LearningfromGenerativeDesignSysteminthe60’spapers.cumincad.org/data/works/att/ecaade2018_169.pdf · Open Design, however, produces many varia-tions, none of which is original or

Learning fromGenerative Design System in the 60’s

Case Study of Agricultural City Project by Kisho Kurokawa

Maki Kasahara1, Kiwa Matsushita2, Akihiro Mizutani31Tokyo Denki University 2Shibaura Institute of Technology 3Toyohashi Universityof [email protected] [email protected] [email protected]

The concept of generative design in Architecture and Urbanism can be found inthe 60's before the wide availability of computer technology. This paper decodesone of the urban projects by Metabolist in 1960, which was intended to be agenerative system applicable to other sites and evolves over time. Through ouranalysis, we de-code the formulation process, and verified our hypothesis byre-coding into the program using the software, Rhinoceros and Grasshopper. Wefound that the determinate factors rule more at the macro level of the project, butthe parameters are set by taking the local conditions into account. At the microlevel, the system leaves more freedom to accommodate various needs, reflectingthe philosophy of the Metabolists. The investigation on this historical predecessorcan provide useful insights for parameter settings in future generative systemdesign.

Keywords: Generative Design, Grasshopper, Kisho Kurokawa

INTRODUCTIONThe emergence of Open Design is dramatically de-mocratizing the concept of design and its influencehas spread into the architectural profession. Archi-tecture had been an authorial discipline; it is ex-pected that all dimensions are pre-calculated, as thearchitect envisions, on construction drawings, and abuilding is to be built exactly as depicted in thosedrawings (Carpo 2011). Thus, the final outcome ofthe building is determined and fixed in the architect’sdrawings prior to construction.

Open Design, however, produces many varia-tions, none of which is original or final, accordingto the users’ preferences or situations. In Open De-

sign, the design is a generative system, and the roleof the designer is shifting towards a developer of sys-tems that enable anybody to adopt and create de-sign, rather than as a creator of a singular perfect ob-ject (Atkinson 2011). The generative system becameeasily accessible with the development of computa-tional technology. It is an especially useful asset tobe able to construct a simulation of a complex entitylike a city, which requires theprocessingof numerousvariants, including the time-module as the fourth di-mension of the space (Weber 2009).

The root for generative design system can betraced back to late 1950s and early 1960s, whena new generation of avant-garde architects, such

GENERATIVE DESIGN - Volume 2 - eCAADe 36 | 95

as Team 10 and Japanese Metabolists energeti-cally launched new ideas of architecture and ur-banism (Smithson 1974). Facing the rapid popula-tion increase, they advocated ideas such as mobil-ity, expendability, indeterminacy, and openness forchange and growth, which, for them, transcendedthe more static philosophy of older Modernist archi-tects representedbyC.I.A.M. Itwas a timewhen somearchitects were aware of the potential of computa-tional technology to be a strong generative designtool especially for urban design in the future, eventhough, at that point, such technology was not read-ily accessible. Kenzo Tange (1913-2005) was one ofthose who was strongly interested in new computa-tional technology and systematic urban design. Inhis “Tokyo Plan 1960,” three different parts (Residen-tial Complex, Office and Traffic) were separately butsystematically designed based on the same rules andmodules by three different teams in his laboratoryat Tokyo University (Mizutani 2013). The system wasbased on many determinate factors, such as popula-tion, transportation flow and a set of urbanmodules,which enabled the smooth collaboration of multipledesigners.

As a developer of the generative design sys-tem, an important task for an architect / designerwould be to define parameters. Even though theymay be based on determinant factors, priorities andrules have to be set, and different sets of parame-ters would result in different outcomes, not just vari-ants. Nonetheless, in the design of a complex en-tity like a city, the design system may not be able togenerate all elements based on determinant factors,but should allow some room for subjective and in-determinate inputs in order to accommodate diverseneeds.

This paper treats “Agricultural City Project(1960)” by Kisho Kurokawa (1934-2007), as a prede-cessor of generative design. We analyze the projectas an architecture of system and decode it. Then, weverify the hypothetical codes by writing a programto regenerate the model using the computer soft-ware, Grasshopper for Rhino3D, and investigate on

its system, that which the original agents could notachieve to a full extent due to the limitation of tech-nical availability at their time. We also aim to clarifythe balance between the determinate and indeter-minate factors in its design by differentiating whatcan be automatically generated by the program.

