An Artificial Life Approach to Configuring Architectural Space

An Artificial Life Approach to Configuring ArchitecturalSpace

Tim Ireland11De Montfort [email protected]

This paper presents a method of configuring architectural space that articulatesthe coupling of an organism with its environment; expressing the spatiality ofunfolding engagement in the world. The premise is that space is a consequence ofcohesion, effected through constraints and processes of enaction. An ArtificialLife model is presented as an analogue of a bottom-up approach to architecturaldesign that takes into account that we as organisms interact with our ever presentchanging environment and redefine our spatial domain depending on our sensoryinteraction with said environment.

Keywords: Configuration, Agency, Agent-based modelling, Self-organisation

INTRODUCTIONSpatial problems are complex. A key constraint inarchitectural practice is the general reliance on tra-ditional methods to organise architectural layouts,which tend to flatten spatial problems into some-thing quantifiable so they can be managed andplanned. Approaching the configuration of space inthe standard way raises the question whether anyrichness is lost? There is often a qualitative discon-nect between the articulation of spatiality in the builtenvironment and the spatiality of being. Materialproperties of objects, and the environment, can bedepicted and practical measurements (such as di-mension, distance, angle, area, and so on) can beutilised productively to communicate and engineerour mental and physical environment. Spatial prob-lems are inherently situated in the world, which wemanage and solve within the confines of geometry.This is the strengthof geometry: that it states generallaws about geometrical objects and scenarios thatwe can then apply back to the real world. Simple

operands (i.e., reflection, rotation, subtraction andso forth) can be utilised in solving spatial problemsto manipulate parameters without questioning theissue of space. Such mathematical operations pro-vide a cognitive basis for ordering and manipulatingthe environment, and enable us tomanage everydaytasks. More importantly, they allow us to communi-cate past, current and future spatial scenarios. In a se-quence of lectures on 'The Relation of Space and Ge-ometry to Experience' Norbert Wiener claimed that"geometry is the scienceof a 'form' intowhichwecastour spatial experiences" (Weiner 1976, p95). Space,he argues, is experiential and that geometry is an ab-straction of that experience, being a set of rules bywhich experience may be replicated.

The spatiality of an organism is an effect of itsdistributed cognition. The concept introduced byHutchins (1995) is understood as "the ability of an or-ganism to interact with its environment for the pur-pose of satisfying its physiological (internal and ex-ternal) and social needs in order to survive and sus-

Generative Design - Biological - Volume 2 - eCAADe 33 | 581

tain itself" (Cardenas-Garcia 2013). The capacity ofan organism to affect its environment is a result ofits mobility and the effect of the environment onthe organism (relative to its objective and subjectiveneeds), instigating the organism to act in some par-ticular way. On the basis that design is a construc-tive activity (Glanville 2006) a distributed cognitionapproach to design is proposed, whereby an artifi-cial archetypal organism is utilised for the purpose ofconfiguring architectural arrangements. A novel ap-proach to arranging room layouts is presented; tak-ing a behavioural approach that builds on aspectsof spatial character and adopts distributed cognitionas a driver to generate layouts. The behaviour ofnatural phenomena is leveraged, such that their ac-tivities and goal seeking behaviour is bent towardsdesigning. Looking to distributed (swarm) systemsthe collective (social) behaviour of natural phenom-ena is utilised to capitalise on their constructive (i.e.nest building) and configurational (i.e. food foragingand agglomeration of slimemoulds) activities. Artifi-cial organisms are used to represent spatial regions,which self-organise according toassociationparame-ters. The resulting configurations are akin to bubble-diagrams and may thus serve as a basis from whichto develop actual architectural arrangements.

