. In this test, the effect of the blue particles created a localized effect on the red particles; particles in the vicinity of the blue particles are more affected than the further ones.
TEST A
The renders of the springs are used as a tool to evaluate the effect of the active particles on the passive particles and show the localized deformation (like in Test C), where the rest of the cube arrangement remains unaffected. TEST B
TEST C Population A
TEST A TEST B TEST C TEST D
Population B
stiffness repulsion value value
stiffness repulsion value value
0.2 20 0.2 20
0.2 0.2 0.2 0.2
1 1 7 17
1 1 1 1
TEST D AGENT BEHAVIOUR | AADRL 2020
- 28
4.3 CUBE TWO POPULATION: TEST VIEW 1 Using the chosen particle-spring attributes (high stiffness in the springs with high repulsion value of the particles) allows creating control of the blue particles over the red ones within the arrangement. The arrangement transforms from the static shape of the cube to a more fluid form which exhibits the active particles' impact on the passive ones. The distribution that is created resembles a shell-like shape with holes around the blue particles caused because of their repelling force.
Frame 0
Frame 200
Frame 400
Frame 600
Frame 200
Frame 400
Frame 600
VIEW 2
Frame 0
29 -
AGENT BEHAVIOUR | AADRL 2020
4.4 CUBE TWO POPULATION: RESULTS
The renders of the springs are used as a tool to evalute the affect of the active particles on the passive particles and show the resulted deofrmation. The renders show different stages within the evolution from the initial set-out (cube... Frame 00) to the shell (Frame 800).
FRAME 400 AGENT BEHAVIOUR | AADRL 2020
FRAME 600
FRAME 600
FRAME 600
FRAME 600 - 30
Starting with the same set-out (and the same red-blue distribution), we get differentiated particle outcomes through varying the particle attributes in a gradient way. So, by densifying some areas and thinning other areas' particle properties a different red distribution. The varying particle values used in tests: radius, bounce, and stickiness. By varying these a different distribution of passive particles is achieved each time.
BOUNCE
4.5 PARTICLE ATTRIBUTES: BOUNCE & STICKINESS
STICKINESS
The 'Bounce' and 'Stickiness' of the particles are introduced with different particle radius in order to enable the particles to collide (and therefore to stick together or to bounce away from each other).
31 -
AGENT BEHAVIOUR | AADRL 2020
4.6 PARTICLE ATTRIBUTES: RADIUS
Different distributions of particle radius from different directions: top-down, down-top, frontback, back-fron, inside-out, outside-in.
AGENT BEHAVIOUR | AADRL 2020
- 32
REPULSION
4.7 PARTICLE ATTRIBUTES: REPULSION & REST LEGNTH
REST LEGNTH
Tests showing different rest-length in the blue springs and different repelling force from the blue particles used as a method to implement control of the blue (active) particles over the red (passive) particles. The blue particles are used as a way to sculpt the red.
33 -
AGENT BEHAVIOUR | AADRL 2020
AGENT BEHAVIOUR | AADRL 2020
- 34
CHAPTER ΙV TWO POPULATIONS | FURTHER EXPLORATION
35 -
AGENT BEHAVIOUR | AADRL 2020
5. FURTHER EXPLORATION
Following the research in the cube the same logic is applied a larger scale. Setting a distribution of the two populations we are interested in pushing to an extreme the ratio of the two populations, choosing a ratio of 1 to 21. The blue population is introduced as the leader of the movement while the red gets passively influenced. we start analyzing different spring properties regarding the strength of the connection. The blue population expands with highly aggressive collision attributes and as it moves it starts “carving the red one”. Weak or strong connections applied in the passive population allow different type and amount of influence from the blues. Firstly we introduce a cataloguing of the basic behaviors for this distribution of particles. The rendering of the springs is introduced as an evaluation criteria for our system. It is evident that in the strong spring system the population is being deformed in specific areas While in the case of weak spring connections the red population acts almost as a liquid, Ηowever it seems harder to be controlled by the repulsive properties of the blue.
AGENT BEHAVIOUR | AADRL 2020
- 36
5.1 SPRING PROPERTIES
37 -
AGENT BEHAVIOUR | AADRL 2020
5.1 SPRING PROPERTIES | REST LENGTH
Population A Rest Length changes according to a focal point
AGENT BEHAVIOUR | AADRL 2020
Population B
stiffness rest value length
stiffness rest value length
0.2
0.2
1.00-10.00
1
- 38
5.2 PARTICLE PROPERTIES | STICKINESS
STICKINESS = 1.00 - 50.00
STICKINESS = 1.00 - 50.00
Ιt seems harder to be controlled by the repulsive properties of the blue. Based on this statement we are examining tinteraction of these two populations varying the particle attributes according to gradient maps while keeping strong spring connections. We started exploring the potential patterns by altering the pull and push force for each cluster of the blue population. So we examine how altering radius gradiently affects the deformation. How sticky particles in different sides of the tower could alter it shape and how the behavior of a resistant shell with a soft interior is getting differentiated from the behavior of a soft and sticky exterior with a rigid core.
Population A
39 -
Population B
stiffness rest value length
stiffness rest value length
0.2
0.2
1.00-10.00
1
AGENT BEHAVIOUR | AADRL 2020
5.2 PARTICLE PROPERTIES | RADIUS
0.10 <RADIUS< 0.49
AGENT BEHAVIOUR | AADRL 2020
0.10 <RADIUS< 0.49
0.10 <RADIUS< 0.49
- 40
5.2 PARTICLE PROPERTIES | RADIUS SPRING PROPERITES | STIFFNESS
0.5< STIFFNESS <100
0.5< STIFFNESS <100
Following these experiments we are attempting to push the repulsion force of the two particle systems to an extreme. As the blue clusters move out of the initial set-out they start to influence the passive population massively. We Keep the collision properties almost the same in these series of simulations while we are trying to choreograph the red cloud behavior variating its particle properties. The behavior which emerges is no more fragmented in distinct areas and a more homοgeneous behavior appear in the red population. The blue can sculpte the behavior of the red by distance remaining balanced. The embedded spring attributes and the variation of the particle properties gave us the behaviour of a system which is achieved by the choreography of two.
Population A stiffness rest value length 0.2
41 -
1.00-10.00
Population B stickiness 100
stiffness rest value length 0.2
1
stickiness 100
AGENT BEHAVIOUR | AADRL 2020
FURTHER EXPLORATION | HIGH REPULSIVE PROPERTIES | CHOREOGRAPHING THE TWO
max stickiness
max radius
max stiffness
AGENT BEHAVIOUR | AADRL 2020
- 42
43 -
AGENT BEHAVIOUR | AADRL 2020
FURTHER EXPLORATION | HIGH REPULSIVE PROPERTIES | PASSIVE RED CLOUD
max stickiness
max radius
max stiffness
AGENT BEHAVIOUR | AADRL 2020
- 44
45 -
AGENT BEHAVIOUR | AADRL 2020
FURTHER EXPLORATION | HIGH REPULSIVE PROPERTIES | PASSIVE RED CLOUD
AGENT BEHAVIOUR | AADRL 2020
- 46
47 -
AGENT BEHAVIOUR | AADRL 2020
AGENT BEHAVIOUR | AADRL 2020
- 48
49 -
AGENT BEHAVIOUR | AADRL 2020
AGENT BEHAVIOUR | AADRL 2020
- 50