Architectural Applications of Complex Adaptive Systems

TLDR: Methods and case studies of approaching architectural design and fabrication utilizing Complex Adaptive Systems utilizing CASs are presented, which provide frameworks for managing large numbers of elements and their inter-relationships.

Abstract: This paper presents methods and case studies of approaching architectural design and fabrication utilizing Complex Adaptive Systems (CASs). The case studies and observations described here are findings from a continuing body of research investigating applications of computational systems to architectural practice. CASs are computational mechanisms from the computer science field of Artificial Life that provide frameworks for managing large numbers of elements and their inter-relationships. The ability of the CASs to handle complexity at a scale unavailable through non-digital means provides new ways of approaching architectural design, fabrication, and practice.

Full text

Architectural฀Applications฀of฀Complex฀Adaptive฀Systems
Cory฀Clarke฀and฀Phillip฀Anzalone
XO฀(eXtended฀Ofce)฀
Abstract
This฀paper฀presents฀methods฀and฀case฀studies฀of฀approaching฀architectural฀design฀
and฀fabrication฀utilizing฀Complex฀Adaptive฀Systems฀(CASs).฀These฀case฀studies฀
and฀observations฀described฀here฀are฀ndings฀from฀a฀continuing฀body฀of฀research฀
investigating฀applications฀of฀computational฀systems฀to฀architectural฀practice.฀฀CASs฀
are฀computational฀mechanisms฀from฀the฀computer฀science฀eld฀of฀Articial฀Life฀
that฀provide฀frameworks฀for฀managing฀large฀numbers฀of฀elements฀and฀their฀inter-
relationships.฀The฀ability฀of฀the฀CAS฀to฀handle฀complexity฀at฀a฀scale฀unavailable฀
through฀non-digital฀means฀provides฀new฀ways฀of฀approaching฀architectural฀design,฀
fabrication฀and฀practice.
1฀INTRODUCTION
This฀paper฀documents฀ndings฀from฀investigations฀of฀the฀application฀of฀Complex฀
Adaptive฀Systems฀(CASs)฀to฀the฀practice฀of฀architecture.฀฀CASs฀are฀computational฀
mechanisms฀from฀the฀computer฀science฀eld฀of฀Articial฀Life฀that฀provide฀frameworks฀
for฀managing฀large฀numbers฀of฀elements฀and฀their฀inter-relationships.฀Some฀
algorithms฀that฀are฀classied฀as฀CASs฀are฀Cellular฀Automata,฀Lindemayer฀Systems,฀
Turing฀Machines฀and฀Flocking฀algorithms.฀There฀are฀many฀potential฀applications฀for฀
these฀systems฀in฀the฀architectural฀design฀process:
•฀Tools฀for฀the฀automated฀design฀of฀large฀scale฀buildings฀and฀urban฀projects;
•฀Techniques฀for฀the฀design฀of฀serialized฀buildings฀(e.g.฀mass฀housing,฀franchise฀
buildings,฀pre-fabricated฀construction)฀that฀are฀contextually-sensitive฀and฀
differentiated;
•฀Platforms฀for฀roboticized฀self-assembling฀and฀self-adjusting฀buildings;
•฀Methods฀for฀designing฀adaptable฀buildings฀with฀redundant฀structures฀with฀
the฀ability฀to฀more฀successfully฀withstand฀damage฀and฀catastrophic฀events.
Critical฀to฀our฀investigation฀is฀the฀marriage฀of฀CASs฀with฀suitable฀geometric฀and฀
structural฀systems.฀Complex฀computational฀systems฀have฀found฀little฀application฀in฀
contemporary฀architectural฀practice฀because฀their฀effective฀realization฀is฀often฀not฀
possible฀with฀traditional฀geometries,฀such฀as฀the฀Cartesian฀coordinate฀system,฀or฀

traditional฀structural฀methods,฀such฀as฀trabeated฀construction.฀In฀parallel฀with฀our฀
investigation฀of฀applications฀of฀CASs฀to฀architectural฀design฀has฀been฀the฀development฀
of฀formal฀and฀structural฀systems฀with฀geometries฀to฀match฀the฀CAS฀morphology.