CONFIGURING ARCHITECTURAL LAYOUTSPlanning is the usual way of systematically work-ing through the arrangement of activities in a build-ing, and for all intents and purposes is the processthroughwhich anunderstanding of the buildingpro-gram is determined. Working out the organisationof a building is one of the most important and tax-ing aspects of the design process concernedwith thephysical arrangementof objects andareas to fulfil therequirements of the diverse human activities perti-nent to a particular building scenario. The successof the plan is in abstracting such problems into two-dimensions to define a plane to render them man-ageable, so that the numerous intertwined compo-nents may be arranged. Experience is an asset inplanning, but may be counter-productive as solu-

tions may remain hidden on the basis of 'what oneknows', or 'what has been done before'. Architects of-ten fall back on previous plans (as a template) for in-spiration, enabling one to judge and construct solu-tion's by interlacing the template with design criteriaof a particular scenario. The problem of organisingplan layouts is combinatorially hard and has receivedmuch attention in the fields of architecture and engi-neering, particularly since the computer came to beutilised as a tool for analysis and design.

The field of automatic plan generation is com-posed of two distinct approaches. (1) optimisation,which automates planning to present a single 'bestsolution' according to specified objective function(s),(aligned to the objective of shortest path inwiring di-agrams) and (2) enumeration, which presents possi-bilities to enable exploration of alternatives. The lat-ter presents a 'world of alternatives' to open up thedesigner to possibilities (Steadman 1970; Mitchell,Steadman and Liggett 1977; Flemming 1986, 1990).Another distinction is between methods that are in-teractive and engage the user. The LOOS modelby Ulrich Flemming (1986, 1990) generated 'looselypacked' arrangements of rectangles. This diagram-matic resolution maintained a level of ambiguitygenerating partial solutions; seen as a step in theprocess towards resolving arrangements, but not toproduce a final layout. Flemming focused on the in-termediate stage of space allocation to generate ar-rangements in which rectangles describing crucialspatial relations between the primary elements areallocated. Circulation spaces were not specified atthe outset, and so the process generated arrange-ments containing "gaps or holes that are used laterto allocate auxiliary spaces or that are added to pre-viously allocated spaces once the shape of the circu-lation area has been determined" (Flemming 1986,p192).

Methods to automate the generation of plannedlayouts diminished in the mid 80's, by those who pi-oneered them, because architects in practice wereturned off by the approach and due to three basiclogical difficulties:

582 | eCAADe 33 - Generative Design - Biological - Volume 2

1. The strengths of association between spacesweredefined inpracticeby surveyingpatternsofmovement in existingbuildingsof the sametype -whichmay of course be the result of theplans considered. Fundamentally the spatialrelations used to generate a plan were basedon existing plans so in essence what was gen-erated was a replication, or alternatives of thesame spatial arrangement.

2. Circulation tended to be privileged aboveall other generic functions. This is certainlytrue of the first group, whereby the result-ing layouts became optimised on this basis,which tended to produce centralised layouts,tightly clustered around the most stronglyconnected space(s).

3. Circulation spaces were specified at the out-set, which is the opposite to what tends tooccur in practice, where circulation is often aconsequenceof the roomarrangement. Note:There is an issue of scale here. Specifying cir-culation in a small house is not unreasonable;as only a central circulation space, such as ahall from which all spaces are accessible is re-quired. Else, in the case of two levels a stairwith landing on the next level is also required,from where, again all other spaces are acces-sible.

Recent and currentwork in thefieldof auto-matic plan generation:There has been a resurgence of interest in theautomatic planning of layouts using a variety ofcomputing techniques: such as shape grammars(Duarte 2003); evolutionary methods (Rosenman1997; Rosenman and Gero 1999; Jo and Gero 1998;Elezkurtaj and Franck 2002, 1999); physically basedmodelling (Arvin and House 1999, 2002, Arvin 2004);agents (Ophir 2009; Ireland 2010); three-dimensionalplanning and conceptual form generation (Hsu andKrawczyk 2003, 2004; Derix 2010; Ireland and De-rix 2003). These efforts have occurred both in anacademic and commercial context. The latter has

occurred as a result of architectural practice com-ing to terms with computation and large architec-tural practices employing computational design/re-search teams to operate at the edge of practice andacademia. This re-interest seems to have occurred,not least because architectural planning is a complexproblem and a key aspect of the design process, butbecause no model has yet met the requirements ofpractice, and new computational techniques providealternative approaches towards achieving this.

