Scribd
Upload
666 views43 pages

An Evolutionary Architecture - John Frazer PDF

Uploaded by

Mînecan Ioan Alexandru
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
666 views43 pages

An Evolutionary Architecture - John Frazer PDF

Uploaded by

Mînecan Ioan Alexandru
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
You are on page 1/ 43
AN EVOLUTIONARY ARCHITECTURE John Frazer THEMES vit ARCHITECTURAL ASSOCIATION + LONDON Themes Vi An Evolutions Avehtectr hasbeen nen publised 6 corde vith an exon al at the Nester Assccston ram 27 nua to 25 Ferny 1905, The Ahtoctua Assertion woul ete hk the falowing compares nds andsnteiten or ‘heir conribuon: Aedes, Autograpnies, Bulg Cente Tus. Bho Cp Technologies Calheve ‘Gyscape tet Sences Ld, pi London is Roderick. Newton, RS Components icon Grohe Comurer Systm. Jer Sotho exition and tne pbcaon Feoon. have been made posse ty gat jonas tor th hits Comet of Egan Front Co: The Unveil Constr, wating model ot st _Bomnengimerocieonarmert AA bpore Unt 1 inGanneraton nb Urey ot User 1900, Proto Goatey Buckman cata of Bary Design eval ao sto ing vou enormer, Mant Rasa Peter ran ar So Frc 188 The fests uno he model which has Boon Goren tne ene to aves wih he ubtcaton of te book Temes Vi An Evotuionay Aetecturohas been produced fr AA Pubetons tough the AA Pt Stu by Demis Crompton techal coerrato and Paola Jomsion Greducon ext, with Jonny Meher ana Maryn Spr ‘The pubteaton and exhiton have been coord ‘breugh he fee ofthe Chonan, Alo Batu Exhibitions crgnzer Lind Brown ‘Salary manaper Andrew Macken ©1905 John Fazer and the Avec Aasoton, [6 Becton Square London WCB 36S '58N 1 670800477 Faint by Trg Lk Printed in London by E.. Bond id SECTION2 ‘THE NATURE OF THE EVOLUTIONARY MODEL PART |: THE GENERAL PROPOSITION ‘The evolutionary model requires an architectural concept to be described in a form of ‘genetic code’. This code is mutated and developed by ‘computer program into a series of models in response to ‘environment. The models are then evaluated in that environment and the ‘code of successful models used to reiterate the cycle until a particular tage of development is selected for prototyping in the real world. The real-world prototype is expected to be capable of interactive response to ‘the changing environment on a short time-scale, but ths is not essential inthe theoretical model In order to achieve the evolutionary modal itis necessary to define the following: a genetic code-script, rules forthe development of the code, ‘mapping of the code toa virtual model, the nature ofthe environment for the development of the model and, most important, the criteria for selection. {tis further recommended that the concept is process:driven; that is, by form-generating rules which consist not of components, but of processes. It is suggested that the system is hierarchical, with one process driving the next. Similarly, complex forms and technologi ‘should be evolved hierarchically from simple forms and technolog Generative Description In order to create a genetic description itis frst necessary to develop an architectural concept in a generic and universal form capable of being expressed in a variety of structures and spatial configurations in tesponse to different environments. Many architects already work in this ‘way, using a personal set of strategies which they adapt to particular design circumstances. Such strategies are often pronounced and Consistent to the point where projects by individual architects are instantly recognizable, Al that is required is that this generic approach i ‘explicit and sufficiently rigorous to be coded. ‘The process that we are describing has evident parallels with the way much conscious design occurs: itis also similar to the way in which See Bion Hota tere ith ohn Frazer ‘Lotus No 79,189, 9p 16-25, many vernacular archetypes and successful prototypes have been developed and adapted for diferent sites, environments and individual requirements. However we do not propose @ return to vernacular forms of building ‘evolution, for that tradition can no longer meet the requirements of ‘contemporary urban life. Nor can we advance the evolutionary process by constructing and evaluating full-size prototypes, as was the practice in ‘the past, wth the construction ofthe Gothie cathedrals, for example. This ‘would take too long and involve unacceptable costs, in terms of both ‘money and, in the case of structural failure, human life. ‘Computer Modeling and Simulation We suggest thatthe prototyping and feedback expressed in vernacular ‘architecture by actual construction should be replaced by computer modelling and simulation. At present, computer modelling tends to ‘occur after @ design is substantially completed, and only minor ‘modifications result. ts unusual to find the ‘what i type of model that is common in the field of economics. There are sever this. Inputing the data for a fully designed building ‘and expensive, and the modelling required for envi ‘not yet necessarily compatible with the modelling required for the ‘production of working drawings. Once the model has been loaded into the computer, there are only certain kinds of alteration which can easily 'be done. Despite the claims of the CAD salesmen, the truth is that itis ‘generally not easy to make changes, atleast not of the kind that would help to develop alternative strategies. One