03 swarm urbanism

In his book Emergence: The Connected Lives of Ants,Cities and Software, Steven Johnson presents the city as amanifestation of emergence.1 The city operates as adynamic, adaptive system, based on interactions withneighbours, informational feedback loops, patternrecognition and indirect control. ‘Like any emergentsystem,’ notes Johnson, ‘the city is a pattern in time.’2

Moreover, like any other population composed of a largenumber of smaller discrete elements, such as colonies ofants, flocks of birds, networks of neurons or even theglobal economy, it displays a bottom-up collectiveintelligence that is more sophisticated than the behaviourof its parts. In short, the city operates through a form of‘swarm intelligence’.

‘Emergence’ has become a highly popular term in recentarchitectural discourse, but it is worth recalling that the termitself does not necessarily refer to contemporary design issues.3

On the contrary, it could be argued that emergence could beviewed most clearly in traditional urban formations. For it isprecisely the less self-conscious forms of urban aggregationthat characterise the development of traditional settlements,from medieval villages to Chinese hutongs or Brazilian favelas,that fits best the simple rules of emergence, such as ‘ignoranceis useful’ or ‘pay attention to your neighbours’. These forms ofurbanism constitute a relatively homogeneous field ofoperations, where individual components do not stand out, butconform to the pervasive logic of their surrounding environment.In this sense, we might understand emergence as operatingwithin the framework of what Gianni Vattimo calls ‘weakthought’ (pensiero debole).4 This is not to say that the signaturebuildings of the contemporary city do not offer examples ofemergence. Rather, emergence is most recognisable in theproliferation of architectures of the everyday.5

Yet ‘emergence’ does have a highly contemporary relevance.Importantly, Johnson extends the principle of emergence to theoperations of certain software programs. If cities and softwareprograms display a similar emergent logic, how might we makeuse of digital technologies to model a city? Let us begin with anote of caution: the complexity of material computation withinthe city far exceeds anything that we might be able to model asyet through digital computation.6 Nonetheless, it would seemimportant to address this question, and explore the potential ofcomputational methodology for modelling urban form.

It is clear from the outset that whatever computationalmethodology is adopted it must itself follow the logic of swarmintelligence. In other words, it needs to exceed the capacities offractals, L-systems, cellular automata and other systems thatoperate largely within their own discrete internal logic. Fractalsand L-systems are limited for modelling patterns of growth inthat they are programmed to behave in a particular way, and ingeneral cannot adjust their behaviour in response to externalstimuli.7 Meanwhile, although cellular automata can respond totheir neighbours, they are fixed spatially, and therefore tied tocertain underlying grids. What we are looking for, then, is amulti-agent system comprised of intelligent agents interactingwith one another and capable of spatial mobility.

Standfirst to come. Neil Leach Standfirst to come. Neil Leach Standfirst tocome. Neil Leach Standfirst to come. Neil Leach Standfirst to come. NeilLeach Standfirst to come. Neil Leach Standfirst to come. Neil LeachStandfirst to come. Neil Leach Standfirst to come. Neil Leach Standfirst tocome. Neil Leach Standfirst to come. Neil Leach Standfirst to come. NeilLeach Standfirst to come. Neil Leach Standfirst to come. Neil Leach

Aerial view of Rocinha Favela, a large slum on the hills behindCopacabana in Rio de Janeiro, Brazil.

pheromone route agentseek pheromone range = 35seek angle range = +/- 15°max Force = 1.5max Velocity = 4.0wander = 1.65

in relation to stigmergic builder:seek pheromone range = 45



cluster agentseek pheromone range = 20seek angle range = +/- 25°cohesion = 3.5alignment = 2.4seperation = 1.65

in relation to pheromone agent:seek pheromone range = 25





network agentrest length = 20dampening = 0.15max Force = 5.5max Velocity = 4.0cohesion = 1.95


infrustructure agentrest length = 50dampening = 0.45max Force = 5.5max Velocity = 4.0cohesion = 6.15












