Luis Orozco CO-DESIGN AND AGENT-BASED METHODS FOR MULTI-STOREY WOOD BUILDING SYS- TEMS RESEARCH REPORTS Institute for Computational Design and Construction 14 RESEARCH REPORTS Institute for Computational Design and Construction Edited by Professor Achim Menges, AADipl(Hons) Luis Orozco CO-DESIGN AND AGENT-BASED METHODS FOR MULTI-STOREY WOOD BUILD- ING SYSTEMS © 2024 Institute for Computational Design and Construction University of Stuttgart Keplerstrasse 11 70174 Stuttgart Germany D 93 RESEARCH REPORTS Institute for Computational Design and Construction 14 ISBN 978-3-9824862-2-2 All rights, in particular those of translation, remain reserved. Duplication of any kind, even in extracts, is not permitted. The publisher has no responsibility for the continued existence or accuracy of URLs for external or third-party internet websites referred to in this book, and does not guarantee that any content on such websites is, or will remain, accurate or appropriate. To my father, for always supporting me and believing in me Foreword Luis Orozco’s dissertation examines new possibilities for multi-storey timber con- struction that arise from an integrative approach to the development of computational design methods, digital fabrication processes and novel building systems, known as co-design. The work is based on the research on multi-storey timber building systems of the Cluster of Excellence Integrative Computational Design and Con- struction (IntCDC) at the University of Stuttgart. Luis Orozco initially considered Co-design from the perspective of enabling interdisciplinarity during the design process. Building on this, agent-based methods are being researched that bring different, often conflicting design criteria into alignment in an integrative and inter- active process. As a basis for this, a systematisation of timber building systems and their data structures were created, and then agent-based modelling for the placement of columns, the subdivision of slabs segments and the arrangement of shear webs within hollow-box slabs were developed and evaluated. These methods have been convincingly investigated for the design of timber buildings at different scales and with different levels of detail and were successfully published in internationally recognized journals. Die Dissertation von Luis Orozco untersucht neue Möglichkeiten für den mehrgeschossigen Holzbau, die aus einem integrativen Ansatz der Entwicklung von computerbasierten Planungsmethoden, digitalen Fertigungsprozessen und neuartigen Bausystemen hervorgeht, der als Co-Design bezeichnet wird. Die Arbeit basiert dabei auf der Forschung zu mehrgeschossigen Holzbauweisen des Exzellenzclusters Integrative Computational Design and Construction (IntCDC) an der Universität Stuttgart, So wird Co-Design von Luis Orozco zunächst aus dem Blickwinkel betrachtet, Interdisziplinarität während des Entwurfsprozesses zu ermöglichen. Darauf aufbauend werden agentenbasierte Methoden erforscht, die verschiedene, oft gegenläufige Entwurfskriterien in einem integrativen und interaktiven Prozess zum Abgleich bringen. Als Grundlage hierfür wurde auch eine Klassifizierung für Holzbausysteme und deren Datenstrukturen erstellt und dann agentenbasierte Modellierungen für die Platzierung von Stützen, die Unterteilung von Deckenplatten und die Anordnung von Schubstegen innerhalb von Hohlkastendecken entwickelt und evaluiert. Diese Methoden werden für den Entwurf von Holzgebäuden in verschiedenen Maßstäben und mit verschiedenen Detailierungsgraden überzeugend untersucht und in international anerkannten Fachzeitschriften erfolgreich publiziert. Professor Achim Menges, AADipl(Hons) Current timber building applications are insufficient to address the global en- vironmental and housing crises. This work conducts research into new mays of working to address these shortcomings. It demonstrates that timber can be used in a broader range of structures if its use is supported by the correct organisational and computational methods. The Introduction frames the work within the current mass timber construction industry. The State of the Art describes current organisational strategies for design, as well as scientific research in the field of agent-based modelling. The research includes a conceptual definition that defines building systems and their data structure. The Developed Methods provides summaries of the experimental implementations of both organisational and computational methods from the included publications. The Conclusion discusses how the work expands the range of applicability of timber buildings. Article A was published as a peer-reviewed research article in the Buildings journal in 2023. Article B was published as a peer-reviewed research article in the Computer Aided Design (CAD) journal in 2023. Article C was published as a peer-reviewed research article in Sustainability journal in 2023. Article D was published as a peer-reviewed research article in the International Journal of Architectural Computing (IJAC) in 2022. Article E was published as a peer-reviewed research article in the Structures Journal in 2023. Article F was published as a peer-reviewed article and presentation in eCAADe 2021: Towards a New, Configurable Architecture in 2021. Article G was published as a peer-reviewed article and presentation at IASS 2023: Integration of Design and Fabrication in 2023. Article H was published as a peer-reviewed article and presentation at the World Conference of Timber Engineering (WCTE) in 2023. The author of this dissertation was the first