This฀paper฀describes:
•฀The฀concept฀of฀CASs฀and฀their฀presentation฀in฀architectural฀terms,฀addressing฀
ideas฀of฀context,฀site,฀program฀and฀form;
•฀Development฀of฀structural฀and฀construction฀systems฀that฀provide฀a฀means฀of฀
constructing฀results฀from฀CAS฀based฀design฀systems;
•฀Case฀studies฀of฀the฀application฀of฀CASs฀in฀architectural฀design.
2฀CAS฀CONCEPTS฀IN฀ARCHITECTURAL฀DESIGN
CASs฀are฀a฀class฀of฀dynamic฀systems฀from฀the฀eld฀of฀Articial฀Life.฀These฀systems฀
typically฀involve฀sets฀of฀discrete฀elements฀that฀change฀state฀(typically฀visualized฀
by฀changes฀in฀color),฀based฀on฀the฀iteration฀of฀a฀simple฀set฀of฀deterministic฀rules฀
(Wuensche,฀Lesser฀1992).฀The฀elements’฀states฀change฀as฀a฀function฀of฀their฀current฀
state฀and฀the฀rules,฀producing฀an฀unpredictable,฀complex฀global฀behavior฀pattern.฀
A฀CAS฀changes฀its฀state฀based฀on฀the฀rules฀of฀the฀system.฀These฀rules฀are฀applied฀to฀
the฀current฀state฀of฀every฀element฀in฀the฀system฀to฀calculate฀each฀element’s฀new฀state.฀
Despite฀the฀fact฀that฀these฀systems’฀behaviors฀are฀dened฀by฀completely฀deterministic฀
rules,฀the฀behaviors฀of฀these฀systems฀(with฀sufcient฀elements฀in฀play)฀are฀so฀complex฀
that฀they฀cannot฀be฀predicted.฀Since฀the฀new฀state฀of฀the฀system฀is฀determined฀by฀the฀
current฀state,฀different฀initial฀states฀of฀a฀CAS฀can฀foster฀entirely฀different฀emergent฀
behavior.฀To฀change฀the฀behavior฀of฀the฀system฀a฀CAS฀can฀merely฀be฀started฀with฀new฀
initial฀state.
Different฀initial฀states฀in฀CASs฀will฀lead฀to฀different฀global฀behavior฀and฀patterns.฀฀In฀
this฀way฀the฀interaction฀of฀the฀elements฀within฀a฀CAS฀can฀be฀thought฀of฀as฀a฀‘calculator’฀
(Neumann,฀1966),฀computing฀outputs฀based฀on฀inputs.฀This฀capacity฀to฀derive฀results฀
from฀a฀complex฀matrix฀of฀inputs฀is฀an฀obvious฀analog฀to฀the฀design฀process฀–฀which฀
must฀take฀the฀complex฀matrix฀of฀requirements฀–฀program,฀site฀and฀structure฀–฀and฀
translate฀them฀into฀an฀architectural฀outcome.