cluste

r age

ntne

twor

k age

ntph

erom

one r

oute

agen

t

Kokkugia, Melbourne Docklands, Melbourne, Australia, 2008 previous spread: The decentralised structure of swarm, or multi-agent, systems changes the nature of hierarchy in urbanism. Hierarchies of scale andintensity are of course imperative to urbanism, however the swarm logic developed for the Melbourne Docklands flattens the hierarchy within thedesign process. All elements of the urban fabric are conceived of as possessing agency, enabling them to interact without a sequential design hierarchy;instead the hierarchy of intensities at a macro scale is an emergent outcome of their self-organising operation. As such, urban elements includinginfrastructure are not a priori, but rather one of many co-dependent systems that self-organise together to generate a mutually resilient organisation.

Kokkugia, Melbourne Docklands, Melbourne, Australia, 2008below: Agency operates through two main processes within this proposal: first by using design agents to self-organise urban matter, and second toencode intelligence into urban elements and topologies. In the first category agents operate to self-organise programme through a process of stigmergicgrowth. This type of collective behavior is similar to the logic of termite colonies in aggregating matter to form their mounds. The second category ofagents works in a similar manner to the processes that govern the self-organisation of slime mould cells into minimal path systems or the collectiveorganisation of ants to create bridges. These urban agents are primarily used to generate infrastructural and circulatory networks.

58

fatim

Highlight

fatim

Highlight

fatim

Highlight

fatim

Highlight

fatim

Highlight

fatim

Highlight

fatim

Highlight

fatim

Highlight

Swarm IntelligenceThere are a number of ways of modelling swarmintelligence within a computational framework. ManuelDeLanda outlines a model of agent-based behaviour thatcould be developed to understand the decision-makingprocesses within an actual city.8 These agents should beseen as concrete, singular individual agents, and not asabstract agents that embody the collective intelligence ofan entire society. DeLanda’s research to date is based oninstitutional organisations rather than urban forms of thecity, and while he envisages the possibility of a modelwhich uses a system of intelligent agents capable ofmaking their own decisions and of influencing others intheir decisions in order to generate urban form in someway, he has yet to develop this model.

The term ‘swarm urbanism’ has been used fairlyextensively within design circles. Often this refers to aform of ‘swarm effect’, where a grid is morphedparametrically using either digital tools or Frei Otto’s ‘wetgrid’ analogue technique.9 Such techniques, whileproducing interesting effects, are limited in that they areeither topologically fixed (as with a morphodynamic lattice)or base geometrically fixed (as with the wet grid), andcannot make qualitative shifts in form and space outsideof these set-ups. The advantage of a genuinely bottom-upemergent system of swarm intelligence where individualagents with embedded intelligence respond to one anotheris that it offers behavioural translations of topology andgeometry that can have radically varied outputs.10

One practice that does use swarm intelligence as afully bottom-up multi-agent design tool is Kokkugia, anetwork of young Australian architects operating fromNew York and London.11 They have deployed thistechnique at a macro level for a project in the Docklandsin Melbourne, an urban redevelopment currently underconstruction focusing on the extension of the CentralBusiness District into a disused port territory, and haveextended it to a micro level with the design of actualbuildings, as with their Taipei Performing Arts Centre.

With their swarm urbanism projects, the concern ofKokkugia is not to simulate actual populations (of people

or institutions) or their occupation of architecture, but to deviseprocesses operating at much greater levels of abstraction that involveseeding design intent into a set of autonomous design agents which arecapable of self-organising into emergent urban forms. They aretherefore not interested in mapping the motion of swarming agents togenerate an urban plan as a single optimal solution, but rather indeveloping a flexible system embodying a collective self-organisingurban intelligence: ‘An application of swarm logic to urbanism enablesa shift from notions of the master-plan to that of master-algorithm asan urban design tool. This shift changes the conception of urbandesign from a sequential set of decisions at reducing scales, to asimultaneous process in which a set of micro or local decisions interactto generate a complex urban system. Rather than designing an urbanplan that meets a finite set of criteria, urban imperatives areprogrammed into a set of agents which are able to self-organise.’12