THE SELECTION OF THE SUBJECTWe chose “Agricultural City Project (1960)” by KishoKurokawa, whowas one of the foundingmembers ofMetabolism, as the subject of the study. Kurokawaadvocated a “Master System” for city planning, whichhe defined as a “Four-dimensional Master Plan,”taking Time Module into consideration (Kurokawa1967). Kurokawa’s Agricultural City Project, first pre-sented in the Metabolist booklet issued for the 1960World Design Conference held in Tokyo, is an un-realized, visionary project, which proposed agricul-tural settlements on an expandable and griddedstreet network aswell as artificial ground, raised four-meters above ground level to avoid damage fromriver flooding. It was designed as a reaction toKurokawa’s own experience of the Ise-bay Typhoonin 1959 and its damage to actual agricultural set-tlements. The project gained international acclaimupon presentation at the World Deign Conference,which led it tobeexhibited at the “VisionaryArchitec-ture” exhibition at the Museum of Modern Arts, NewYork.

Although the computer was not used for thedesign, as a member of the design team of “TokyoProject 1960,” Kurokawa was aware of this upcom-ing technology and was interested in making archi-tectural and urban forms throughwriting codes, withwhich his mentor, Kenzo Tange, as well as othercontemporary artists and designers were startingto experiment (Casey 2010). The key concept ofMetabolism, the responsiveness to the environmentand reproducibility in the cities of growing popu-lation, suited well with algorithmic design, and theAgricultural City Project was designed as the MasterSystem.

Kurokawa, on the other hand, had a strong per-

96 | eCAADe 36 - GENERATIVE DESIGN - Volume 2

sonal belief in aesthetics, which are also apparent. Inthis project, he acted as the ”parameter-setter (Mas-ter Programmer) as well as the designer. His dualrolesmake it an interesting example to study the bal-ance between determinate and indeterminate fac-tors in architectural design.

Figure 1Kisho Kurokawa:Agricultural CityProject 1960 (thefirst version:Prototype)

Figure 2Kisho Kurokawa:Agricultural CityProject 1960 (thesecond version)

THE ANALYSIS OF THE PROJECTThe Project consists of two versions: the prototypeand its variation onto the actual site. The prototypehas a 5 by 5 grid, of which each squaremeasures 100-metersby100-meters, and theartificial raisedgroundwas placed in various patterns in each square (Figure1). This 500-meter square prototype was planned tohouse an average size of an existing community, ap-proximately 340 households of 1700 people as wellas public buildings, such as Shinto shrines, schools,and community centers on the artificial ground. Thegrid acts as the streets as well as the supply lines forwater and electricity.

In the second version placed on the actual site,the edges of the grid stretch in all directions and theformation of each cluster is not a clean square, im-plying the possibility for the future expansion (Figure2). We made a hypothesis that some parts of expan-sion system of the inter-depending communities areplanned based on logical design rules as a variationof the prototype, since Kurokawa did not intend thisdesign specifically to this site, but to be applicable towherever a danger of flood from rivers exists.

Thewritten description of the project claims thatthe site is in Ama county, Kanie town in Aichi prefec-ture, which was Kurokawa’s hometown. By juxtapos-ing the site plan onto the map, however, we foundthe site of the project was situated in Kuwana city,Tado town in Mie prefecture, which was located 20kilometers west from Ama county. (We also foundout that in the second version, the scale of a clus-ter was reduced by 60% in order to fit the site con-dition. For clarity, however, we assumed the dimen-sionwas 100% for this analysis.) The strong feature ofthe selected site is the shape of the river meanderingand that it leaves a roughly circular island in themid-dle. The region that this project was conceived forhad a history of flood damage, and one of the pro-tectionmeasures that villagers took was to constructa ring-shaped embankment (Wa-ju) and tomake set-tlements inside of it. Thus, Kurokawamayhave inten-tionally selected this site, which naturally had aWa-julike feature, rather than his hometown. This analysis


led us to make an assumption that the shape of theriver is an important factor in determining the posi-tioning of the project.