There฀are฀several฀properties฀shared฀by฀CASs฀upon฀which฀we฀have฀focused฀our฀research.฀
The฀properties฀listed฀below฀have฀provided฀us฀with฀a฀means฀of฀linking฀the฀processes฀and฀

behaviors฀of฀CASs฀to฀the฀design฀and฀production฀of฀architecture.฀While฀these฀properties฀
may฀not฀belong฀to฀all฀CASs,฀they฀provide฀a฀list฀of฀traits฀that฀can฀be฀combined฀and฀
exploited฀in฀the฀production฀of฀architectural฀design฀tools.฀The฀properties฀are฀discrete฀
composition,฀algorithmic฀relationships,฀exogenous฀control฀and฀scalability.฀
2.1฀Discrete฀Composition
The฀underlying฀substructures฀of฀CASs฀are฀discrete฀matrices฀of฀elements,฀treated฀as฀
independent฀entities,฀operating฀in฀parallel.฀Despite฀the฀discrete฀nature฀of฀these฀systems฀
at฀the฀local฀scale,฀they฀are฀able฀to฀exhibit฀global฀behaviors฀with฀varying฀degrees฀of฀
continuity.฀The฀trait฀of฀having฀a฀discontinuous฀substructure,฀but฀varying฀continuity฀
at฀the฀level฀of฀superstructure,฀is฀one฀of฀the฀behaviors฀that฀make฀CASs฀suitable฀for฀
translation฀into฀architecture.฀This฀trait฀provides฀an฀analog฀to฀the฀way฀in฀which฀
architecture฀is฀produced;฀assembled฀from฀discrete฀elements฀–฀either฀programmatic฀
(rooms฀and฀zones)฀or฀structural฀(beams฀and฀columns)฀–฀that฀dene฀spaces฀with฀
varying฀degrees฀of฀continuity฀(e.g.฀spatial,฀programmatic,฀acoustical,฀environmental).
2.2฀Algorithmic฀Relationships฀
The฀behavior฀of฀a฀CAS฀is฀prescribed฀by฀sets฀of฀rules฀and฀algorithms฀that฀dictate฀
the฀relationships฀between฀elements฀within฀the฀system.฀฀The฀relationships฀between฀
elements฀in฀a฀CAS฀are฀iteratively฀re-evaluated฀based฀on฀these฀rules,฀providing฀the฀
means฀for฀the฀system฀to฀adapt฀to฀internal฀and฀external฀changes.฀The฀clear฀denition฀of฀
relationships฀between฀elements,฀and฀the฀reiteration฀of฀those฀relational฀rules฀over฀time,฀
nds฀an฀analog฀in฀the฀design฀process฀–฀where฀design฀intentions฀are฀described฀in฀terms฀
of฀relationships฀(privacy,฀lighting,฀proximity),฀and฀the฀nal฀product฀emerges฀from฀the฀
continual฀reiteration฀and฀renement฀of฀these฀relationships.
฀
A฀specic฀example฀of฀this฀can฀be฀seen฀in฀the฀computational฀algorithm฀described฀as฀
ocking.฀Basic฀ocking฀systems฀(Figure฀1)฀dene฀a฀set฀of฀desired฀relationships฀between฀
elements฀in฀a฀ock฀(such฀as฀ideal฀distance฀between฀elements฀and฀preferred฀position฀
relative฀to฀other฀elements)฀and฀attempt฀to฀maintain฀these฀relationships฀by฀continually฀
readjusting฀the฀positions฀of฀each฀element฀(Reynolds,฀1987).฀When฀the฀rules฀of฀proximity฀
and฀position฀are฀applied฀to฀a฀group฀of฀elements฀they฀exhibit฀complex฀adaptive฀behavior฀
similar฀to฀a฀ock฀of฀birds.฀This฀simple฀set฀of฀rules฀of฀proximity฀and฀position฀can฀nd฀
application฀in฀architecture,฀for฀example,฀as฀rules฀for฀positioning฀of฀programmatic฀or฀
structural฀elements.฀฀฀
2.3฀Exogenous฀Control฀(Contextual฀Awareness)
Along฀with฀the฀ideas฀of฀internal฀constraints,฀many฀CASs฀have฀the฀capacity฀to฀react฀to฀
exogenous฀control,฀which฀in฀architectural฀terms฀can฀provide฀mechanisms฀for฀contextual฀
Figure 1 - Paths described by the motion
of ocking particles .

