Post on 28-Mar-2015
USER-LED DEVELOPMENT OF ANINTERACTIVE EVOLUTIONARY DESIGN (IED) SYSTEM
Ian Graham
ADAPTIVE COMPUTING IN DESIGN AND MANUFACTURE 2008SPECIAL SESSION ON USER-CENTRED INTELLIGENT SYSTEMS
GALLERY
Research Associate · Design Practice Research Group · Department of Design and Technology
ConclusionsQuestions and discussion
User-led development of an IED system
SEMINAR FORMAT
IED system development & adoption by industryThe Evolutionary Form Design (EFD) systemOutreach activities and user trials
User-led development of an IED system
IED system development & adoption by industry
With a few notable exceptions in design exploratione.g. Genometri, AffinnovaIED does not seem to have been widely adopted by designers
User-led development of an IED system
IED system development & adoption by industry
Other reasons to explore: Lack of awareness in the design industry? Instinctive resistance to the technology? The ‘look’ or ‘style’ of the output of many existing IED systems?
Acknowledged problems: Operator fatigue
User-led development of an IED system
IED system development & adoption by industry
‘Limited end-user involvement during IED system development’Are researchers reluctant to test their systems in the real world?
‘The current limitations are obvious without user-involvement'Are these limitations inherent in the technology or are we simply not there yet in terms of development?
User-led development of an IED system
IED system development & adoption by industry
Any questions or comments on:
IED adoption in industry ?
The way IED systems are developed ?
EFD
1. Implementation2. Representation3. Noteworthy features
Control of Boolean interactions Edge Blending Strategies Automation of geometric and aesthetic optimisation
4. Limitations5. Strengths
THE SYSTEM
Implementation
THE EFD SYSTEM
ImplementationScreenshot
THE EFD SYSTEM
Representation
THE EFD SYSTEM
Representation
create unite subtract
intersection
THE EFD SYSTEM
Representation
vector direction - x, y, z
primitive typeorigin - x, y, z
sign (boolean operator)
size multipliershape proportions - x, y, z
Genotype
blend multiplierblend frequency
blend radii
Boolean interaction
team-forming tactics
THE EFD SYSTEM
Representation
vector direction - x, y, z
primitive type
origin - x, y, z
sign (boolean operator)
size multiplier
shape proportions - x, y, z
CHROMOSOMES
type chromosome
segment bin. dec. decoded
block
sphere
cylinder
cone
create
unite
subtract
intersect
00 0
01 1
10 2
11 3
00 0
01 1
10 2
11 3
sign chromosome
segment bin. dec. decoded
cone
block
sphere
cylinder
sphere
create
intersect
create
unite
subtract
THE EFD SYSTEM
Implementation
= Dominant features carried over four generations
generation one generation two generation three generation four
g4p8-olive
Ancestral diagram
THE EFD SYSTEM
Implementation
G e n e r a t i o n O n e
G e n e r a t i o n T w o
G e n e r a t i o n T h r e e
Generation Four
Family tree diagram of ‘Cobra’ sculpture
THE EFD SYSTEM
Control of Boolean interactions Edge Blending Strategies The team-forming algorithm Automation of geometric and aesthetic optimisation
Noteworthy features
THE EFD SYSTEM
Control of Boolean interactions Edge Blending Strategies The team-forming algorithm Automation of geometric and aesthetic optimisation
Noteworthy features
THE EFD SYSTEM
Post-creation Boolean operationsControl of Boolean interactions
1
2
3
4Blue - createMagenta - uniteOrange - subtract
5
9
6
7
8
THE EFD SYSTEM
Control of Boolean interactions
All interfering bodies Selected targetsAdjacent creation order
conecreate
spherecreate
spheresubtract
cylinderunite
PrimitiveSign
The three methods of post-creation Boolean target selection
THE EFD SYSTEM
Edge Blending StrategiesCreation of the ‘Parrot Fish’ sculpture
‘Parrot Fish’ before and after genetic blending
THE EFD SYSTEM
Control of Boolean interactions Edge Blending Strategies The team-forming algorithm Automation of geometric and aesthetic optimisation
Noteworthy features
THE EFD SYSTEM
Edge Blending StrategiesSmall (1-2mm) blends
smooth edges highlights
THE EFD SYSTEM
Edge Blending StrategiesMedium (5-10mm) blends
round edges
fillets
fillets
THE EFD SYSTEM
Edge Blending StrategiesLarge (25-100mm) blends
new shapes created
THE EFD SYSTEM
Edge Blending StrategiesWhole-object (post-Boolean) blending
1. Sphere created
2. Interfering block created
3. Block subtracted from sphere
4. Object edges blended
THE EFD SYSTEM
1. Sphere created
3. Block edges blended
