İ
STANBUL TECHNICAL UNIVERSITY
INSTITUTE OF SCIENCE AND TECHNOLOGY
M.Sc. Thesis by
Mahmut Çağdaş DURMAZOĞLU, Arch.
Department
:Informatics
Programme:
Architectural Design Computing
JUNE 2008
AN AGENT-BASED SYSTEM TOOL TO SUPPORT
DESIGNERS FOR FREE-FORM SHAPE
İ
STANBUL TECHNICAL UNIVERSITY
INSTITUTE OF SCIENCE AND TECHNOLOGY
M.Sc. Thesis by
Mahmut Çağdaş DURMAZOĞLU, Arch.
(710051005)Date of submission
:05 May 2008
Date of defence examination:
09 June 2008
Supervisor (Chairman): Prof. Dr. Gülen ÇAĞDAŞ
Dr. Hakan TONG
Members of the Examining Committee Prof.Dr. Ahsen ÖZSOY (İ.T.Ü.)
Prof.Dr. Orhan HACIHASANOĞLU (İ.TÜ.)
Assist.Prof.Dr. Yüksel DEMİR(İ.T.Ü.)
JUNE 2008
AN AGENT-BASED SYSTEM TOOL TO SUPPORT
DESIGNERS FOR FREE-FORM SHAPE
İ
STANBUL TEKNİK ÜNİVERSİTESİ
FEN BİLİMLERİ ENSTİTÜSÜ
SERBEST BİÇİMLİ MİMARİ TASARIM ALANINDA BİÇİM DEĞERLENDİRMESİNDE KULLANILACAK
ETMEN-TABANLI BİR MODEL: DROP
YÜKSEK LİSANS TEZİ
Mim. Mahmut Çağdaş DURMAZOĞLU
(710051005)
HAZİRAN 2008
Tezin Enstitüye Verildiği Tarih : 05 Mayıs 2008
Tezin Savunulduğu Tarih : 09 Haziran 2008
Tez Danışmanı :
Prof.Dr. Gülen ÇAĞDAŞ
Dr. Hakan TONG
Diğer Jüri Üyeleri
Prof.Dr. Ahsen ÖZSOY (İ.T.Ü.)
Prof.Dr. Orhan HACIHASANOĞLU (İ.T.Ü.)
Y. Doç.Dr. Yüksel DEMİR (İ.T.Ü.)
iii PREFACE
I would like to express my heartfelt thanks to both of my thesis supervisors, Prof. Dr. Gulen CAGDAS and Dr. Hakan TONG for accepting to start working on this subject with me and for their continuous support, guidance and challenging insights. I also would like to acknowledge ir. Andrew Borgart and Dr.ir. Rudi M.F. Stouffs from Delft University of Technology, I am deeply indebted to Mr. Borgart and Mr. Stouffs for their generous support and guidance and for being my mentors during the period when I was in Delft University of Technology as an exchange student.
For their unconditional love, support and encouragement, I would like to thank my parents Meliha and Salih DURMAZOGLU. I would also like to thank my friends for their support. And my deepest thanks to Sehnaz CENANI, for all of her support and for being there.
iv CONTENTS
ABBREVIATIONS vi
LIST OF FIGURES vii
ÖZET x
SUMMARY xii
1. INTRODUCTION 1
1.1 Free-Form Design 3
1.2 Opinions about Free-Form Architecture in Academic World and Practice 3
1.3 Problem Definition 5
1.4 Research Objectives 5
2. AGENT DEFINITION and AGENT SYSTEMS 7
2.1 Definition of Agent Systems 10
2.1.1 Components of the Agent Models 12
2.1.2 Behavioural Complexity 14
2.1.3 Implementation Complexity 15
2.1.4 Domain dependence 15
2.1.5 Social ability 16
2.2 Agent models 16
2.2.1 Swarm Agent Model 17
2.2.2 Cognitive agent Model 20
2.2.3 Motivated agent Model 22
2.3 Review of Agent Based Systems Used To Assist Design 26
2.4 Agent-Based System Examples 27
2.4.1 A Virtual Reality Design Environment with Intelligent Objects and
Autonomous Agents 27
2.4.2 Agent Generated Architecture 37
3. FREE FORM DESIGN 43
3.1 The Course of Free-Form Design throughout the History 43
3.2 Symbiotic Relationship with Computers 55
3.3 Free-form Bodies in Digital Architecture and Industrial Design 58 4. AN AGENT-BASED SYSTEM TOOL TO SUPPORT DESIGNERS
FOR FREE-FORM SHAPE EVALUATION: DROP 73
4.1 Rain-Flow Analysis as a Method to Analyse the Distribution of
the Forces on Free-Form Bodies 73
4.2 Theoretical Model of the DROP 76
4.2.1 Implementation Environment 77
4.2.2 Phases of the Model 79
v
4.3 Application of the DROP 83
4.3.1 Kresge Auditorium 91
4.3.2 Los Manantiales Restaurant 93
4.4 Agent Behaviour in the Model 97
5. CONCLUSION 100
5.1 Main Findings 101
5.2 Main Contributions of the Thesis 102
5.3 Potential Users of the Implemented Model and Application Domain 103
5.4 Future Work 104
REFERENCES 106
vi ABBREVIATIONS
2D : Two Dimensional
3D : Three Dimensional
ACADIA : Association for Computer Aided Design in Architecture AI : Artificial Intelligence
CAAD : Computer Aided Architectural Design CAD : Computer Aided Design
CAE : Computer Aided Engineering CAM : Computer Aided Manufacturing
CATIA : Computer Aided Three Dimensional Interactive Application CNC : Computer Numerical Control
CPU : Central Processing Unit
CSCD : Computer Supported Collaborative Design
eCAADe : Education and Research in Computer Aided Architectural Design in Europe
EDM : The Express Data Manager FEM : Finite Element Method GDA : Generative Design Agent
IEEE : Institute of Electrical and Electronics Engineers IT : Information Technology
MEL : Maya Embedded Language
NCSA : National Center for Supercomputing Applications RAM : Random Access Memory
SIA : The Swiss Architects and Engineers Association UCVA : User Centred Virtual Architecture
VDE : Virtual Design Environment VR : Virtual Reality
vii LIST OF FIGURES Page Nr Figure 2.1 Figure 2.2 Figure 2.3 Figure 2.4 Figure 2.5 Figure 2.6 Figure 2.7 Figure 2.8 Figure 2.9 Figure 2.10 Figure 2.11 Figure 2.12 Figure 2.13 Figure 2.14 Figure 2.15 Figure 2.16 Figure 2.17 Figure 2.18 Figure 2.19 Figure 2.20 Figure 2.21 Figure 2.22 Figure 2.23 Figure 3.1 Figure 3.2 Figure 3.3 Figure 3.4 Figure 3.5 Figure 3.6
: An agent interacts with its environment through its sensors and effectors... : An agent model with 5 processes... : The general agent model... : The swarm agent model ... : A virtual meeting room in Second Life ... : Global behavior of the swarm, the number of avatars present
increases: Floor, Beam and Roof agents expand and Chair agents multiply…... : A swarm can adapt to different situations without specific
programming... : The cognitive agent model... : A virtual meeting room in Active Worlds... : Interaction diagram showing communication within the virtual
meeting room society when a new avatar enters a full
room... : The motivated agent model... : A virtual meeting room simulated in Java 3D... : Agents experimenting by modifying the scale and position of
the Wall and Floor objects... : Agents learn to cause interesting events to occur, such as
room expansion which causes the number of avatars using the room to increase... : A user entering the virtual building... : The connections between user, modelling tool and agents... : Modelling with Solid and Voids: Fictitious interior (left), a