and฀environmental฀awareness,฀and฀inuencing฀the฀system฀to฀incorporate฀additional฀
design฀intentions.฀฀CASs฀have฀two฀primary฀mechanisms฀for฀reacting฀to฀environment฀
and฀context฀–฀initial฀state฀denition฀and฀avoidance฀behavior.฀฀
Since฀all฀CASs฀operate฀on฀a฀deterministic฀set฀of฀rules,฀the฀initial฀state฀of฀the฀system฀
ultimately฀determines฀its฀outcome.฀฀Thought฀of฀in฀this฀way,฀the฀outcome฀of฀a฀CAS฀can฀
be฀seen฀as฀a฀registration฀of฀the฀rules฀upon฀a฀particular฀initial฀state.฀With฀a฀simple฀
CAS,฀such฀as฀Cellular฀Automata,฀the฀entire฀range฀of฀possible฀outcomes฀for฀every฀rule฀
from฀a฀single฀initial฀state฀can฀be฀mapped฀(Wuench,฀Lesser฀1992).฀฀In฀an฀architectural฀
application฀the฀initial฀state฀can฀be฀dened฀in฀terms฀of฀the฀environment,฀and฀the฀
behavior฀of฀the฀CAS฀would฀be฀a฀direct฀expression฀of฀the฀relational฀rules฀of฀the฀system฀
within฀that฀environment.
Along฀with฀the฀ability฀to฀register฀the฀initial฀state฀of฀their฀environment,฀CASs฀can฀react฀
to฀external฀change฀via฀mechanisms฀such฀as฀avoidance฀behavior.฀This฀behavior฀is฀
comprised฀of฀basic฀mechanisms฀that฀monitor฀the฀incremental฀growth฀of฀a฀CAS฀for฀
collisions฀with฀objects฀or฀elds฀in฀the฀environment฀and฀provide฀for฀additional฀reactions.฀
If฀elements฀in฀the฀system฀are฀currently฀on฀a฀collision฀course฀the฀monitoring฀mechanism฀
signals฀for฀new฀behavioral฀rules฀to฀be฀invoked฀that฀react฀to฀the฀potential฀collision.฀
The฀obvious฀architectural฀implication฀of฀the฀collision฀detection฀and฀avoidance฀logic฀
is฀to฀install฀the฀system฀with฀an฀awareness฀of฀surrounding฀buildings฀and฀landscape.฀
However,฀the฀behavior฀can฀also฀be฀applied฀to฀more฀abstractly฀zoned฀volumes฀within฀the฀
systems฀surrounding฀context฀as฀well.฀The฀growth฀of฀a฀system฀could฀respond฀to฀zoning฀
envelopes,฀acoustical฀envelopes,฀sun฀and฀shade฀volumes,฀view฀corridors,฀program฀
areas,฀and฀any฀other฀abstract฀constraints฀that฀can฀be฀represented฀volumetrically.
Collision฀detection฀can฀be฀further฀extended฀as฀a฀mechanism฀for฀control฀by฀inverting฀the฀
logic฀of฀avoidance฀behavior฀so฀that฀the฀collision฀detection฀mechanism฀can฀be฀used฀as฀
a฀tool฀for฀searching฀out฀attractive฀areas฀of฀an฀architectural฀context.฀If฀the฀avoidance฀
behavior฀were฀to฀be฀changed฀to฀tend฀towards฀potential฀collision฀instead฀of฀away,฀the฀
morphological฀tendencies฀of฀the฀CAS฀will฀grow฀towards฀specic฀areas฀in฀a฀site฀that฀are฀
designated฀as฀desirable.฀Furthermore฀the฀volumes฀of฀desirability,฀and฀undesirability,฀
can฀be฀weighted฀in฀inuence฀to฀produce฀a฀multivalent฀map฀of฀context.