possibilty would be to have ‘modelling systems which enable some form of evaluation at very early ‘sketch’ stage. Unfortunately, despite substantial investment, theres still 2 lack of software capable of doing this, largely because of @ ‘misunderstanding about the function ofthe sketch or design doodle. In ‘addition, the architectural design ethos is unsympathetic to systematic comparative design evaluation and development. Designers tend to trust ‘their intuitive preconception and then modify it until it ‘works’ ~ with ‘works’ often being interpreted in an almost mystical rather than {unetional sense. It is ironic thatthe fixed ways of representing and abstracting building {orm which developed within the limitations of the drawing board and the techniques of projective geometry should have been carried over 80 Fr amor exe stern approach 10 fervour meeling ee 1H Pere. Paste Moseling "The Fee Modeling of he Lage of Sracure and Soace'in'T Maver tnd H Wag les CAAD Faw 87, [poowodegs ofthe Second neratonal Eenteronce on Corgute-Adna Boag Furs (Eever 1088 pp 190-208, Fer descigtion of these sysams S00 CF Eat ‘Shane Gommars ond tha Generation cf Desgrs in J Rooney and P'Stenaran ade, Preps of Computer ‘Aided Deszgn rvarfOpen Urvesty 186, WJ. Michel The Logi of Architecture — Design Computation and Coontion (ha Press 1800, Voletiedse,Detormae Rao do Favcaectre (ats 1888, Fora ul description fis technique See March, "h ookan Desorption of {Cais of Bult Farms in March od) ‘te aretectureo Form Combepe Uniersy Press 1970, directly into the computer. Geometrical forms could have remained plastic and fluid in the computer; instead they have become rigiified. ‘We need tofind an alternative to our drawing board obsession with fixed forms, and it seems that we have to think in terms of language ~ of a vocabulary or syntax. By this, we do not mean the kind of simplistic approach that is based on the use of large configurable elements and a related shape grammar or spatial syntax. We do not mean the endless permutational exercises much beloved by computer theorists who can produce, for example, every known Palladian plan (plus a few new ones) after devising a so-called Palladian Syntax, erative Adaptation We are proposing an alternative methodology whereby the model is ‘adapted iteratively in the computer in response to feedback from the evaluation. In order to make these significant changes to the database, it essential to have new forms of datastructure with a better understanding ofthe logical relationships inherent in the building model | first experimented with a special datastructure with the Reptile system Jn 1968, but this was specific tothe rules and geometry of one structural system. What we are now proposing is a technique applicable to a wide range of architectural concepts and geometries, all conceived as Generative systems susceptible to development and evolution, all possessing that quality characterized by Viollet-le-Duc as ‘style’: ‘the ‘manifestation of an idea! established upon a principle. CODING THE DATA ‘Once the concept has been described in terms of generative rules, the next stop is to code it in genetic terms. The idea of coding can be ilustrated by Lionel March's coding of plan schema and overall building massing. This is a more conventional and compact method than ours, ‘but of course March's intentions are also very different. For example, the plan of Le Corbusier's Maison Minimum is coded in hexadecimal representation (that is, counting to the base 16, -0,1,2.8,A.B,C,D.E,F). ‘So the plan is expressed as FF803F71180EFEO3GF. This can be expanded in binary, where each hex digit is translated to a 4-bit sequence (7 to 0111, € to 1110, etc). This is then grouped in 9-bit blocks ready to fold into a9 by 9 array. Each cel ofthis array then maps to a cel in a Venn diagram (a technique of Boolean algebra for expressing logical ‘operations diagrammatically). The diagram finally undergoes a metric transformation to become the correctly dimensioned plan. In @ similar manner, the overall three-dimensional form of Mies van der Rohe's Seagram Bullding can be coded #8 10083EFEOFO0. This may sound complicated, but the coding is very economical and some elements of this system are implicit in the invisible datastructures of any computer ‘modelling system. Evolutionary Techniques In our general model itis necessary to develop this coded version of the concept to address a higher level of complexity. This is done by applying genetic algorithms, classifiers or other evolutionary programming techniques, as described inthe previous section. The genetic code is bred into populations which are developed into abstract models suitable for evaluation in a simulated environment. The criteria for selection must be carefully considered and specified, The genetic code of the selected ‘models is then used to breed further populations in @ cyclical menner. Abstract models can be ext prototyping at any time. This externalization requires a further step of transformation or mapping nalized for further examination oF ‘Transforming the Output ‘The model used for evaluation is highly abstract and represents structure, spaces and surfaces in a manner which ean only be accessed by the environmental modelling program. Externalization of the model may include transformations which can affect its dimensions or form in ‘a more