residence by Snozzi (center), grid overlay as modelling support (right)... : Design in the Sculptor environment: An object (left), its
decompositions (center), gravity and collusion applied (right). Sculptor’s user interface is intentionally simple and not based on graphical interface elements... : The graph of connections between rooms generated by the
Navigator... : Simple Generative structure... : Generative Output... : Relationship families... : Second order agents... : Facade of the “Ospedale degli Innocenti”... : Vitruvian Man by Leonardo Da Vinci... : Santa Maria delle Fiore by Filippo Brunelleschi... : Space House,1933... : A constructed Bubble House, and a worker spraying concrete
to the inflated balloon... : Sketch and model of the Endless House... 11 12 14 18 18 19 19 20 21 22 24 25 25 25 29 29 29 30 32 39 39 40 41 45 46 47 49 50 51
viii Figure 3.7 Figure 3.8 Figure 3.9 Figure 3.10 Figure 3.11 Figure 3.12 Figure 3.13 Figure 3.14 Figure 3.15 Figure 3.16 Figure 3.17 Figure 3.18 Figure 3.19 Figure 3.20 Figure 3.21 Figure 3.22 Figure 3.23 Figure 3.24 Figure 3.25 Figure 3.26 Figure 3.27 Figure 3.28 Figure 3.29 Figure 3.30 Figure 3.31 Figure 3.32 Figure 3.33 Figure 3.34 Figure 4.1 Figure 4.2 Figure 4.3 Figure 4.4 Figure 4.5 Figure 4.6 Figure 4.7 Figure 4.8 Figure 4.9 Figure 4.10 Figure 4.11 Figure 4.12 Figure 4.13 Figure 4.14 Figure 4.15 : Monsanto House... : Model and section of the “Arango House” by John Lautner... : H2O Expo by NOX... : Saltwaterpavilion and its interior... : Experience Music Project of Frank Gehry... : Construction methods that was used in the Gehry’s Zollhof
Complex in Dusseldorf... : Sydney Opera House and detail of its roof... : Concrete ribs of the Sydney Opera House... : Final design of the shells of Sydney Opera House... : The symbiotic relationship between the historical setting and
the “Friendly Alien”... : Kunsthaus Graz’s “noozles”………... : Kunsthaus’s facade... : BMW Pavilion by Bernhard Franken... : The theme of the BMW Pavilion “water drops”... : 3D Model of the Pavilion that shows the structural system... : Aluminum ribs that described the form and plexi-glass panels
covering the surface of the BMW Pavillion... : Pictures of Mur by Vito Acconci ... : Interior of the Island in the Mur... : Island in the Mur: a structure made of steel, glass, rubber,
asphalt, water and light... : SolarCar by Ross Lovegrove ………... : Lovegrove’s wide preferences in materials from left to right,
Speakers, Go Chair, Digital Camera... : Sketch of the water bottle and the final product... : Miran Gallery: complex curved shell by Decoi... : Miran Gallery was designed as space that has both display
and storage functions………... : from left to right, Vitra Fire Station, Bergisel Ski Jump
and BMW building... : Olympic Aquatics Centre and its relationship with Olympic
Park... : Olympic Aquatics Centre with and without the additional
seats... : Nuragic & Contemporary Art Museum in Cagliari... : FEM analyses of the Bernhard Franken’s “The Bubble”... : Workflow diagram of the DROP... : Duplication of the initial surface... : Positioning the duplicated surface... : Selecting the duplicated surface as an emitting source... : Defining rate of the particles………... : Assignment of the geoconnector to the surface... : Defining gravity... : Defining collusion options... : Simulation of the rain flow analysis... : Simulation of the rain flow analysis tested on the Kresge
Auditorium... : Simulation of the rain flow analysis tested on the Los
Manantiales Restaurant... : Analysis with 150 particles... : Pictures of the Kresge Auditorium………... : Mathematical Equation that was used to calculate the
dimensions………..…….... 52 52 53 54 55 57 58 58 59 60 60 61 62 62 63 63 64 64 65 66 66 67 68 68 69 70 71 72 75 80 83 83 84 85 85 86 86 87 89 89 90 91 92
ix Figure 4.16 Figure 4.17 Figure 4.18 Figure 4.19 Figure 4.20 Figure 4.21 Figure 4.22 Figure 4.23 Figure 4.24 Figure 4.25 Figure 4.26
: 2D drawings of the Kresge Auditorium... : First step of the modeling phase of Kresge Auditorium... : Second step of the modeling phase of Kresge Auditorium... : Final 3D model of the Kresge Auditorium in MAYA... : Results of the analysis of the Kresge Auditorium... : Los Manantiales Restaurant... : Sketches of the Los Manantiales Restaurant that shows its
features... : 2D and 3D drawings of the Los Manatiales Restaurant... : 3D solid model of the Los Manantiales Restaurant... : Results of the rain-flow analysis conducted on the Los
Manantiales Restaurant (on solid model)... : Results of the rain-flow analysis conducted on the Los
Manantiales Restaurant (on wireframe model)... 92 92 93 93 93 94 94 95 95 96 96
x
SERBEST BİÇİMLİ MİMARİ TASARIM ALANINDA BİÇİM
DEĞERLENDİRMESİNDE KULLANILACAK ETMEN-TABANLI BİR MODEL: DROP
ÖZET
Tasarımcıların, genellikle de mimarların tasarım sürecinde hasaba kattıkları birçok parametreden birisi de fikirlerini tasarladıkları haliyle somutlaştırabilmektir. Bazı durumlarda, özellikle de üzerinde çalışılan obje büyük ölçekli değilse ve deneme-yanılma gibi pratik yöntemlerle tasarlanabiliyorsa bu amaca ulaşmak o kadar da zor değildir. Fakat mimarlar tasarlanan objenin tatmin edici ve istenen son tasarım durumuna ulaşmak için birçok kez değişik tasarım olasılıklarını değerlendirmek zorundadırlar. Serbest biçimlerle ilgilenen mimarlar ise ulaşılmak istenen sonuç ile tasarlanan objenin üretim ya da inşasından sonraki durumu arasındaki, yani zihindeki tasarımla üretilen sonuç arasındaki benzerliği sağlamak için daha fazla çaba göstermek zorundadırlar. Tasarımcıların tasarladıkları objenin başarılı bir biçimde üretilebileceği ya da inşa edilebileceği ve işlevini bekledikleri ya da tasarladıkları şekilde yerine getireceğini test edebileceği çeşitli araçlar bulunmaktadır. Tasarımların dijital ortamda geliştirilen görsel benzetim modelleri mimari tasarım süreci içerisinde önemli yere sahip, tasarımcıya tasarımını test etme olanağı sağlayan araçlardandır. Fakat bu modeller tasarımın uygulanabilirliğine dair yeterli veri sağlamamaktadır. Bu nedenle, tasarımcıya, objenin tasarlanan biçimde üretilip üretilemeyeceği konusunda fikir verebilecek bir araç ihtiyacının bulunduğu çok açıktır.