This approach tends to produce a result which – if not reducible to asingle steady-state condition – will eventually coalesce into a near-equilibrium, semi-stable state always teetering on the brink ofdisequilibrium.13 This allows the system to remain responsive tochanging economic, political and social circumstances. Kokkugiatherefore sees the urban condition as one of constant flux: ‘Our urbandesign methodology does not seek to find a single optimum solutionbut rather a dynamically stable state that feeds off the instabilities ofthe relations that comprise it.’14

Rhizomatic UrbanismOne way of taking this approach further from a theoretical perspectiveis to appropriate the notion of the ‘rhizome’ from Gilles Deleuze andFélix Guattari as an urban planning strategy. In their seminal work, AThousand Plateaus, Deleuze and Guattari seem to offer a theoreticalmodel that resonates closely with the logic of emergence. For example,they refer extensively to multiplicities, to packs of wolves and to thelogic of the crowd.15 Meanwhile, one of the central tenets of theirphilosophy is ‘population thinking’ – the idea that ‘the population notthe individual is the matrix for the production of form.’16 Moreover,they touch upon the logic of the city itself as a space of flows. Deleuzeand Guattari describe the town/city as a network, a phenomenon oftransconsistency, that ‘exists only as a function of circulation, and ofcircuits’.17 For cities and towns themselves must be understood asamalgams of ‘processes’, as spaces of vectorial flows that ‘adjust’ todiffering inputs and impulses, like some self-regulating system.18

stigmergic builder agent

seek pheromone range = 200seek angle range = +/- 55°

positive feedback mechanismflocking coelesce to steady state


negative feedback: water edge avoidance







stigmergic builder agentseek angle range = +/- 55°





rgic builder agentngle range = +/- 55°




329220.21932098124.01942329412.25383948k 23412923905.234222

329220981212.253823412923905 2

pheromone route agent

bourke st

lt collin

lonsdale st

lonsdale st

lonsdale st










in relation toao stigmergic builder:seek pheromone range = 65wwer = wewiwz s8293 = 231093[]w23\saw e2 c02cw sa adae2323wegt232313wer = wewiwz s8293 = 231093[]w23\wer = wewiwz s8293 = 231093[]w23\wer = wewiwz s8293 = 231093[]w23\


in relation toao stigmergic builder:seek pheromone range = 65wwer = wewiwz s8293 = 231093[]w23\saw e2 c02cw sa adae2323wegt232313wer = wewiwz s8293 = 231093[]w23\wer = wewiwz s8293 = 231093[]w23\wer = wewiwz s8293 = 231093[]w23\






in relation to stigmergic builder:seek pheromone range = 65 pheromone route agent

seek pheromone range = 100seek angle range = +/- 25°max Force = 7.5max Velocity = 8.0wander = 0.65