4. Blended block subtracted from sphere
2. Interfering block created
Edge Blending StrategiesPrimitive (pre-Boolean) blending
THE EFD SYSTEM
Edge Blending Strategies
Edge Blending StrategiesComparing equivalent objects with alternative blending strategies
THE EFD SYSTEM
Control of Boolean interactions Edge Blending Strategies The team-forming algorithm Automation of geometric and aesthetic optimisation
Noteworthy features
THE EFD SYSTEM
Comparison of the standard and team-forming modesThe team-forming algorithm
Object formed from 1 genotype Object formed from a team of 5 phenotypes
1 genotype (30 chromosomes)
object is a single phenotypegenotype
(6 chromosomes)object is a team of phenotypes
5 phenotypes
3 segment chromosome
2 bit segment
1 segment chromosome
6 bit segment
THE EFD SYSTEM
The team-forming algorithm
EXAMPLE POPULATIONS
THE EFD SYSTEM
The team-forming algorithm
g3t10-grey
g2t2-green g2t6-yellow g2t5-magenta
Parents of ‘g3t10-grey’ team-members with associated objects
P62
P53
P32P58
P61P50
P8
P41
P21
P39
P66
P53
P42
THE EFD SYSTEM
Control of Boolean interactions Edge Blending Strategies The team-forming algorithm Automation of geometric and aesthetic optimisation
Noteworthy features
THE EFD SYSTEM
Automation of geometric and aesthetic optimisation
1
initial population
17
population converged
5
object type established
11
solution found
target: bounding box, volume of 0.1000m3
solution: block, volume of 0.1005m3 (31x47x69cm)
achieved in: 11 generations (population size: 14)accuracy: (0.1005 - 0.1) 0.1 = 0.005 = 0.5%
THE EFD SYSTEM
target: bounding box, volume of 0.1000m3
solution: block, volume of 0.1005m3 (31x47x69cm)
achieved in: 11 generations (population size: 14)accuracy: (0.1005 - 0.1) 0.1 = 0.005 = 0.5%
Automation of geometric and aesthetic optimisation
1 2 3 7654 14 15 16 201918178 9 13121110
0.4
0.5
0.6
0.3
0.2
0.1
0.0
Generation
Maximum Fitness
Mean Fitness
Show
THE EFD SYSTEM
Automation of geometric and aesthetic optimisation
Volume - Target : 0.05m3
0.04998611 0.028%16solutiongeneration accuracyconvergence
Optimisation examples
THE EFD SYSTEM
Automation of geometric and aesthetic optimisation
Surface Area - Target : 0.8m2
0.799749 14 0.033%
solutiongeneration accuracyconvergence
Optimisation examples
THE EFD SYSTEM
Automation of geometric and aesthetic optimisation
Dimensions - Target : 201050cm
20105117 0.667%25solutiongeneration accuracyconvergence
Optimisation examples
THE EFD SYSTEM
Automation of geometric and aesthetic optimisationMachine-based aesthetic optimisation
‘Computer Aided Aesthetics inEvolutionary Computer Aided Design’
Aesthetic measures represented geometrically:
Simplicity · Stability · Edge Smoothness · Face SmoothnessExpandability · Surface Area · Volume · Hardness · Softness
THE EFD SYSTEM
1. Lack of hands-on control
Limitations
2. Can be ‘hit-and-miss’
3. Restricted range of 3D forms
THE EFD SYSTEM
1. Lack of hands-on controlObjects cannot be edited directly using the CAD interface and then returned to the population for further evolutionary development.
Limitations
2. Can be ‘hit-and-miss’
3. Restricted range of 3D forms
THE EFD SYSTEM
1. Lack of hands-on controlObjects cannot be edited directly using the CAD interface and then returned to the population for further evolutionary development.
Limitations
2. Can be ‘hit-and-miss’In the early stages of evolution (to ensure variability), especially if the user is aiming for a particular form.
3. Restricted range of 3D forms
THE EFD SYSTEM
1. Lack of hands-on controlObjects cannot be edited directly using the CAD interface and then returned to the population for further evolutionary development.
Limitations
2. Can be ‘hit-and-miss’In the early stages of evolution (to ensure variability), especially if the user is aiming for a particular form.
3. Restricted range of 3D formsDesigners of consumer products rarely use simple solid modeling these days, tending to use hybrid surface/solid modelers.
THE EFD SYSTEM
1. Lack of hands-on controlObjects cannot be edited directly using the CAD interface and then returned to the population for further evolutionary development.
Limitations
Any other limitations ?
2. Can be ‘hit-and-miss’In the early stages of evolution (to ensure variability), especially if the user is aiming for a particular form.
3. Restricted range of 3D formsDesigners of consumer products rarely use simple solid modeling these days, tending to use hybrid surface/solid modelers.