One focus of attention in modelling spatial ar-rangement that is of particular interest to this workis (what may be called) automatic space adjacencyanalysis, because it is focused on the pre-planningstage aimed at generating partial solutions. Archi-tects archetypally use space adjacency analysis (suchas bubble diagrams) as a way of considering the lay-out of functions in a floor plan, to explore relation-ships among the sizes, adjacencies and approximateshapes of spaces needed for various activities (White1986, Do and Gross 2001). A recent example of auto-matic space adjacency analysis is the tool developedby the computational design research team at Aedas(UK) Ltd. for the design of the Abu Dhabi EducationCouncil's new headquarters (Abrahams 2011). Theirmodel used an attract-and-repel algorithm to organ-ise the spatial adjacencies stipulated in the buildingbrief. (See also Arvin 2004).

Planning is an inherently top-down activity,which tends to focus on adjacency and connectiv-ity. There is an inclination to locate areas accordingto functional requirements with an emphasis on con-nectivity according to a scale of importance. Thosewhich have a high correlation are placed adjacentto one another, or as close as possible, while oth-ers are located at decreasing distances according tothe importance of their connectivity. "An adjacencyobjective is a topological objective that influencesthe distance between two spaces. For example, twospaces that have a large amount of traffic betweenthemmaybe specifiedwith an immediate adjacency"(Arvin and House 2002). Spatial relations, naturally,constitute much greater variance than the typical fo-


cus allows for. Whilst it is clear that topology is a fun-damental aspect of reasoning about space qualita-tively, it only accounts for particular distinctions. Inthe context of synthesising spatial relations we needto be able to examine a relation and account for con-ditions of both connection and parthood. Taking amereotopological approach allows a more encom-passing approach to establishing spatial relations be-tween constituents: (see Casati and Varzi 1999).

The approach of Hsu and Krawczyk (2003 and2004) is of particular interest to this study. They de-scribe rooms as having characters, making the anal-ogy between configurations of people and the taskof configuring rooms. "If we consider a space is aperson, and a group of people who gather accordingto similar requirements, we can then assume everyspacehas its own 'space character', either in a very ab-stract or practical sense, or both" (Hsu and Krawczyk2003). The relations between rooms, and to a site, arethus transformed into 'spatial characters' arrangedaccordingly in a space-planning program. Hsu andKrawczyk define space adjacency as behavioural, ap-plying certain traits of human activity to the param-eters of a space. For example, a room taking advan-tage of, or requiring, a view is perceived 'a watcher',requiring a location against the building perimeterto take advantage of a particular attraction. The ap-proach taken in this study is similar, but opens up thepotential of spatial arrangements in amanner reflect-ing pattern formation in natural systems; which arenot specifically constrained by topological relations.The model thus reflects a naturalised conception ofspace (Ireland 2015) and presents a bottom-up ap-proach to spatial configuration.

AN ARTIFICIAL LIFE APPROACHVarious organisms have developed the capacity tomodify their environment in such a way that theyconstruct artefacts. These structures embody thesubject's intelligence, and whilst human-beings maybe understood to create artefacts 'par excellence'their constructs are ingrained by patterns of inhabi-tation, which (from an evolutionary perspective)may

be extended downwards. Scrutinising built struc-tures enables us to consider 'lived-space' retrospec-tively as a system of social relations and to therebyextrapolate particular rules, or patterns, of inhabita-tion. Bill Hillier and Julienne Hanson (1984) trans-ported themselves within the plans of built formsto review their organisation, and illustrated how theconfiguration of space alterswhen specified from thediscrete perspective of each room location. Identi-fying architectural-space to be heterogeneous theyillustrate buildings to be social-systems determinedby the dynamics of habitation. Perceived in thisway architectural-space exhibits structure and con-stitutes organisation, becoming a sort ofmedium, es-tablished through a system of relations.