Figure 3Types of ArtificialGrounds

We identified the cluster surrounded by theforked river as the central and initial community,since it is themost complete and similar to the proto-type. The center of the cluster roughly matches withthe center of the Wa-ju like ring, which we assume tobe the “startingpoint”. It is the farthest point from theriver and thus safest from flood damage.

We also discovered by analysis that the horizon-tal grid lines are parallel to the part of the river whichhas the least inflection. We defined the line of riverwhich is almost straight a “reference line” for the grid.It also coincides with the angle of existing rice paddyfields, which can be considered an inherent trace ofthe local spatial characteristics. The second cluster,which is southof thefirst one seems tobeamirror im-age of the first on the other side of the reference line.From those two clusters, the other clusters seem tobe placed at least one module apart and shifted oneto three grids from the adjacent one in order to avoidinterference from the other clusters and to adjust tothe different conditions of the actual site. The in-side of some grids have artificial ground, onto whichhouses and other buildings were to be built. Thereare six types for artificial ground (Figure 3), but wecouldnot find the apparent rules for their placementsand frequency. The bridges (extension of axis) that

connect different clusters are always in pair, implyingone-way traffic on each bridge, and they seem to ap-pear where the density of settlements is high, but wecould not find logical rules for their placement.

While we could find certain logical rules for thepositioning of the starting point, the expansion ofthe grid from the “reference line,” and the formationand location of clusters, rules for the types and place-ments of artificial ground as well as bridges that con-nect different clusters does not seem to depend onthe determinant factors.

VERIFICATIONOF THE ANALYSISThe hypothesis of the design process based on theanalysis above was verified by writing an algorithmicprogram to replicate the project using Grasshopper.The process is as follows (Figure 4 & 5):

Step1a : The“ReferenceLine” fromthe riverThe “reference line” which is the basis of the grid isfound from the river at the least curved line.

Step 1b: Positioning of “Starting Point”The “starting point” for the centre of the initial com-munity is defined as the farthest point from the riverin the Wa-ju like circle.

Step 2: Formation of GridThe grid of 100-meters by 100-meters is drawn fromthe “reference line.”

Step 3: Positioning of ClustersThe initial community is formed on the grid and hasthe starting point as its centre. The second clusteris formed as a reflection of the first one on the otherside of the “reference line.” The shifting direction ofthe other clusters from the first two are deliberatelydetermined, but the numbers of modules betweenthe clusters are randomly distributed.

Step 4: Placement of Artificial GroundAlthough all the clusters seem to be based on theprototype, none of them are a complete 500-meter


Figure 4Program Flow


Figure 5Program Flow 2


square grid, having voids or extending some gridsout of the 500-meter square. The sizes of voids are2- or 3-grids wide or long, and the locations seem tobe slightly off-centre. In the program, the shifting di-rection of the centre of each cluster is randomly se-lected. Also, the extensions of the grids are randomlyregenerated. While we specified six different types ofartificial grounds in this project, we could not find ap-parent rules for their distribution. Thus, in thismodel,the locations of different types of artificial groundsare decided by the roulette, at the ratio they appearin Kurokawa’s original version (Figure 6).

Step 5: BridgesA pair of bridges are placedwhere the densities of ar-tificial grounds are highest in adjacent two clusters.

DISCUSSIONComparing the reproduction by Grasshopper (Figure7) with Kurokawa’s original version, we found the fol-lowing three layers of design systems in the Agricul-tural City Project:

At a more macro level, the positioning of the en-tire project, the main frames of structure and the di-rection of expansion follow logical rules, responsiveto the site and natural conditions, such as the riverand topography. The set of parameters at this levelcan easily be applied to different sites to create an-other version of Agricultural City.

At the second level, more social factors rule theformation of clusters. For instance, the typical gridand cluster sizes are determined based on the aver-age size of an existing community. The rules for dis-tance between clusters as well as the placement ofconnecting bridges may also be included in this cat-egory. The program could reproduce a similar designto the original at this level. The clear difference of thereproduction from the original at this level, however,is that the clusters in the reproduction aremore com-plete and similar to the prototype. We could considerthat Kurokawa’s original version is representing onemoment in the process of development, while the re-production shows a more complete stage. The pa-

rameters for this evolution process are not yet appar-ent, butpossiblydependonfluctuating social factors,such as migration.