max Force = 3.5max Velocity = 4.0wander = 0.85

in relation to pheromone route agent:avoid route pheromone = 85.5

















329220981212.253823412923905.2235094230887223792198733112891347761

329220981212.253823412923905.2235094230887223792198733112891347761

329220981212.253823412923905.2235094230887223792198733112891347761

329220981212.253823412923905.2235094230887223792198733112891347761

329220981212.253823412923905.2235094230887223792198733112891347761

329220981212.253823412923905.2235094230887223792198733112891347761

32

32

32

32 - 2S

32 - 2S

32 - 2S

32 - 2S

32 - 2S

32 - 2S

32 - 2S

32 - 2S

32

32

32

32

28

7G

98

98

98

98

C2

26

26

26

26

56

56

56

88

329220981212.253823412923905.2235094230887223792198733112891347761

329220981212.253823412923905.2235094230887223792198733112891347761

329220981212.253823412923905.2235094230887223792198733112891347761

329220981212.253823412923905.2235094230887223792198733112891347761

329220981212.253823412923905.2235094230887223792198733112891347761

329220981212.253823412923905.22350942308872

329220981212.253823412923905.22350942308872

329220981212.253823412923905.22350942308872

329220981212.253823412923905.22350942308872

329220981212.253823412923905.22350942308872

329220981212.253823412923905.22350942308872

329220981212.253823412923905.22350942308872

929812.25384129905.2509408872

329220981212.253823412923905.22350942308872

329220981212.253823412923905.22350942308872

329220981212.253823412923905.22350942308872

329220981212.253823412923905.22350942308872

329220981212.253823412923905.22350942308872

329220981212.253823412923905.22350942308872

329220981212.253823412923905.22350942308872

329220981212.253823412923905.22350942308872

329220981212.253823412923905.22350942308872

329220981212.253823412923905.22350942308872

329220981212.253823412923905.22350942308872

329220981212.253823412923905.22350942308872

329220981212.253823412923905.22350942308872

329220981212.253823412923905.2235094230887223792198733112891347761

329220981212.253823412923905.2235094230887223792198733112891347761

329220981212.253823412923905.2235094230887223792198733112891347761

329220981212.253823412923905.2235094230887223792198733112891347761

329220981212.253823412923905.2235094230887223792198733112891347761

329220981212.253823412923905.2235094230887223792198733112891347761

329220981212.253823412923905.2235094230887223792198733112891347761


329220.21932098124.01942329412.25383948k 23412923905.234222


329220.21932098124.01942329412.25383948k 23412923905.234222


329220.21932098124.01942329412.25383948k 23412923905.234222




Kokkugia, Melbourne Docklands, Melbourne, Australia, 2008 The Melbourne Docklands proposal is an investigation into an urban designmethodology based on the emergent capacities of swarm intelligence. Thisspeculative project posits a further intensification of the masterplan in a manner thattransforms its urban typology through the concept of urbanism as an ecosystem.

One practice that does use swarm intelligence as a fully bottom-up

multi-agent design tool is Kokkugia, a network of young

Australian architects operating from New York and London.

61

fatim

Highlight

fatim

Highlight

fatim

Highlight

fatim

Highlight

fatim

Highlight

fatim

Highlight

fatim

Highlight

fatim

Highlight

fatim

Highlight

fatim

Highlight

fatim

Highlight

fatim

Highlight

fatim

Highlight

fatim

Highlight

fatim

Highlight

fatim

Highlight

fatim

Highlight

Deleuze and Guattari use the interaction between awasp and an orchid to illustrate their concept of therhizome. The example is a familiar enough one – of aninsect being attracted to a plant, and thereby serving tocross-pollinate that plant.19 The wasp is of course being‘housed’ by the orchid, thereby giving the description acertain architectural relevance. But what interestsDeleuze and Guattari most of all is the interactionbetween wasp and orchid. The orchid has developedattributes that attract the wasp, but so too the wasp hasdeveloped a pattern of behaviour that serves the orchid.As Deleuze and Guattari observe, wasp and orchid enterinto a mutual reciprocity, such that the wasp has adaptedto the orchid, no less than the orchid has adapted to thewasp. Deleuze and Guattari refer to this as a form ofmutual ‘becoming’. The wasp becomes like the orchid,and the orchid becomes like the wasp, or – more precisely– the wasp has evolved in response to the orchid, just asthe orchid has evolved in response to the wasp.

Importantly, for Deleuze and Guattari, we mustperceive both wasp and orchid in terms of a multiplicity.They form an ‘assemblage’, an ‘acentred multiplicity thatis subjected to continuous movement and variation’.20 AsGreg Lynn explains: ‘The multiple orchids and wasps unifyto form a singular body. This propagating unity is not anenclosed whole, but a multiplicity: the wasps and orchidsare simultaneously one and many bodies. What isimportant is that there is not a pre-existing collectivebody that was displaced by this parasitic exchange ofsexual desire but rather a new stable body is composedfrom the intricate connections of these previouslydisparate bodies. Difference is in the service of a fusionalmultiplicity that produces new stable bodies throughincorporations that remain open to further influence byother external forces.’21