The model presented draws on the theory oforganism-environment relations by Barry Smith andAchille Varzi (2002). They define elemental formsof interaction between organism's, to reflect con-trasting sorts of interaction and how these affectthe niche of an organism: the niche of an organismequating to its territory. In short what is proposed isa general hypothesis for creating causally relevant spa-tial regions that generate spatial formation in a cell-likemanner; on the basis that the cell is the primal organ-ism. A cellmay be defined, in abstract, as a nichewiththe ability to distinguish self from non-self that actsaccording to differences in the environment, whichmean something and that this meaning has spatialconsequences according to the significance of thedifference relative to the state of the perceiving 'self'.The basic component of the model is an actant (aterm borrowed from Bruno Latour (1996) to refer toan autonomous entity-in-its-environment), which isan artificial cell-like organism that represents a regionof space. Actants coalesce with one another accord-ing to their relations to form an aggregation; whichis deemed to represent a pattern of habitation perti-nent to the spatial-regions represented.

Smith and Varzi identify four elemental formsof interaction: (a) coupling, (b) nonchalance, (c) en-counter (which may be a collision or impingement)and (d) contrast (which may be conflict or incompat-


Figure 1Scale ofconsolidation

ibility). Whilst (b) expresses commonality the otherexchanges lead to deformations of the organismsniche: the latter (c and d) in terms of negative de-formation and (a) to positive deformation. Theseforms of interaction are extended for this study to de-fine relation potentials, establishing forms of associ-ation between one actant and another. These asso-ciational parameters distinguish the spatial-propertyof a relation as a scale, not of dimension, but as agradient or degree of consolidation. The relation-potentials between one actant and another are thussub-sumptive. (See Figure 1).

The basic component of the computermodelAn actant represents a region of space, depicted bya boundary composed of 'boundary-nodes' that arelinked, and describe the actants form. The bound-ary is a mutable entity, because the boundary-nodeshave the capacity to affect andbeaffected. Their con-figuration therefore affects the actants conformation.The boundary consisting of nodes, which act as theactants receptors, and effectors, are affected by dif-ferences detected in the environment. These nodes(referred to as boundary-receptors) are agents whichmove collectively while emitting and responding todifferences. These differences are created by the ac-tants emitting pheromone, which acts as a signal

to other actants because each actants pheromoneis unique. As a difference , the pheromone can af-fect the actant by constituting a centrifugal or cen-tripetal force on the nodes, thereby affecting the ac-tants current location relative to the difference de-tected. Fundamentally an actant constitutes an ab-stract swarm, with the capacity to distinguish selffrom non-self. The pheromone an actant emits actsas a unique signal, identifying itself to other actants.Each actant thus has an identity, which its compo-nents share and to which other actants refer. A dif-ference may therefore be 'observed' by the actant,through its boundary-receptors as something whichis not an aspect of its identity. The actant will thusrespond to the difference detected by positioning it-self according to the significance: i.e., the associationwith the other actant detected. Consequently, theactants configure themselves according to those ac-tants they have an association (or dissociation) withby responding to their signals. Configuration arisesin the model as a result of boundary conformation,determined by the way the boundary-receptors re-spond to differences detected. (See figure 2).

The form of association and behaviour of an ac-tant is determined by its capacity to sense, and dis-tinguish differences present in its environment. Thedifferences thereby have meaning for an actant, act-ing as a signal, according to the significance of the


difference detected. The actants thereby respond toa difference according to the association. If the as-sociation is positive an actant will move towards thedifference (i.e. towards the source of production andtherebymoving in thedirectionof an associate). Oth-erwise the actantwill back off,moving away from thesource in a direction elsewhere from its dissociate.Figure 3 shows actants responding positively to dif-ferences they detect in their environment in variouscircumstances. In the first instance to a point source,secondly to a trail and thirdly to another actants sig-nal.