Figure 6Step 4 : Placementof Artificial Groundin the originalversion

Figure 7RegeneratedAgricultural CityProject


At a more micro level, the types and locations of ar-tificial grounds as well as the location of individualhouses, however, could not be replicated automati-cally. In the regeneration, we could not exactly repli-cate the original form without entering specific val-ues to components. The overall image of the project,however, is not affected by the differences. It sug-gests that their formation is also not based on appar-ent rules, but on other kinds of input possibly fromresidents and community. It may reflect Kurokawa’shumanistic approach to the city that it should alwaysbe a system, of which self-directed individuals havecontrol (Kurokawa 1959).

CONCLUSIONWe analyzed the Agricultural City Project by KisyoKurokawa as one of the predecessors for early gen-erative design and verify their system by re-codingusing Grasshopper. Though the study, we made thefollowing observations:

The indeterminant factors (the parts that couldnot be regenerated automatically) appeared at alllevels of the project, but this tendency becomesmore apparent at a more micro scale leaving morefreedom for each settlement’s or resident’s prefer-ences and choices. This idea corresponds to the phi-losophy of the Metabolists. The algorithmic proce-dure clarified the two layers of urban systems: theone that regulatesmore permanent urban infrastruc-ture and the other that generates ever-changing ur-ban elements and functions.

Themaster system in this project is not designedto be an autonomous and universal system, but theparameterswere setby taking the local conditionandtradition (i.e. Wa-ju) into account. It can be a use-ful reference when we design system architecture ingenerative design.

In the present work, we could not find the ap-parent rules to regenerate elements at a more microlevel. We speculate that other systemsare required toprovide inputs for regeneration of the project, suchas simulation through a multi-agent or deep learn-ing system. Those social and phenomenological in-

puts would direct the process of evolution over timeand clarify different layers of parameters that affectthe urban structure, which is only possible with com-puter technology. Further development of researchwould investigate the algorithmic systems for morecommunity and humanistic factors which Kurokawaenvisioned but were not technically possible to fullydevelop at that time.

REFERENCESAtkinson, P. 2011, ’Orchestral Manoeuvres in Design’, in

vanAbel, B., Klaassen, R., Evers, L. and Troxler, P. (eds)2011, Open Design Now, BIS Publishers

Carpo, M. 2011, The Alphabet and the Algorithm, MITPress, Cambridge, USA

Casey, R. and Chandler, M. 2010, Form+Code in Design,Art and Architecture, Princeton Architectural Press,Princeton

Kurokawa, K. 1959, ’Shin Tokyo Keikaku (New TokyoPlan)’, Kenchiku Bunka, pp. 27-31

Kurokawa, K. 1967, ’Teiju Tan-i Keikaku no Rinen toHouhou (The Concept and Method for PermanentUnit Plan)’, Kenchiku Bunka, pp. 93-125

Mizutani, A. 2013, Reevaluation and application of urbandesign method of Kenzo Tange’s laboratory -Focusingon the contemporary significance of themethod usinga computer-, Ph.D. Thesis, Shibaura Institute of Tech-nology

Mizutani, A. 2016, ’RE-EVALUATIONOFCOMPUTATIONALDESIGNOFTANGELABAND ITS IMPORTANCEOFTO-DAY:Focusingon computer simulationmethods andits theory’, Journal of Environmental Engineering, Ar-chitectural Institute of Japan, 81(723), pp. 487-494

Smithson, A. 1974, TEAM 10 PRIMER, MIT Press, Cam-bridge, USA

Weber, B., Muller, P., Wonka, P. and Gross, M. 2009, ’Inter-active geometric simulation of 4D cities’, Eurograph-ics, 28, pp. 481-492


Documents

LearningfromGenerativeDesignSysteminthe60’spapers.cumincad.org/data/works/att/ecaade2018_169.pdf · Open Design, however, produces many varia-tions, none of which is original or