Tezin amacı, tasarımcılara serbest biçimli tasarımların üretimi konusunda, tasarlanan biçimin dijital ortamda analizi ve değerlendirmesi yapılarak, yardımcı olabilecek etmen-tabanlı bir aracın – bir program bileşeni anlamında – geliştirilmesidir. Tasarımcılara serbest biçimli tasarım üretimi konusunda yardımcı olabilecek bir program bileşeni geliştirilmesi konusundaki asıl önemli olan konu bu program bileşeninin kullanılacağı problem alanının tanımlanmasıdır. Program bileşeninin kullanılacağı problem alanına, serbest biçimli tasarım alanındaki strüktürle ilgili, kullanılan malzemelerle ilgili ve biçim üretimi konusundaki güçlükler araştırılarak karar verilmiştir. Tezin odaklandığı asıl problem alanı ise serbest biçimli tasarım alanındaki biçim üretimi konusu ile ilgili problemlerdir.
Tasarımlarında sadece düzensiz eğrilerle açıklanabilecek ve serbest biçimli mimari tasarımla birlikle düzensiz eğrilerden oluşan biçimlerin tasarlanmasında yararlanılan geleneksel olmayan geometrik yaklaşımı da açıklamak için kullanılan bir estetik bütünlüğe ulaşabilmek mimarların mimarlık tarihi boyunca ulaşmak istedikleri bir durum olmuştur. Akışkanlığı ve hareketi tasarımlarında ifade edebilme isteğinin, Art Nouveau’nun bitkisel formlarında ve barok mimarlığın formları birbiri içerisinde akıyormuş gibi düzenleyen stilinde görülebileceği gibi farklı mimari dönemlerde de etkin olduğu izlenmektedir. Günümüzde de, bu estetik bütünlüğe ulaşma isteği, avant-garde öncülleri gibi biçim arayışlarında ve kullandıkları temalarda fazla deneysel davranmaktan korkmayan yeni nesil mimarlar tarafından da sürdürülmektedir.
Tasarımda, bahsedilen estetik bütünlüğe ulaşmak bilgisayarların yardımıyla daha da kolaylaşmıştır. Bilgisayar teknolojisindeki yeni gelişmeler ve bilgisayarların mimari tasarımdaki sürekli artan kullanımlarıyla mimarlık alanında önemli yenilikler
xi
olmuştur. Bunlardan en fazla öne çıkanı, serbest biçimli tasarım örneklerinin sayısındaki artış ve serbest biçimli mimari tasarım örneklerinin, bilgisayar teknolojisinin kullanımının sağladığı faydalar sayesinde bilgisayar yardımıyla yönetilen yöntemlerle tasarlanmalarını ve üretilmelerini sağlayan üretim tekniklerindeki değişimlerdir. Bu gelişmelerin sonucunda, güncel mimari tasarımların birçoğu bilgisayarla üretilmiş ya da yeni donanım ve yazılımların kullanımı ile üretilecek duruma gelmişlerdir. Bilgisayarların, çeşitli çizim ve modelleme yazılımlarının mimari tasarım sürecinde aktif olarak kullanılmaları ile birlikte tasarımcılar her zaman tasarlamayı hayal ettikleri karmaşık biçimlerin üretilmesi ve tasarlanması ile ilgili konulara daha hakim duruma gelmişlerdir. Yine de, düzensiz biçimlerin tasarımı ve üretilmesi konularına tamamen hakim olabilmek için ilk önce bu biçimlerin tasarımı ve üretilmesi ve bu alandaki problemler hakkında bilgi sahibi olmak gerekmektedir. Serbest biçimli tasarım alanındaki objelerin biçim arama sürecindeki problemlerin açıkça tanımlanması ile mimarların bu alışılmışın dışındaki biçimleri tasarımlarında kullanabilme istekleri gerçekleşebilir.
Serbest biçimli mimari tasarımdaki biçim-arama süreci büyük ölçüde yüklerin biçim üzerindeki dağılışına bağlıdır. Yüklerin strüktür üzerinde dağılımı ve tasarımın strüktürel kararlılığa sahip olup olmadığı serbest biçimli tasarımda biçim arama sürecini şekillendiren kritik noktalardır. Tasarımın yükler altındaki davranışına karar verebilmek için kullanılan az sayıda pratik yöntem vardır. Bu yöntemlerden en kullanışlısı tasarımın kendi yükü altındaki davranışını “yağmur analizi” yöntemini kullanarak bulmaktır. Tezde “yağmur analizi” yöntemi düzensiz eğrilerden oluşan yüzeylerdeki yük ve biçim arasımdaki ilişkiyi araştırma amacıyla kullanılmıştır. Yağmur analizi yöntemini tasarımın, yüklerin yüzeydeki dağılımına göre performansını test etmek için kullanma fikri başka bir araştırmadan alınmıştır. Borgart, de Leuw ve Hoogenboom’ a göre “yağmur taneciklerinin akışı gibi, yükler de kabuk struktürlerin yüzeyinden taşıyıcılara doğru en çok eğime sahip eğriler üzerinde dağılırlar”. (Borgart, A., ve diğerleri., 2005). Kabuk strüktürlerdeki yüklerin dağılışına dair olan bu varsayım, serbest biçimli yüzeylerin uygulanabilirliğine; serbest kenarlardaki yükler, bükme momentleri ve yüklerin yoğun olarak toplandığı bölgeler göz önünde tutularak karar vermede kullanılmıştır. Eero Saarinen’in tasarladığı Kresge Oditoryumu, ve Felix Candela’nın Los Manantiales Restoranı geliştirilen modelin güvenilirliğinin, daha önce bahsedilen serbest biçimli mimari tasarım örneklerinde görülebilen tasarım hataları göz önünde tutularak test edileceği örnek çalışmalar olarak seçilmiştir.