2.4฀Scalability
Many฀CASs฀exhibit฀a฀fractal฀behavior฀-฀the฀behavior฀of฀a฀small฀number฀of฀elements฀is฀
congruent฀to฀the฀same฀system฀with฀exponentially฀more฀components.฀This฀scalability฀
of฀behavior฀has฀two฀benets฀to฀the฀study฀and฀implementation฀within฀architectural฀
practice.฀The฀rst฀is฀one฀of฀study฀and฀testing;฀since฀the฀behavior฀of฀a฀large฀system฀

is฀similar฀to฀that฀of฀a฀small฀system,฀methods฀for฀design฀can฀be฀modeled฀in฀a฀limited฀
capacity฀and฀still฀translate฀to฀the฀larger฀scale฀of฀a฀building฀or฀urban฀project.฀The฀
second฀benet฀of฀the฀scalability฀of฀CASs฀is฀their฀potential฀to฀be฀nested฀fractally.฀Their฀
scalability฀makes฀them฀essentially฀scale-less;฀a฀particular฀set฀of฀rules฀that฀functions฀
well฀as฀a฀global฀organization฀system฀could฀be฀used฀at฀the฀scale฀of฀an฀urban฀project,฀
while฀each฀‘cell’฀could฀in฀turn฀be฀lled฀with฀a฀smaller฀version฀of฀a฀CAS฀with฀a฀detailed฀
behavior฀at฀the฀scale฀of฀building฀or฀dwelling฀unit.
฀
3฀STRUCTURAL฀SYSTEMS
The฀characteristics฀of฀the฀organizational฀strategy฀outlined฀above,฀and฀the฀diverse฀range฀
of฀possible฀complex฀forms฀that฀may฀result,฀require฀a฀structural฀solution฀with฀similar฀
geometric฀and฀algorithmic฀traits.฀Our฀research฀focused฀upon฀the฀three-dimensional฀
differential฀space-truss.฀
The฀traditional฀space-truss฀is฀a฀lattice฀structure฀of฀standard฀elements,฀typically฀leading฀
to฀architecture฀of฀regular฀geometrical฀forms,฀as฀in฀the฀geodetic฀domes฀of฀Buckminster฀
Fuller฀and฀projects฀such฀as฀I.M.฀Pei’s฀Javits฀Convention฀Center฀in฀New฀York.฀The฀
traditional฀space-truss฀employs฀standard฀elements฀because฀of฀constraints฀of฀design,฀
analysis฀and฀fabrication฀–฀constraints฀now฀surmountable฀through฀computer-based฀
techniques.฀The฀differential฀space-truss฀uses฀non-standard฀elements;฀by฀allowing฀
each฀element฀to฀be฀unique฀it฀can฀take฀on฀complex฀three-dimensional฀curvilinear฀form฀
as฀well฀as฀basic฀linear฀geometry.฀
During฀our฀research฀we฀built฀and฀tested฀several฀scale฀models฀of฀three-dimensional฀
differential฀space-trusses฀with฀non-uniform฀elements฀(Figure฀2).฀The฀manufacturing฀
of฀the฀structural฀elements฀requires฀the฀use฀of฀Computer฀Numerically฀Controlled฀(CNC)฀
fabrication฀techniques.฀The฀specic฀biases฀and฀limitations฀of฀these฀construction฀
methods฀can฀be฀built฀into฀the฀relational฀rules฀of฀the฀CASs฀as฀further฀means฀of฀control.฀
The฀differential฀space-truss฀has฀three฀major฀traits฀that฀match฀many฀of฀the฀behaviors฀
and฀properties฀of฀CASs:฀discrete฀composition,฀lattice฀geometry฀and฀scalability.
3.1฀Discrete฀Elements
Space-truss฀systems฀are฀comprised฀of฀two฀basic฀components,฀linear฀struts฀and฀
connecting฀nodes;฀through฀the฀manipulation฀of฀these฀two฀types฀of฀elements฀the฀system฀
can฀yield฀a฀diverse฀range฀of฀complex฀three-dimensional฀forms.฀The฀uid฀nature฀in฀
which฀a฀space-truss฀can฀transition฀between฀curvilinear฀form฀and฀linear฀geometry฀
is฀similar฀to฀that฀of฀monolithic฀structural฀systems฀such฀as฀cast฀concrete,฀yet฀it฀is฀
constructed฀from฀repetitive฀discrete฀elements.฀Its฀range฀of฀formal฀expression,฀coupled฀
Figure 2 - Fabricated prototype of
differential space-truss.