Figure 2An actant: the basiccomponent of themodel

The actants are building blocks to generate patternsof configuration autonomously, whose behaviourmay be steered and manipulated to suit particularobjectives, by affecting their associations. An actantmoves through the collective actionsof its boundary-nodes, whichmove relative to their distance from thenucleus and nearest boundary-node neighbour. Theformer is a simple attract-repel mechanism: if tooclose to the nucleus move away and if too far awayit move towards it. The latter a repel mechanismfrom the closest boundary-node of the same niche.This results in a wandering-like behaviour in whichthe collectivemoves in a unifiedmanner, reminiscentof the movement of amoebae. An actant wandersin this way for a period until, if no other actants aresensed, one of its boundary-receptors is selected tobecome a 'hunter'. Having been selected the hunt-ing boundary-receptor will move away from the nu-cleus, extending its search space to check for asso-ciate boundary-receptors beyond the niche's imme-diate vicinity. If another boundary-receptor is per-ceived the hunter will position itself according tothe relation between the two activity-niches: seeright-hand image of figure 3. Otherwise the hunter

switches state back to boundary-receptor and settlesback. This hunting action is analogous to the cellu-lar extensions of amoeboid type cells used inmovingand feeding. The propulsion of the extension can af-fect the course of the niche's wandering. If no asso-ciate is sensed after another period of wandering thehunting behaviour is repeated. The autonomy andsensorial capacity of an actant means that its formis changeable. It is a mutable figure affected by theconditions in which it is situated, which is diachronic:being affected by motion, the actants compositionand individual relations.

Pheromone ContingencyThe form of association and behaviour of an ac-tant is determined by its capacity to sense, anddistinguish differences present in its environment.An actant is equipped with the capacity to 'smell',which is enabled through the capacity to distinguishcontrasting forms and levels of pheromone. Thepheromone thereby has meaning for the actant, act-ing as a signal. The actants thereby respond to thepheromone according to the association, followingpheromone 'uphill' (towards the source of produc-tion and thereby moving in the direction of an as-sociate) if there is a positive relation and 'downhill'(moving away from the source in a direction else-where from a dissociate) if there is a negative rela-tion. Thismechanismdefines attract-repel behaviourreminiscent of predator-prey relations. With this inmind the emission of pheromone is determined bythe state of the boundary-nodes, such that they onlyemit pheromone when open to being found; imi-tating slime mould behaviour whereby the individ-ual spores only emit pheromone when they are ina state of hunger; causing them to aggregate. Like-wise, a boundary-node only emits pheromone whenit is not evading and seeking, thereby discouragingaggregation with dissociates. Since pheromone actsas a signal it would be perverse to remain signallingto a dissociate whilst evading, and to inform an asso-ciate (to whom one is a dissociate) of one's presence,thereby aiding evasion. To ensure an actant which is


Figure 3Stills from thecomputer modelshowing anactivity-nicheresponding topheromone. (Left)to a pheromonesource, (centre)following apheromone trail,and (right)engaging with anassociateactivity-niche.

dissociated to that which perceives it an associate isnot disadvantaged its sense of pheromone is moreacute: creating asymmetry in the actants capacity tosense pheromone. Referring to the hunter/prey con-dition we can see that an actant that is prey (it is dis-sociate to another for which it is associate) wants toevade. It therefore needs to be alert to any encroach-ment from dissociates, and so the threshold for sens-ing dissociate pheromone is therefore lower thanfor sensing an associate. Alternatively the hunterwants to be, as it were, 'quick off the mark'. Suc-cessful evasion/invasion is dependent upon the sit-uation determined by the trajectory of actants andthe presence (history) of pheromone. This meansthat the hunter/prey condition is opportunistic, be-cause whether the prey evades or the hunter attainsis a matter of directionality. The situation tends tobe better for the 'hunter' if the approach is from the'front'; which equates to the direction of movementbecause the pheromone, as a deposit, tends to forma trail.