DROP, etmen-tabanlı bir aracın, tasarımcılara serbest biçimli mimari tasarımların üretilmesi ve değerlendirilmesi konularında yardımcı olabileceğini göstermek için geliştirilmiş ve bu amaçla kullanılmıştır. Bu tez çalışmasının sonuçları, serbest biçimli mimari tasarımların uygulanabilirliğine dair yeni yöntemler oluşturmakta kullanılabilir. Gelecekte serbest biçimli mimari tasarımların tasarım sürecinde karşılaşılabilecek diğer problemleri çözmede tasarımcılara yardımcı olabilecek etmenlerin tasarlanması planlanmaktadır. Yapılacak eklentilerle birlikte geliştirilen modelin, tasarımcıların serbest biçimli mimari tasarım konusunda daha kapsamlı ve açık bir anlayışa sahip olmalarına yardımcı olabileceği düşünülmektedir.
xii
AN AGENT-BASED SYSTEM TOOL TO SUPPORT DESIGNERS FOR FREE-FORM SHAPE EVALUATION: DROP
SUMMARY
One of the important parameters of the design process is to be able to materialize the scheme in the mind. In some cases, especially if the object of the work is acceptable in size or if it can be done with some practical methods like trial and error, realizing that goal would not be so difficult. But architects have to search through several design possibilities back and forth in order to achieve the final satisfactory and desired design state. But in order to achieve the resemblance between the intended result and the final state of the design or product after the construction phase in the domain of free-form design, architects have to take one more step further. There are some tools that can help designers to test their design to see if it will be successfully constructed or functioned as close as possible as the way they imagined or expected; visual simulation models in digital environments of the designs are such important tools for architectural design process, but these models does not provide enough data to determine the design’s applicability. Therefore, the need of a tool that would work with the designer in order to determine the possibility of whether the design could be built or not as the way designer imagined, is vital.
The aim of this thesis is to explain that an agent based system tool – in the sense of a program component – can help designers in the free-form shape generation by analysing and evaluating the form in digital media. The main issue about developing a program component, that is capable of assisting designers in free-form shape generation, is defining the problem domain that tool will be used. The problem domain, which the tool would be affective against, was decided by examining the difficulties within the design domain, such as difficulties concerning the structural components, materials and the form finding process, the research which is explained in this paper only focuses on the difficulties concerning the form finding process of the free-form design.
The desire to achieve an aesthetic condition in their designs which can only be explained by the irregular curves; which is also used to describe the free-form architecture and the unconventional geometric approach that was used to produce these irregular curved forms, have always been haunted the architects throughout the architectural history. The roots of this yearning to capture the flow and movement and to be able reflect this notion into their designs can be traced back to the different architectural eras, such as; decorative, flowerlike forms of Art Nouveau and immersed style of Baroque architecture. In addition, this quest is pursued by a new generation of architects who does not afraid of being too experimental in their search of forms and themes not so different than their neo-avant-garde predecessors.
Being able to achieve that aesthetic state in design has become easier with the help of computers. By means of recent developments in the computers and their ascending use in architectural design, there have been serious innovations in the domain of architecture: the most important ones among them are the increase in numbers of observations of free-form architecture more than ever before and the
xiii
change in construction techniques; which have led these examples of free-form design become products of computerized procedures with the benefit of recent use of digital computer technology. As a result of these developments, most of the contemporary designs are computer generated or about to emerge as a result of the implementations of latest computer hardware and software. With the aid of computer animation software, designers can master complicated forms that they ever dreamed of. Nevertheless, in order to master these irregular forms, first of all one have to be capable of outlining the knowledge about the design and production phases, and some difficulties in this domain. By successfully and clearly outlining and describing the problems in the form- finding process of the artefacts which belongs in the free-form design domain, this lifelong search of architects to design these unusual forms, could finally be realized.
Form finding process of the free-form design is mainly depended on the distribution of the forces. How the forces distributed along the structural system and whether the design meets the structural stability or not, are the critical questions that define the form finding process of the free-form design. There are not many practical ways to determine if the design is capable of baring these forces or not. But one of the most practical among the approaches is using the rain-flow analysis method to test the performance of the design solution under these forces. In the thesis, rain-flow analysis is used for searching the relationship of form and force in (irregular) curved surfaces. The idea of using rain-flow analysis method to test the performance of the design according to the distribution of the forces was borrowed from another research. That research stated that, “Like a rain flow loads will flow along curves with the steepest ascent on the shell surface to its supports.” This hypothesis for the flow of forces in shell structures was used to determine the feasibility of the free-form surfaces according to three design flaws: loads on the free-edges, bending moments and drain curves (valleys). In order to test the tool’s reliability, Eero Saarinen’s Kresge Auditorium in MIT Campus and Felix Candela’s Restaurant Los Manantiales in Xochimilco, Mexico are selected as case studies for which the tool tested out according to previously mentioned design flaws regarding to free-form surfaces.
The digital tool “DROP” was developed and used to exhibit the ways that an agent based tool could be used to assist designers for free-form shape generation and evaluation. The outcome of this thesis can be used to specify novel ways to determine the applicability of the form of the free-form architectural designs. In the future, we aim to develop other agents which would deal with other issues related to the design process of free-form architectural designs; like analysing the form according to the choice of materials or computing the optimal size of the divisions on the surface. This way the tool would give us a clear and through insight regarding designs.
1 1. INTRODUCTION
This thesis focuses on the development of an agent based system tool – in the sense of a program component - to help designers for free-form shape generation in digital media by analyzing the geometry of the free-form they design according to the distribution of the forces (applicability of the form). Free-Form architecture could be described as simply as designing volumes, surfaces: at the end artefacts, which are irregular, curved. Architecture has come a long way from just designing on the paper to use of computers. With the expanding use of computers as a new design media, perception of architecture has changed. Today computers used not only as drawing tool but also like an assistant at the design process and the total number of designers who use computers during the whole design process from scratch to the last detail, is rapidly increasing. The reason for this choice mainly lies on the many benefits of using computers in design process. Architects have become more liberated in the way they design with the developments in the computer technologies. Almost every shape can be realized within the digital media as well as in the physical world with the aid of new digital design techniques. After this evolution in the computer technology, many architects have begun to explore the possibilities of new topological surface organizations alternative to the Cartesian volumes.
In the thesis, we attempt to present a partial insight regarding free-form architecture, and then we propose a novel way/tool to use during the design processes within this new domain of architecture called free-form design. Working on a new evolutionary design domain in architecture like free-form design, requires extensive and undivided attention both on design and production phases of these artefacts. Aim of this research project is to develop an agent system based tool to help designers for free-form shape generation in digital media. The key reason of doing a research on free-form shape generation is to able give designers control over the design process of free-form buildings. Computers are not new tools for the architectural practice. Architecture and industrial design firms have been using computer aided drafting programs for more than a decade and that technology is advancing aggressively since then. Nevertheless the unconventional relevance of dynamic modelling software and animation applications to architecture and industrial design is a rather
2
a recent phenomenon. Architects like Ben Van Berkel and designers like Karim Rashid stand aside from the rest and use software intended for making of special effects and computer game development like Maya, 3ds max, Rhino, FormZ to represent their designs in ways beyond the potentials of traditional CAD software. More and more computer became the major tool of the design profession because of its ability to enhance efficiency in the design process and for its remarkable flexibility as an artistic instrument. Greg Lynn has explicitly admitted this increasingly symbiotic relationship between designer and computer to Alexander Stille in an interview. Greg Lynn stated, “If it comes down to it, I would have to give the software 51 percent of the credit for the design of my buildings.” (Stille, 1998). This research aims to be the next step of this exiting evolution and try to develop the tool, which will act as a colleague to the designer by interacting him/her with the means of sharing of ideas and pointing out the problems.