Deleuze and Guattari describe this process through theconcept of the rhizome: ‘Wasp and orchid, asheterogeneous elements, form a rhizome.’22 The logic ofthe rhizome should be distinguished from that of the tree.As John Marks explains: ‘The model of the tree is

hierarchical and centralised, whereas the rhizome is proliferating andserial, functioning by means of the principles of connection andheterogeneity … The rhizome is a multiplicity.’23 Central to theconcept of the rhizome is the principle of ‘becoming’, of forming arelationship with the other, as in the case of wasp and orchid, wherethe one de-territorialises the other: ‘The wisdom of plants: even whenthey have roots, there is always an outside where they form a rhizomewith something else — with the wind, an animal, human beings (andthere is also an aspect under which animals themselves form rhizomes,as do people, etc).’24 By extension, we could understand the city asforming a rhizome with its inhabitants.

This opens up an intriguing way of understanding the relationshipbetween humans as ‘agents’ within this system and the fabric of thecity as a form of exoskeleton to human operations. We need todistinguish between the city as a site of material composition – as anamalgam of traces of construction – and the city as the site of spatialpractices. The former can be read in terms of an accretion of materialdeposits, and the latter can be read in terms of choreographies ofagents whose freedom of movement is constrained by these materialdeposits. It as though the city is ‘formed’ by registering the impulses ofhuman occupation, much as the sheets on our beds, for example,record the movements of our bodies through the night. But so, too, thecity constrains the possibilities of human movement through its veryphysicality. There is, therefore, in Deleuzian terms, a form of reciprocalpresupposition between city and occupants. The city modifies itsoccupants, no less than the occupants modify the city. Over time thefabric of the city evolves through interaction with its inhabitants.

The task of design therefore would be to anticipate what would haveevolved over time from the interaction between inhabitants and city. Ifwe adopt the notion of ‘scenario planning’ that envisages the potentialchoreographies of use within a particular space in the city, we can seethat in effect the task of design is to ‘fast forward’ that process ofevolution, so that we envisage – in the ‘future perfect’ tense – the wayin which the fabric of the city would have evolved in response to theimpulses of human habitation. These impulses are likewiseconstrained and influenced by that fabric in a form of unendingfeedback loop between inhabitants and city, than echoes therelationship between wasp and orchid.

Quite how such a relationship could be modelled digitally remainsan interesting challenge for urban designers. 4

Notes

1. Steven Johnson, Emergence: The Connected Lives of Ants, Cities andSoftware, Schribner (New York), 2002. See also Mitchell Waldrop,Complexity: The Emerging Science at the Edge of Order and Chaos,Simon and Schuster (New York and London), 1992; John Holland,Emergence: From Chaos to Order, Oxford University Press (Oxford), 1998;Eric Bonabeau, Marco Dorigo and Guy Theraulaz, Swarm Intelligence:From Natural to Artificial Systems, Oxford University Press (New York andOxford), 1999; Neil Leach, ‘Swarm Tectonics’ in N Leach, D Turnbull and CWilliams (eds), Digital Tectonics, John Wiley & Sons (London), 2004, pp70–77.2. Johnson, op cit, p 104.3. See, for example, Michael Hensel, Achim Menges and MichaelWeinstock, AD Emergence: Morphogenetic Design Strategies,July/August 2004.4. See Neil Leach (ed), Rethinking Architecture: A Reader in CulturalTheory, Routledge (London), 1997, p 147.5. It could also be argued, for example, that certain avant-garde buildingsoperate collectively as a form of ‘movement’, even though they stand outmarkedly from their surroundings.6. For a discussion of ‘material computation’, see Neil Leach ‘DigitalMorphogenesis’, in AD Theoretical Meltdown, Jan/Feb 2009, p 35.7. One of the limitations of fractals is that they typically involve thesubdivision of an already known whole, while L-systems remain inherentlyhierarchical.8. See the interview with Manuel DeLanda in this issue, pp xx–xx.9. See, for example, Patrik Schumacher, ‘Parameticism: A New GlobalStyle for Architecture and Urban Design’, pp xx–xx of this issue.10. It could be argued, however, that the surface tension in the water of awet grid acts to self-organise the grid, and as such could be seen as abottom-up form of material computation.11. Kokkugia is a collaboration between Rob Stuart-Smith, Roland Snooksand Jonathan Podborsek.12. See the Kokkugia website: www.kokkugia.com, accessed on 3 March 2009.