A DISTRIBUTED COGNITION PERSPECTIVEOF CONFIGURATIONThe model presents a process whereby configura-tion is the result of the multiplicity of interactionsbetween the actants with their own timing, spacing,goals, means and ends. At a basic level the attract-repel mechanism (described above analogically aspredator-prey) relates to contrasting social relations,

defining conditions analogous to situations betweenareas which have conflicting social properties, suchas one being public the other private or environmen-tal properties where the effect of one is noise whilstthe requirement of another is quiet. In such casesone actant will seek to evade the other. However,the activities people perform, and the associationsbetween them are not necessarily fixed. Our activi-tiesmaybehabitual but they fluctuate dependingonphysiological and social needs. Whilst the model asit stands does not account for such 'fluctuating' con-ditions the configurations generated are the resultof the actants individual timing, spacing and goals.(See figure 4). The result of this is that, whilst the re-sulting configuration satisfies the individual actantsassociates the arising configuration is different eachtime, because history is a significant aspect of themodel. Also, the actants' 'behaviour' is tensive, be-cause an actant that has settled (having satisfied itsassociations)maybecomeunsettled by other actantsactions. This can cause the overall configuration tounravel, because if a settled actants associate is un-settled it is then caused to move; spoiling the settledactants state of harmony, causing them to re-seektheir state of cohesion. This is good, because the fi-nal configuration rests on the harmony of all actantsrealising their individual relation potentials. Config-uration in the model is aggregative.

The individual actants conformation and thusthe concluding configuration are determined by the


behaviour of the population. Looking back at themiddle image of figure 3, we see an actant respond-ing to a pheromone trail. The actant is in the processof adapting to its environment according to the dif-ferences it 'perceives'. Relating this to how people re-spond to their changing environment, and how ouractivities change or the way we alter our surround-ings to reflect (for example) a changing climate wecan see how the actants transformation reflects this:changing from one stable state to another. Beforedetecting the pheromone it is in one state, but hav-ing detected 'a difference' it responds to that differ-ence (and follows the pheromone trail) and settlesinto another state once that difference has been 'ac-commodated': i.e. it reconfigures its boundary con-formation to changing conditions. The model re-flects how an organism moves from a stable spa-tial domain, representing a given understanding of

said spatial domain at a specific time, to another sta-ble but different spatial domain due to the organismsensing changes in its environment and adapting tosuch changes. For example, in much the same waythat the internal state of an organism may changeaccording to external perturbations, an inhabitantsensing a changing climate may alter the configu-ration of his or her living quarters to accommodateor embrace changing external conditions: therebysatisfying physiological needs. The model is an ana-log of a bottom-up approach to architectural designthat takes into account that we as organisms inter-act with our ever present changing environment andredefine our spatial domain depending on our sen-sory interaction with said environment. The changesaffecting our sensory interaction are not only phys-iological, but are social too. For example, whilst wemight engage in seasonal changes of our living quar-

Figure 4Actants settled indifferentconfigurationsaccording to theirassociations


ters due to perceiving changes in our spatial domainconsequent to external conditions, changesmay alsooccur as a result of social or life events.

The relation between one actant and another iscausal, the significance of which is contingent on theconditions particular to each actant. Each actant de-posits a uniquepheromone, thereby identifying itselfto other actants. The actants thus configure them-selves according to those actants they have an asso-ciation (or dissociation) with by responding to theirpheromone. The point being that the 'form' of thepheromone relates to what meaning it holds for theindividual actants. The pheromone may representmany other factors other than whether one activityrelates to another: such as daylight, noise, a view,and so forth because the pheromone simply repre-sents a difference; which (if the pheromone repre-sents somethingof significance to an actant) is some-thing which either attracts or repels an actant.