The main motivation to start this research has been to develop a model and a prearranged framework, which is capable of outlining the knowledge about the design and production phases, and some difficulties in this domain. The main reason behind developing such a framework is to have it as a foundation of the future work related to the tool we propose. By successfully and clearly outlining and describing the problems in the design process of the artefacts which belongs in the free-form design domain, making necessary additions to the existing design tool could be possible and a lot less time consuming in the future. Thus, we focused on pointing out the possible problem areas within free-form design domain while developing the tool capable of dealing with at least one of these difficulties in the domain of free-form design.
In the research, the main technique used to determine whether the geometry of the free-form design is applicable or not is the rain-flow analysis, which is simulated in Maya. Rain-flow analysis is used for analyzing the relationship of form and force in (irregular) curved surfaces. The idea borrowed from a research made by Andrew Borgart, March de Leuw and Pierre Hoogenboom. Bogart, de Leuw and Hoogenboom stated, “Like a rain flow loads will flow along curves with the steepest ascent on the shell surface to its supports.” (Borgart, de Leuw and Hoogenboom, 2005). That hypothesis for the flow of forces in shell structures was used to determine the applicability of the free-form surfaces according to three design flaws: loads on the free-edges, bending moments, drain curves (valleys).
3 1.1 Free-Form Design
Irregular curves that used to describe free-form architecture and the unusual geometric approach, which used in the process of producing these -as almost liquid- forms have always been something that architects have wanted to achieve throughout the architectural history. This disposition can easily be seen in the examples of different architectural eras, such as; decorative, flowerlike forms of Art-Nouveau, nested/immersed style of Baroque architecture and the organic design repositories of twentieth century.
Among various exhilarating changes in the architecture, spatial form has evolved into a new plane with the recent developments in the information age. With the increasing use of computers in architectural domain, digital free-form architecture had become seen frequently than ever before. Computers had become to use in several stages of architectural design process concerning form, structure, and planning. Most of the contemporary designs are computer generated or about to emerge as a result of the implementations of latest computer hardware and software.
By means of recent developments in the computers, there have been serious innovations in the domain of architecture especially in spatial form. Construction techniques have increasingly become products of computerized procedures with the benefit of recent use of digital computer technology. Nowadays computers are not just used as a tool to draw in the design process as they used to be, they have taken place in almost every phase of the construction process. Nevertheless, with the spreading use of computers in the design process with new techniques such as CAD (computer aided design), CAE (computer aided engineering), CAM (computer aided manufacturing) and CNC (computer numerical control), form has reached a fascinating state. With the aid of computer animation software like Maya, Rhino, 3Dsmax, Cinema 4D, designers can master complicated forms which consist of irregular curves that they ever dreamed of. Computers can also assist the designers in construction phase.
1.2 Opinions about Free-Form Architecture in Academic World and Practice The Free-Form design is mainly described by designs of different international design groups and firms like Gehry Partners, Asymptote, Decoi, Eisenman Architects, Greg Lynn Form, Nox, Mecanoo, UN Studio, Oosterhuis NL, VVKH Architects, Foreign Office Architects and many more. These firms have enormous contribution to the establishing and illustrating of the new architectural
4
morphologies, different design strategies, documentation, and explanation of the design and construction processes. After the construction of Gehry’s 20th century Guggenheim Museum in Bilbao, architecture evolved into something new by means of novel geometric and digital approaches in architectural design. With the new computerized techniques that Guggenheim introduced, domain of free-form architecture has elevated into another ground; and finally digitally produced architectural form began to appear more frequently not just on the screen, but also on real environment. Moreover, a multidimensional and virtual digital environment with astounding developments in the field of digital design progressively replaced conventional emphasis on spatial form, its static space, stability, along with limitations imposed on structure. Without these limitations, a designer can freely roam through the limitless design possibilities.
Free-Form architecture has found its place in the academic world with different courses held in universities. Some of these courses are/were being lead by architects whom mentioned above. Focal points of these courses vary from “paperless architecture” experimentations within the digital space and experimentations with various CAD/CAM tools.
In the Fall 1994, The Columbia University’s Graduate School of Architecture, Planning and Preservation, under the direction of Dean Bernard Tschumi, established the first “paperless design studios”. As digital design technology becomes the norm in schools and practices, architects are pushing materials manufacturers to expand their modes of production through digital technology as well. The proliferation of the digital knowledge at the turn of the century has changed the pedagogy, practice, and general public’s perception of architecture. Gehry’s 2000 EMP is a model for what digital architecture can be in the 21st century (Rosa, 2001).
Lars Spuybroek – head of the Rotterdam based office NOX architects – is directing the design studio in Digital Design Techniques department at the University of Kassel, Germany along with Christian Troche and Geri Stavreva from 2002. Spuybroek mainly focuses on the interactive relationship between architecture, art, media, and computing through digital and analog techniques.
The Digital Design Techniques department is dedicated to the research into the usage of computing in three different phases of the building process: before a building is built, during the building process, and after it has been built. In other words, the department focuses on digital design techniques, digital manufacturing
5
techniques, and the digital interactive environments. The main characteristic of these techniques is that form is generated through rule-based procedures. Because of the complex nature of the teaching of digital design techniques, issues like design methodology, software proficiency, philosophy and architectural theory, panelization and manufacturing, were dealt with on different levels. In addition, the students who attend the digital design techniques program can learn software programs and programming in separate courses.
1.3 Problem Definition
Publications in the field described in the previous section are highlighting the principles, theories, and methods concerning the subject on individual cases. They provide a solid foundation for the up and coming design content and have a large impact on the design theory. Nonetheless, there is not enough number of researches about dynamics of the design process of free-form architecture and its limitations. As mentioned before, most of the literature is focus on case based analysis of various methods of form generation, novel manufacturing techniques, and individual buildings as examples of free-from architecture. Furthermore, there is a lack of research in the field of decision-making and form generation regarding design process of free-form architecture. Therefore, designers have to think with a limitation while designing because of not having enough knowledge about possible new methods that could be used in form finding process of the free-form architecture. Designers who has limited or no knowledge about the limitations about material, structure etc. which gives the design artefact its final shape, bound to spend considerable amount of time while deciding the form of the building and might have to compromise some elements that are crucial in the context of their design just to make their design applicable to realized. Therefore exploring the possible limitations concerning this emerging field and developing a tool to help designers to solve problems that resulted because of these limitations and optimize the design successfully is crucially important and the focus point of this research.
1.4 Research Objectives
The thesis suggests that with describing and documenting the difficulties concerning the design stage of free-form design like structural characteristics, material and form finding process related issues; the core of free-form domain can be captured in a way that we can easily examine and identify different design solutions regarding our design within the free-form design domain.