THE EFD SYSTEM
Strengths
2. No preliminary modelling needed
1. Generic
3. Conceptually simple
4. Less of an inherent ‘style’ then other IED systems
THE EFD SYSTEM
Strengths
2. No preliminary modelling needed
1. GenericNot focussed on any particular product.Offers flexibility across a wide spectrum of design,(consumer products, furniture, architecture, sculpture etc.).
3. Conceptually simple
4. Less of an inherent ‘style’ then other IED systems
THE EFD SYSTEM
Strengths
2. No preliminary modelling neededBypasses users’ preconceptions.Users need no prior experience in using CAD
1. GenericNot focussed on any particular product.Offers flexibility across a wide spectrum of design,(consumer products, furniture, architecture, sculpture etc.).
3. Conceptually simple
4. Less of an inherent ‘style’ then other IED systems
THE EFD SYSTEM
Strengths
2. No preliminary modelling neededBypasses users’ preconceptions.Users need no prior experience in using CAD
1. GenericNot focussed on any particular product.Offers flexibility across a wide spectrum of design,(consumer products, furniture, architecture, sculpture etc.).
3. Conceptually simpleUsers quickly grasp how their interactions control the evolutionary process
4. Less of an inherent ‘style’ then other IED systems
THE EFD SYSTEM
Strengths
2. No preliminary modelling neededBypasses users’ preconceptions.Users need no prior experience in using CAD.
1. GenericNot focussed on any particular product.Offers flexibility across a wide spectrum of design,(consumer products, furniture, architecture, sculpture etc.).
3. Conceptually simpleUsers quickly grasp how their interactions control the evolutionary process.
4. Less of an inherent ‘style’ then other IED systemsCapable of creativity-enhancing and innovative form design.
THE EFD SYSTEM
Strengths
2. No preliminary modelling neededBypasses users’ preconceptions.Users need no prior experience in using CAD
1. GenericNot focussed on any particular product.Offers flexibility across a wide spectrum of design,(consumer products, furniture, architecture, sculpture etc.).
3. Conceptually simpleUsers quickly grasp how their interactions control the evolutionary process
4. Less of an inherent ‘style’ then other IED systemsCapable of creativity-enhancing and innovative form design.
THE EFD SYSTEM
Any other strengths ?
Any questions or comments on the EFD system ?
THE EFD SYSTEM
OUTREACH ACTIVITIES AND USER TRIALS
Engaging design professionals
Creativity experiments
Schools and creative engagement
Design-recreation activities
User-integrated design
Engaging design professionals
1. Introduce the EFD system to a variety of industrial and product design fields, such as: Consumer products Furniture Automotive styling Architecture Sculpture
2. Workshop
3. Exhibition
4. Interviews
Creativity experiments
Comparative experiments under controlled conditions:
1. ‘Designers’ vs. ‘Non-designers’, both using the EFD system
2. A group using the EFD system vs. a group using other form-finding techniques
Results judged via an exhibition or on-line
Any other ideas for experiments ?
Schools and creative engagement
Engage young people in schools Early exposure to CAD Combine with other technologies
Design-recreation activities
Creative recreation - evolving virtual sculptures
Assign material properties and create photo-realistic prints
Or rapid prototype and hand finish solid models
User-integrated design
Previous work in the department inhibited by users’ lack of ability in CAD form creation
EFD system can overcome this by generating forms
How well will the outputs from consumers using the EFD system be integrated into customised consumer products by professional designers ?
OUTREACH ACTIVITIES AND USER TRIALS
Engaging design professionals
Creativity experiments
Schools and creative engagement
Design-recreation activities
User-integrated design
Any questions or comments on activities, trials and experiments ?
Conclusions
User-led development of an IED system
SEMINAR FORMAT
IED system development & adoption by industryThe Evolutionary Form Design systemOutreach activities and user trials
Questions and discussion
Conclusions
Principle of user-centred design should be applied to the development of Interactive Evolutionary Design systems
This will increase awareness if IED amongst the design community, creating more of a pull for the technology
Collaboration between computer scientists and human scientists will provide a boost to the progression of IED research
IED can contribute to a change in the way things are designed, and the way users can be involved in design processes
These principles will be adopted in development of the EFD system, which is felt is an ideal candidate for such research methods.
Questions and discussion
User-led development of an IED system
SEMINAR FORMAT
IED system development & adoption by industryThe Evolutionary Form Design systemOutreach activities and user trials
Conclusions
Questions and discussion
BACK TO START
Interactive Evolutionary Design in general Awareness and use of IED in industry Barriers to widespread adoption Sectors to target When should IED systems be exposed to users?
The EFD system Usefulness Potential applications Our proposed activities
User-led development of an IED system:
Adoption by industry The EFD systemIED Development Outreach activities
Conclusions, questions and discussion:
ConclusionsDiscussion points
INDEX
GALLERY