CONCLUSIONWhat has been presented is a conceptual material-isation of spatial configuration, in a way reflectingthe behaviour of spatial formation found in naturalsystems: such as slime mould aggregation. By tak-ing such cases into account, the model here pre-sented can serve as a starting point for an artificiallife approach to generating architectural layouts. Themodel illustrates a novel user-centic method, tak-ing a behavioural approach that accounts for an or-ganisms sensorial engagementwith its environment,and demonstrates a general hypothesis for creatingcausally relevant spatial regions that generate spatialformation in a cell-like manner.

As it stands the model does not generate resultsan architect may utilise. The computer model needsfurther development and, as is, stands as a rhetor-ical device for the theoretical perspective and ap-proach argued for. There are two significant issues,which the results (figure 4) illustrate, that require at-tention. (1) the actants associations are purely sub-jective, which leads to (2) themodels practical imple-mentation. Theactants shouldhaveobjective, aswell

as subjective, parameters to satisfy, such as main-taining a min-max area. Of the relation-potentialspresented (figure 1) only weak overlap, coincidenceand disjunction are accounted for. The computermodel demonstrates a scale-of convergence but ex-tending it to allow for greater variance would enrichits output. Development of the model lies, initially,with resolving these two key aspects. Further devel-opment lies in extending the forms of pheromone,so as to incorporate site contingent factors such asviews, access and environmental constraints, so thatthe actants may respond to greater variances. Fur-ther enhancement of the actants sensorial capacitywould benefit the ability of actants to engage withdifferences present in their environment. A questionthen stands as to whether the computer model is tobe taken as a conceptual diagramming tool (alongthe lines of Flemmings (1990) LOOS programme) oris developed for more practical consideration. Theauthor suggests the former most beneficial, so thatthe model stands as an explorative aid rather than atool for solving planning problems. Developing themodel into 3dimensions would extend it beyond thestandard 2dimensional perspective adopted in plan-ning and serve to articulate Smith and Varzi's (2002)general theory of causally relevant spatial volumes.

THANKSThis paper stems from my PhD thesis (completed2013). I thank my supervisors Philip Steadman andPaul Coates for their input, encouragement and sup-port, and the EPSRC for funding. Thanks also to JaimeCardenas-Garcia for his on going discussion and in-put on distributed cognition and relevance to thecomputer model.

REFERENCESAbrahams, T 2011, 'Office Design by Tetris', WIRED (UK),

9th December 2011Arvin, S.A. 2004, Physically Based Mechanical Metaphors

in Architectural Space Planning, Ph.D. Thesis, TexasA&M University

Arvin, S.A. and House, D.H. 2002, 'Modelling Architec-tural Design Objectives in Physically Based Space


Planning', Automation in Construction, 11(2), p.213–225

Arvin, S.A. and House, D.H. Modeling Architectural De-sign Objectives in Physically Based Space Planning'Modeling Architectural Design Objectives in Phys-ically Based Space Planning', Proceedings of ACADIA1999, Salt Lake City, Utah, pp. 212-225

Bateson, G 2000, 'Form, Substance and Difference', in -, -(eds) 2000, Steps to an Ecology of Mind, The Univer-sity of Chicago Press, Chicago, pp. 454-471

Cardenas-Garcia, J.F. 2013, 'Distributed Cognition: AnEctoderm-Centric Perspective', Biosemiotics, 6(3),pp. 337-350

Casati, R and Varzi, A.C. 1999, Parts and Places: The Struc-tures of Spatial Representation, A Bradford Book, TheMIT Press, Cambridge, Mass

Derix, C. 2010 'Mediating Spatial Phenomena throughComputational Heuristics', Proceedings of ACADIA2010, New York, pp. 61-66

Do, E.Y. and Gross, M.D. 2001, 'Thinking with Diagramsin Architectural Design', Artificial Intelligence Review,15, pp. 135-149

Duarte, J. P. 2003 'A Discursive Grammar for Customiz-ing Mass Housing: The case of Siza´s houses atMalagueira', Proceedings of eCAADe 2003, Graz, pp.665-674

Elezkurtaj, T. and Franck, G. 1999 'Genetic Algorithms inSupport of Creative Architectural Design', Proceed-ings of eCAADe 1999, Liverpool , pp. 645-651