6
In order to capture the essential core elements that that mainly alter the design in the form design domain, we try to entrap the available knowledge about free-form design and agent systems in the literature. We suggest that having that kind of knowledge would be an effective way to identify the main issues regarding the form finding process of the free-form design and then present possible solutions to these issues and a novel way to deal with the form finding process with the help of a digital tool.
7 2. AGENT DEFINITION and AGENTSYSTEMS
Agent- based computing started in the 1970s, since then agents have become used frequently in internet applications, simulations, etc. with their high compatibility to use Artificial Intelligence. Agents have goals and beliefs and execute actions based on those goals and beliefs (Wooldridge, 2000). These features set apart a rational agent from the computational agents that perform actions based on predefined events, like search agents on the web.
Agent systems mainly have a usage in the field of architecture as intertwined with the implementations of virtual worlds. When searching through the literature regarding usage of the agent systems related to the architecture, one definitely came across the rich body of work. Examples regarding this issue could be summarized as:
• agent approaches to sharing and synchronising building model data among CAD and virtual world systems,
• agent models for 3D virtual worlds,
• design agents for 3D virtual worlds, creativity in design, • situated agents,
• situated computing, and
• multidisciplinary design in virtual worlds.
As mentioned before, agents mostly have a usage in the virtual worlds in these examples; in some cases they used for evaluating the novelty of designs agent-based systems also have a use as a virtual meeting room that can change its dimensions in order to enlarge to accommodate all of the participants of a virtual meeting, and in another cases they exist as a door agent which have knowledge about whether let a person or in terms of virtual worlds avatar, in a room or not. However, these are not the only places that agents are handy, in the recent years many novel ways that agents can be used were witnessed.
Maia Engeli and David Kurmann presented a system compromised of an intuitive interactive modelling tool that run in a virtual reality set-up, where the user can use 3D glasses to experience rooms and 3D input devices to model in three dimensions.
8
Another feature of the system was its ability to run in a distributed fashion so that a number of users could look at and modify the same design. The important aspect of the Engeli and Kurmann’s research regarding the use of agents was that the autonomous agents that were added to the system to enhance the designer support. These were the agents that enhance the virtual environment, agents that take over tasks, and agents that help to test the design. Engeli and Kurmann’s research in 1996 was introducing a vision of a new design environment at that time (Engeli, Kurmann, 1996).
Jeffrey Krause is also one of the researchers who have contributed the literature with his works regarding agents within the framework of architecture. In his research, which he submitted to the ACADIA Conference in 1997, Krause tried to describe a behaviour based artificial intelligence experiment in computer generated design and explain the internal representations and procedures of an agent based system through suggesting an alternative design process -to the common hierarchical and top-down design process of architecture- involving multiple autonomous agents moving within the design landscape, simultaneously collaborating, building, degenerating and transforming their environment.
Rabbee M. Reffat’s work is also one of the mentionable works in the literature regarding alternative usage of agent systems in architecture. In his paper, Reffat concerned with developing intelligent design agents that would be capable of inventing creative concepts of design forms, shapes, and compositions while involved in the design process. Reffat explored a completely new area that shows promise with his interpretation of concept invention with his research in 2002. One year later Reffat presented two other papers at the 21th eCAADe Conference at Graz, Austria. The theme of one of his papers was to introduce a framework of semantic–based Virtual Design Environment (VDE) that aims to provides designers of Virtual environments, which Reffat strongly believe that they have the potential to reach beyond the limitations of CAD systems and can be utilised as design tools for architecture (Reffat, 2003), with virtual observers of designers’ actions to investigate the current design and respond to these actions when the need arises. The second paper Reffat presented was, similarly to his other works, a documentation of the results of the developed system of intelligent design agents that supports design exploration and creativity within the domain of architectural shapes.
Another unorthodox research regarding the issue was done by Rob Saunders and John S. Gero and published on the Proceedings of the Ninth International Conference on Computer Aided Architectural Design Futures in 2001. Saunders and
9
Gero’s paper was dealing with the motivation of design agents to promote the exploration of design spaces. Regarding to this novel research field, Saunders and Gero proposed computational models of interest and curiosity based on the detection of novelty. In their research, the behaviour of the model of interest was illustrated by developing a design agent that is motivated to explore the effects of emergent crowd behaviours on the performance doorways.
Mary Lou Maher, Gregory J. Smith and John S. Gero presented a model for a design agent reasoning process that includes sensation, perception, conception, hypothesizing and planning sequence of actions and a model for constructing memory of the agent’s knowledge and interaction with a virtual world which can be explained as a dynamic and changing view of the designed world that is determined by the agents sense data and reasoning at the Workshop on Cognitive Modelling of Agents and Multi-Agent Interaction in 2003. These models were implemented by extending the Active Worlds platform so that each object in the 3D world can have agency and illustrated with a door agent and a multi-office that operate within a 3D world (Maher, Smith, and Gero, 2003). Mary Lou Maher, Pak-San Liew and John S. Gero also presented another paper proposing an agent approach to sharing and synchronizing building model data among CAD and virtual world systems in the conference on IT in Construction in 2003 (Maher, Liew, and Gero, 2003). Another research presented by Ning Gu and Mary Lou Maher at the CAADRIA symposium in 2003 was about a User-centred Virtual Architecture (UcVA) Agent, a kind of rational agent capable of representing a person in virtual worlds and designing virtual worlds based on current needs (Gu, and Maher, 2003). Also in 2003, Gregory J. Smith, Mary Lou Maher, and John S. Gero presented a framework in which agents become the basis for the elements of a 3D virtual world. This framework was presented as having a model for an agent that can interact and reason about the 3D world, and as a model for agent communication. The model was illustrated by the design of a virtual conference room (Smith, Maher, and Gero, 2003).
Mary Lou Maher also continued to work on her and so many other researchers’ mutual point of interest in 2005. Rosenman M.A., Smith G., Ding L., Marchant D. and Maher M.L. proposed an environment which provides real-time multi-user collaboration in a 3D virtual world for designers in different locations. Agent technology is used to manage the different views, creation and modifications of objects in the 3D virtual world and the necessary relationships with the database(s) belonging to each discipline (Rosenman, Smith, Ding, Marchant and Maher, 2005). Like their previous research regarding integrating an User centred Virtual
10
Architecture Agent which has an ability to change the environment based on current needs, into a virtual world in 2003, Ning Gu and Mary Lou Maher also suggested a different approach to develop an agent model that is associated with an individual person in the 3D virtual world as a personal design agent in their paper which they submitted to the CAAD Futures symposium in 2005. Maher and Gu presented a Generative Design Agent (GDA), a kind of rational agent capable of representing a person in a virtual world, i.e. Virtools, Active worlds, and designing, implementing and demolishing 3D virtual places based on the occupants’ current needs in the virtual world (Gu, and Maher, 2005). Mary Lou Maher and Kathryn Merrick experimented with three agent models for generating a dynamic, adaptive environment that changes in response to users’ actions. Maher and Merrick introduced a schema for characterising the implementation and behavioural complexity of agent models for dynamic virtual environments in their paper which was presented in IEEE symposium in 2005 (Maher, and Merrick, 2005).