complex way than was necessary fo more significant requirement is that of materalzation into a buildable the schematic evaluation. A medium. This could be achieved by making part of the code simply represent a material and a precise construction technique, as we did with the model for Walter Segal described earlier. However we ar¢ inclined to think that this final transformation should be process-driven, and that fone should code not the form but rather precise instructions for the formative process. CODING: THE EXAMPLE OF THE REPTILE SYSTEM In 1966 | started developing a flexible enclosure system which consisted of just two folded-plate structural elements that could be combined to form a range of structures in a large numberof shapes and sizes, called itthe Reptile (reptile) system because it consisted of repeated units. The opti structural syst, computer Sramings Jona Fava 1988 onwerds. utomatcaly doveoped and aneuated the computer to produce corer sucha! ‘orm The denge ol ha suc oncosure ‘ios developed rom the aod wih 8 b> A Tio stctral uit. OG Ba Ma Se Ver BD, Se Has Aas Yeu Bag Tas Pie Fu Bas Tos Coded spt arertatons of mts opto, Architect! Design, Agi 1878 pp zai dinosaur metaphor also seemed apt for a component-based approach ‘that | anticipated would soon be obsolete. Already | was looking for a ‘more process-oriented and biological approach, and when the project was published in Architectural Design in 1974 | included a drawing of the structure being dismantled and replaced with a more biomorphie form. The accompanying caption said: ‘The association with an obsolete species is intended to emphasize that such @ component approach to architecture is probably only of transient significance.’ The frst of many computer programs associated with the project was written at the Architectural Association in 1967/68. This was able to produce perspective output of structures that | had designed ~ but only after the coordinates of each unit had been painstakingly digtized. | started to search fora technique of minimizing the data input, and thus conceived the ancestor of all subsequent developments. In 1968, with ‘access to the then massive computing power of the Atlas Titan in Cambridge University mathematics laboratory, | developed a seeding technique that allowed a densely coded description of a minimal configuration of the units to be developed and manipulated into a complex structural form without inputting any further data, ‘The two structural units of folded-plate construction could be orientated in eightoon different ways relative to each other, resulting in over three hhundred useful combinations. These combinations allowed the system to produce enclosures in a wide variety of plan shapes and complex structural forms. And because the units could be combined to form a straight edge allowing simple rectangular openings, the system was ‘entirely compatible with traditional rectangular buildings. There was no ‘need for the kind of special variants or cut units which had limited ‘geodesic structures, for example, to dome-lke shapes. ‘2 secron2 ARABS an &® & & 2& tN Lo pulley wp and out atasconing APY Aid, — Addy yd moves Fea fut dasciotion se J, Fazr ond ‘SM Conn, Concept! Seong {ecru or rcsectrl Design PAC 72, Procaotng ofthe tron Confrence ‘rine Appcaton of Computers in ‘rchtectirl Desa (rine Conerences ‘sin AM 109), Pp. 425-94 ‘The machine code programming wos done ‘tha Cambridge Unerety matherate! ‘Shorey by Reha Panne eth e, reoulgoment of Cratos Lang {The FORTRAN programs were writen Lylota Faget an rosawen set Frndaco Guar, The units were located by 60" axes using integer coordinates (A,B,C). 'B' coordinates had a negative direction, not positive as one would expect, in order to ‘make ‘C’ coordinates (which were not stored) easy to find (AvB+C=0), and to provide a check when inputting data manually ‘0’ coordinates described depth or level ‘of the unit inthe structure. The type and orientation of ‘ unit was described by two digits, the first giving the ‘type and vertical orientation, the second the ctientation in the horizontal plane (T'T"). This whole description of the unit in space took the form (AB,C.0,17T"), for example -66,92,34,21, and was originally packed (using machine-code instructions) into one word of the Cambridge Atlas Titan with a pointer to the next unit in the chain. The units were ‘chained clockwise by levels from the top of a structure ‘downwards, with information about the levels being stored at the beginning of data refering to that level ‘The basic datastructure was set up and manipulated by @ series of machine-code sub-routines and functions. These allowed the descriptions ofthe units to be retrieved, deleted or updated, and additional ‘units to be inserted into the chain. On output the integer location on the 60° grid was translated into ‘normal orthogonal coordinates by a sub-routine which Contained a graphic description ofthe unit its type and orientation. A variety of descriptions were available depending on the detail required in the final drawing, ‘The GINO graphics package was used to produce perspectives and other drawings as required. ‘The seed was a minimal closed configuration which included units in all possible orientations but not necessarily all possible combinations. The develop- mont of a structure was initialized by implanting the

You might also like