Elezkurtaj, T. and Franck, G. 2002, 'Algorithmic Supportof Creative Architectural Design', Umbau, 19, pp.129-137

Flemming, U. 1986, 'On the Representation and Gener-ation of Loosely Packed Arrangements of Rectan-gles', Environment and Planning B: Planning and de-sign, 13(2), pp. 189-205

Flemming, U. 1990, 'Knowledge Representation and Ac-quisition in the LOOS System', Building and environ-ment, 25(3), pp. 209-219

Glanville, R. 2006, 'Construction and Design', Construc-tivist Foundations, 1(3), p. 103–110

Hillier, B. and Hanson, J. 1984, The Social Logic of Space,Cambridge University Press

Hsu, Y. and Krawczyk, R.J. 2003 'NewGeneration of Com-puter Aided Design In Space Planning Methods - aSurveyandAProposal', ProceedingsofCAADRIA2003,Bangkok, pp. 101-116

Hsu, Y. and Krawczyk, R.J. 2004 'Space Adjacency Be-haviour in Space Planning', Proceedings of CAADRIA2004, Seoul

Hutchins, E. 1995, Cognition in the wild, MIT Press, Cam-bridge, Mass.

Ireland, T. 2010 'Stigmergic Planning', Proceedings ofACADIA 2010, New York, pp. 183-189

Ireland, T. and Derix, C. 2003 'The Poly-dimensionalityof Living: An Experiment in the Application of 3-Dimensional Self-Organising Maps to Evolve Form.',Proceedings of eCAADe 2003, Graz, pp. 449-455

Ireland, T. 2015, 'Naturalising Space', in Perrin, C. andMle-cek, M. (eds) 2015, Perception in Architecture: HEREand Now, Cambridge Scholars Publishing, pp. 38-46

Jo, J.H. and Gero, J.S. 1998, 'Space Layout Planning Us-ing an Evolutionary Approach', Artificial Intelligencein Engineering, 12(3), p. 149–162

Latour, B. 1996, 'On Actor-Network Theory. A Few Clarifi-cations PlusMore than a FewComplications', SozialeWelt, 47, pp. 369-381

Mitchell, W., Steadman, P. and Liggett, R. 1977, 'Syn-thesis and Optimization of Small Rectangular FloorPlans', Environment and Planning B: Planning and de-sign, 4(81-88), p. 1

Ophir, Y. 2009, 'Collective Intelligence: An Agent-BasedApproach to Programmatic Organization in Archi-tecture', International Journal of Architectural Com-puting, 7(3), pp. 479-499

Rosenman, M.A. 1997, 'The Generation of Form Us-ing an Evolutionary Approach', in Dasgupta, D. andMichalewicz, Z. (eds) 1997, Evolutionary Algorithmsin Engineering Applications, Springer-Verlag, pp. 69-85

Rosenman, M.A. and Gero, J.S. 1999, 'Evolving Designsby Generating Useful Complex Gene Structures', inBentley, P. (eds) 1999, Evolutionary Design by Com-puters, Morgan Kaufmann, San Francisco, pp. 345-364

Smith, B. and Varzi, A.C. 2002, 'Surrounding Space: TheOntology of Organism-Environment Relations', The-ory in Biosciences, 121, p. 139–162

Steadman, P. 1970, 'The Automatic Generation ofMinimum-standard House Plans', Land use and builtform studies, working paper 23(University of Cam-bridge Department of Architecture)

White, E.T. 1986, Space Adjacency Analysis: Diagram-ming Information for Architectural Design, Architec-ture Media.

Wiener, N. 1976, 'The Relation of Space and Geometry toExperience', in Masani, P. (eds) 1976, CollectedWorkswith commentaries. Volume 1, MIT Press, CambridgeMass., pp. 87-214


An Artificial Life Approach to Configuring Architectural Space

Documents

Transcript of An Artificial Life Approach to Configuring Architectural Space