Mary Lou Maher and Mike Rosenman, used agents to maintain different views of a single design in order to support multidisciplinary collaboration and address issues such as multiple representations of objects, versioning, ownership and relationships between objects from different disciplines in the paper they presented in 2006 (Maher and Rosenman, 2006). Mary Lou Maher and Kathryn Merrick suggest motivated learning agents as a means of creating non-player characters that can both evolve and adapt in persistent computer game worlds (Merrick and Maher, 2006).
2.1 Definition of Agent Systems
As mentioned at the beginning of this chapter, agent- based computing started in the 1970s and from that time, the concept of agents has become important for internet applications, drawing ideas from Artificial Intelligence and Artificial Life. Although there is not a universal definition for the term agent, agents could be described in the context of computer science as intentional systems operate independently and rationally, seeking to achieve goals by interacting their environment (Wooldridge and Jennings 1995). This idea of an agent is illustrated in Figure 2.1.
11
Figure 2.1: An agent interacts with its environment through its sensors and effectors (Maher, Liew and Gero, 2003)
An agent has the ability to operate usefully. However, the increasing interconnection and networking of computers is making this situation rare. Typically, the agent interacts with other agents (Huhns and Stephens, 1999). Hence, the concept of multi-agent system is introduced with the applications of distributed artificial intelligence.
Object-oriented programming is one of the major types of programming methods. In object-oriented systems, objects are defined as computational entities that encapsulate some states, are able to perform actions or methods on this state, and communicate by message passing. There are similarities between agents and objects, but there are also significant differences (Wooldridge, 1999).
The differences between agents and objects can be explained in two distinctive features:
• Agents embody a stronger notion of autonomy than objects, and in particular, they decide for themselves whether to perform an action on request from another agent or not.
• Agents are capable of flexible (reflexive, reactive, reflective/proactive and social) behaviours, and the standard object model has nothing to say about such types of behaviours.
The intelligence of agents also reflects on its direct interaction with multi-agent environments, Huhns and Stephens summarize the characteristics of multi-agent environments as:
12
• Multi-agent environments provide an infrastructure specifying communication and interaction protocols.
• Multi-agent environments are typically open and have no centralized designers. • Multi-agent environments contain agents that are autonomous and distributed,
and may be self-interested or cooperative (Huhns and Stephens, 1999). 2.1.1 Components of the Agent Models
Maher, Liew and Gero developed a multi agent system as the core of a 3D multi-user virtual world. Each object in the world is an agent in a multi-agent system. The agent model provides a common vocabulary for describing, representing, and implementing agent knowledge and communication. Maher, Liew and Gero’s common agent model has sensors, effectors, and five kinds of reasoning: sensation, perception, conception, hypothesizer, and action. This agent model is illustrated in Figure 2.2.
13
The components of the agent model are described below.
• Sensors recognize two kinds of events: sense_data in which the agents identifies relevant data by monitoring the virtual world and the common The Express Data Manager Developer (EDM) data model, and receive_data in which the agent receives a message from another agent. An example of sense-data for a virtual world agent is a change in an object in either the collaborative virtual world, or the EDM database.
• Sensation transforms raw input from the Sensors into structures more appropriate for agent reasoning and learning.
• Perception is a process that finds grounded patterns of invariance in the agent’s representation of the sensed data. For example, a concrete wall agent may know about patterns in the aspect ratios of the dimensions of the wall. Perception is goal driven by concepts that the agents knows about and data driven by the sense_data.
• Conception associates concepts with sense data or patterns of sense data. Concepts are abstractions of experience that confer a predictive ability for new situations. The concept of a load-bearing wall, for example, is a representation of the behaviours and functions of the wall agent, and its meaning is its predictions of possible interaction with other building elements.
• Hypothesizer identifies mismatches between the current and desired situation, which goals are relevant to the current state of the building models and reasons about which goal should be achieved in order to reduce or eliminate that mismatch. It identifies possible actions which when executed will change the building model to meet those goals.
• Action reasons about which sequence of operations on the building model, when executed, can achieve a specific goal.
• Effectors are the means by which actions are achieved. Two types of effectors are: Change_data in which the agent causes a direct change to the virtual world or EDM database, and SendMsg_data in which the agent sends a message to another agent to respond by changing the world.
This agent model is derived from recent developments in cognitively based design agents, where design is considered as a situated act (Gero, 1998). The agents are developed to interact with the design and the design knowledge (Smith and Gero, 2001; Saunders and Gero, 2001).
14 2.1.2 Behavioural Complexity
According to Maher and Merrick agents are systems capable of perceiving their environment through sensors, reasoning about their sensory input using some characteristic reasoning process and acting in the environment using effectors (Maher and Merrick, 2005) (Figure 2.3). Agents are generally implemented using a programming language such as Java or C/C++ providing them with the potential for complex reasoning processes that produce dynamic virtual environments of greater behavioural complexity than those with scripted behaviours. Behavioural complexity measures the richness of the reasoning process that produces the dynamics of an environment. According to Maher and Merrick, behavioural complexity can be measured in five modes: reflexive, reactive, reflective (Maher and Gero, 2002), autonomous (Steels, 1995) and proactive (Wooldridge and Jennings, 1995). Reflexive being the simplest of all of them, each mode requires increasingly sophisticated reasoning.
Figure 2.3: The general agent model (Maher and Merrick, 2005)
• Reflexive behaviour is a pre-programmed response to the state of the environment - a reflex without reasoning. Only recognized states will produce a response. This mode of behaviour is typically achieved using a scripting language to implement behaviours associated with the 3D objects in a virtual world or environment. These scripts define behaviours that are triggered by predefined patterns of events.
15
• Reactive behaviour manifests itself as reasoning about responses within a fixed set of goals. This mode of behaviour is achieved by using agents to control one or more objects in the virtual world. Agents make changes to the objects in the world to work towards achieving goals in response to changes in the state of the environment. Reactive behaviour is a consequence not only the state of the environment but also how that state is perceived by each agent. Perception may vary as a consequence of experience.
• Reflective behaviour also has a fixed set of goals. In addition, it does not simply react but hypothesizes possible desired states of the environment and proposes alternate actions that will achieve those states. This type of behaviour is also achieved using agents. Reflective agents are able to reason about the world at different levels of abstraction.
• Autonomous behaviour does not simply select goals from a fixed set but includes reasoning processes to create new goals in response to new situations. • Pro-active behaviour goes beyond reasoning about goals to be achieved, and
hypothesizes possible undesirable future states of the environment, and proposes alternate actions to avoid those states (Maher, et al, 2005).
2.1.3 Implementation Complexity
Implementation complexity is a measure of the level of programming effort that is required to produce a certain level of behavioural complexity. Implementation complexity depends on properties of the system architecture such as domain dependence and social ability.
2.1.4 Domain dependence
Domain specific systems incorporate modules that tie them to a particular environment or problem within an environment. Examples of such modules are scripts defining the behaviour of a specific object or a set of goals relevant to a particular environment.