13. They compare this to the self-regulating system of the earth, which operates within adynamically stable, yet fragile near-equilibrium condition, that Lynn Margulis has termed‘homeorhesis’. Kevin Kelly, Out of Control: The New Biology of Machines, Social Systemsand the Economic World, Perseus Books (New York), 1994, p 402.14. See the Kokkugia website: www.kokkugia.com.15. On this see Gilles Deleuze and Félix Guattari, A Thousand Plateaus, Athlone (London),1988, p 32 onwards.16. On this see Manuel DeLanda, ‘Deleuze and the use of the genetic algorithm inarchitecture’, in Neil Leach (ed), Designing for a Digital World, John Wiley & Sons(London), 2002, pp 117–118.17. Deleuze and Guattari, op cit, p 432.18. As John Holland puts it: ‘Like the standing wave in front of a rock in a fast-movingstream, a city is a pattern in time.’ John Holland, as quoted in Steven Johnson, op cit, p 27.19. Deleuze and Guattari appear to be referring to the digger wasp (Gorytes mystaceusand Gorytes campestris) and fly orchid (Ophrys insectifera). It is curious that they do notrefer to the particular sexual nature of this relationship. Usually an insect is attracted to aflower by the promise of nectar. Here, however, the sole attraction for the wasp is thepotential of copulation. The orchid looks and smells like a female wasp. It attracts themale wasp, whose excited behaviour serves to dislodge pollen from the plant on to theback of the wasp, which then transfers it to another orchid as it seeks gratificationelsewhere. Biologists refer to this process as one of ‘pseudocopulation’. See FriedrichBarth, Insects and Flowers, trans MA Biederman-Thorson, George Allen and Unwin(London), 1985, pp 185–192.20. Ansell Pearson, Germinal Life, Routledge (London), 1999, p 156. 21. Greg Lynn, Folds, Bodies and Blobs, La Lettre Volée (Brussels), 1999, p 139.22. Deleuze and Guattari, op cit, p 10. Deleuze and Guattari’s opposition to signification isan integral part of their theoretical position. Signification subscribes to the discourse of‘binary oppositions’. Moreover, it belongs to the realm of ‘representation’ rather than‘process’, and can therefore never account for the complexity of the rhizome.23. John Marks, Gilles Deleuze: Vitalism and Multiplicity, Pluto (London), 1998, p 45. 24. Deleuze and Guattari, op cit, p 11.

Text © 2009 John Wiley & Sons Ltd. Images: pp 56-7, 59-63 © Kokkugia LLP; p 58 ©Stephanie Maze/CORBIS

Kokkugia, Taipei Performing Arts Center, Taipei, China, 2008 The areas of the roof enclosing the auditoriums maintain theirexplicit starting geometry while the area surrounding the maincirculation spine has a more complex set of requirements andreforms to negotiate these. The agents are programmed with a set ofspatial imperatives while the material nature of the network createsa tendency towards equilibrium topologies that operate with adegree of structural efficiency. The network structure of the systemgenerates both space-filling lattices and continuous surfaces wherethe network connections are articulated as a web of veins.

In this proposal for a performing arts centre, the roof and spatial lattice aregenerated through a network of semi-autonomous agents. The emergentproperties of this swarm intelligence system generate an active networkedtopology in which agents self-organise in reforming their topology, enablinga gradient interaction between explicit design and emergent processes. Astarting network geometry of the roof is explicitly modelled and then self-organises within various degrees of freedom, enabling parts of the roof tomaintain their original geometry while other parts radically reform bothtopology and geometry. This process generates a material behaviourthrough the negotiation of the internal motivation of the agents and theforce within the network connections.

6362

fatim

Highlight

fatim

Highlight

fatim

Highlight

fatim

Highlight

fatim

Highlight

03 swarm urbanism

Documents