Domain independent systems are comprised of general modules that are relevant to a wide range of environments and problems. Domain independent systems have less implementation complexity than domain specific systems because domain specific systems must be modified each time they are to be applied to a new problem or environment.
16 2.1.5 Social ability
When the objects in a 3D virtual world or a virtual environment are controlled by more than one script or agent there may be a requirement for scripts for agents to communicate. Implementation complexity is affected by the level of communication or social ability of the system, with no communication being the simplest.
No communication implies that every script or agent functions independently of every other script or agent and does not receive any information about their behaviour or the region of the environment they modify.
Indirect communication or stigmergy is communication through environment. There are two types of stigmergy, discrete and continuous. When a script or agent responds to a structural change in the environment made by another agent or script it is responding to discrete stigmergy. Scripts or agents can communicate via continuous stigmergy by depositing pheromones. In nature pheromone is a chemical substance deposited by one individual that triggers some behaviour in another individual. In a virtual world, a pheromone can be modelled as an invisible 3D object with a type and strength. The type of pheromone determines the behaviour that is triggered. The strength of the pheromone determines the time it will take to decay and thus the amount of time it will be present to trigger behaviour from other scripts or agents.
Direct communication implies that messages are sent directly from one script or agent to another script or agent. This form of communication has the greatest implementation complexity. Smith defines an agent society as an aggregation of agents that share a common connection with a virtual world and have some ontological connection with each other (Smith et.al., 2003). For example, a Floor agent and a set of Wall agents might form a society that represents a virtual meeting room. When a group of agents form a society there is frequently a need for them to communicate. Direct communication allows agents to interact with other agents by sending messages from the effector of one agent to the sensor of another. This allows agents within a society to self-organize without flooding the virtual world with additional objects or events.
2.2 Agent models
Agent models describe ways to implement the characteristic reasoning process of one or more agents. Agents using different models may differ in their behavioural or implementation complexity and hence result in dynamic 3D virtual worlds of differing complexity. Mary-Lou Maher and Kathryn Merrick proposed three different agent
17
models in their “Agent Models for Dynamic 3D Virtual Worlds” titled paper in 2005. These three different agent models classified by Maher and Merrick are; swarm agent model, cognitive agent model, and a motivated agent model.
2.2.1 Swarm Agent Model
Swarm intelligence is the property of a system whereby the collective behaviours of unsophisticated agents interacting locally with their environment cause coherent functional global patterns to emerge (Payman, 2004). Maher and Merrick investigated swarm intelligence as the basis for an agent model in a virtual world with the purpose of determining whether it is possible for the objects in 3D virtual worlds to achieve globally coherent behaviour without the complexity of a structured communication protocol and multiple levels of reasoning.
In Maher and Merrick’s swarm based model (Figure 2.4), each agent has two internal reasoning sub-processes, sensation, and action. These processes are facilitated by three structures, sensors, memory, and effectors. Sensors sense the local state of the environment, which is, the environment within some small radius of the agent. This raw data is transformed by the sensation process into sense-data structures more appropriate for reasoning. Sense-data structures incorporate both the most recent raw data and the raw data sensed immediately prior to the most recent data. This process eventually allows a swarm agent to reason about changes in its environment as well as static states. The action process uses a reflexive mode to respond to sense-data by triggering effectors to make changes to the environment. Behaviours are selected by consulting a pre-programmed lookup-table of domain specific microrules that is stored in the agent’s memory. Microrules are boolean conditions about sense_data. When events in a virtual environment cause a microrule to evaluate to true, behaviour is triggered.
18
Figure 2.4: The swarm agent model (Maher and Merrick, 2005)
Maher and Merrick implemented a suite of microrules for reproduction, growth, clustering, stigmergic interaction and death in the swarm agent’s memory. The performance of swarm agents using these rules could be seen with the example of a virtual meeting room implemented in the Second Life (www.secondlife.com) virtual environment. In this implementation, agents function as members of a society in which each agent controls exactly one object in the 3D virtual environment. There are two Floor agents, four Wall agents, four Column agents, four Beam agents, two Roof agents, four Chair agents and a Table agent (Figure 2.5).
Figure 2.5: A virtual meeting room in Second Life (Maher and Merrick, 2005) Each agent can sense other agents and avatars that fall within a radius one and a half times the size of the radius of the object that it controls. If a small meeting room has reached its capacity of three avatars; one of the Wall agents senses the presence of three avatars and constructs sense-data corresponding to “the number
19
of people in the room is three” then The agent refers to its list of microrules and identifies that the current sense-data makes true the condition on the “grow when crowded” microrule. After that procedure the rule is fired and the Wall agent expands. Following this expansion, one of the Beam agents senses the presence of the newly enlarged Wall agent and constructs the sensedata corresponding to “Wall has a new scale of 1.5” like the Wall agent, the Beam agent also refers to its list of microrules and identifies that the current sense-data makes true the condition on the “grow when neighbour grows” microrule. Eventually the corresponding rule is fired and the Beam agent expands. This expansion is gradually communicated to all agents in the society via this process of discrete stigmergy until the entire meeting room has adjusted to the avatars’ presence in the room either by expanding (Wall agents) or multiplying (Chair agents)( Figure 2.6).
Figure 2.6: Global behavior of the swarm, the number of avatars present increases: Floor, Beam and Roof agents expand and Chair agents multiply (Maher and
Merrick, 2005)
While individual swarm agents display only simple, reflexive behaviour, the swarm as a whole is able to adapt to situations not pre-programmed in microrules. For example, the clustering microrules in our implementation are general enough to produce the three situations where chairs have adapted to different arrangements of the projector screen, even when there are avatars in the way (Figure 2.7).
Figure 2.7: A swarm can adapt to different situations without specific programming (Maher and Merrick, 2005)
20 2.2.2 Cognitive Agent Model
Maher and Gero’s original agent model (Maher, et.al, 2002; Maher, et.al, 2003; Smith, et.al, 2003) is shown in Figure 2.8. Agents using this model are capable of constructing increasingly complex interpretations of their environment. Their interpretations influence their selection of goals and actions. These interpretations are constructed using five successive reasoning sub-processes, sensation, perception, conception, hypothesizing, and action. Sensors, memory, and effectors are the three structures that facilitated these reasoning processes.
Figure 2.8: The cognitive agent model (Maher and Merrick, 2005)
Like swarm agents, cognitive agents function as members of a society in which each agent controls exactly one 3D object from the virtual world. But unlike swarm agents, cognitive agents sense the global state of their environment, which includes all agents and avatars that fall within the bounds of the society. Sensation transforms raw data from sensors into sense-data structures, which is more appropriate for reasoning. The perception, conception and hypothesizing processes each build upon the output of the former to build more abstract interpretations of the environment. Perception transforms sense-data into patterns called percepts that are used as the building blocks for concepts about recurring situations. Conception is the process of recognizing concepts and is the basis for hypothesizing desired situations. The hypothesizer identifies mismatches between the current and desired situations and reasons about which goals should be pursued to eliminate or reduce the mismatch.