author of articles A, B, C, D, and F, the second author of publications E and G, and the third author of publication H. The presented work is a cumulative dissertation representing an interdisciplinary research project started in 2019 at the Cluster of Excellence Integrative Computa- tional Design and Construction (IntCDC). The research was conducted within the framework of the research project "Computational Design, Engineering and De- velopment of Digitally Fabricated Multi-Storey Wood Building System", between architecture, structural design, timber engineering, and building physics. The re- search itself sits at the intersection between architecture and computational design and construction. The results are relevant to these fields, and the overall approach provides an example of how architectural design, especially that of multi-storey tim- ber buildings, can benefit from integration with extra-disciplinary knowledge and design simulations. The research was partially supported by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany’s Excellence Strategy – EXC 2120/1 – 390831618. Luis Orozco CO-DESIGN AND AGENT-BASED METHODS FOR MULTI-STOREY WOOD BUILDING SYSTEMS A dissertation approved by the Faculty of Architecture and Urban Planning of the University of Stuttgart for the conferral of the title of Doctor of Engineering Sciences (Dr.-Ing.) Submitted by Luis Orozco from Bogota, Colombia Committee Chair: Professor Achim Menges, AADipl(Hons) Committee member: Professor Jan Knippers, Dr.-Ing. Further committee members: Prof. Dipl.-Ing. Martina Bauer Date of the oral examination: 29.02.2024 Institute for Computational Design and Construction of the University of Stuttgart 2024 CO-DESIGN UND AGENTENBASIERTE METHODEN FÜR MEHRGESCHOSSIGE HOLZBAUSYSTEME Von der Fakultät Architektur und Stadtplanung der Universität Stuttgart zur Erlangung der Würde eines Doktor-Intenieurs (Dr.-Ing.) genehmigte Abhandlung Vorgelegt von Luis Orozco aus Bogota, Kolumbien Hauptberichter: Professor Achim Menges, AADipl(Hons) Mitberichter: Professor Jan Knippers, Dr.-Ing. und weitere Mitberichter: Prof. Dipl.-Ing. Martina Bauer Tag der mündlichen Prüfung: 29.02.2024 Institut für Computerbasiertes Entwerfen und Baufertigung der Universität Stuttgart 2024 Acknowledgements I want to thank Prof. Achim Menges and my friends and colleagues at ICD for the incredible environment I got to work, research, build, and teach in for the last four years. Special thanks to my doktorschwester Anna Krstchil and to Hana Svatoš- Ražnjević, who were with me on every publication and through every deadline, and without whom I would never have written a single word. I am indebted to Felix Amtsberg, Hans Jakob Wagner, Cristóbal Tapia, Theresa Müller, Lorez Riedel, Simon Tremel, Tobias Schwinn, and Long Nguyen for the work they did with me in RP3, and in general. Also to the ITECH Studio 20 and 21 teaching teams, including Simon Bechert and Gregor Neubauer, as well as the ITECH Classes of 21 and 22, for letting me not so much teach as preach for two years, and for supporting my research. Everyone at the ICD and ITKE, as well as the Early Career Board, especially Alya Rapoport, and the whole IntCDC management team made work a social and pleasurable thing to do. Were it not for Profs. Ricardo Castro, Mireille Roddier, Wes McGee, and Karl Daubmann I would never have set out on this particular path, and without Dan Flower, Rafael Petrovic, Jason Harm, Jennifer Kolstad, Paul Ferrer, and Tim Logan I would not have come back to it, in spite of Drs. Paul Crewe and Mark York’s prevailing wisdom. Thanks to John Peel for walking the path with me from afar, and to Philipp Köser for shoving me along when I needed it. I want to thank my friends, who gave me so much and were rewarded by having to put up with me in return. Nick Salthouse, Krissy Massey, Annie Locke Scherer, Olivia Howard, Catherine Ador, Matt Piechowicz, Corey Stone, Elliot Rubin, and Jules Koifman; your indefatigable friendships across time and space kept me sane. Gaby and Alfred Weber, Max Wenzel, Katie Lonson, Lasath Siriwardena, and Tiffany Cheng did so much more than just keep a roof over my head. Käsespätzle nights with Janusch Toppler, Elise Rigolo, Teresa Herzog, Agnes Gambietz, Stine Fischer, Chris Harm, Dasha Biryukova, and Suzie Knuckey, and cooking Yasi Tahouni, Oliver Bucklin, Gina Spiller, and Dylan Wood kept me gastronomically and emotionally well fed. Outings with Anna Trebacz, Christoph Schlopschnat, David Stieler, Regine Rössle, and everyone at StCRC kept me fit. Video games with Oliver and Käthe Gericke, Max Zorn, Grant Galloway, Jojo Seitz, and Aimée Sousa kept me entertained. Above everyone else, I want to thank my dad, Joseph, who as guru, taskmaster, coach, and friend has helped me through everything up to and including this PhD. I simply would not be without you, Felipe, and Laura, and now Susan, Maya, and Owen, to keep me afloat. Finally, Vanessa, for her support in everything in this last year. Luis Orozco Contents Foreword v Acknowledgements xiii List of Abbreviations xix List of Figures xix Abstract xxiii Zusammenfassung xxv Part I Introduction 1 1 Overview 5 1.1 Aim 6 1.2 Structure of Dissertation 6 2 Relevance 9 2.1 Environment 9 2.2 Productivity 11 2.3 Urbanisation 12 2.4 Construction Industry 13 2.5 Building Qualities 15 3 State of the Art 19 3.1 Design Approach 19 3.1.1 Organisational Design Strategies 20 3.1.2 Computational Design Strategies 22 3.2 Timber Construction 25 3.2.1 Building Systems 25 xv Contents 3.2.2 Data Strategies 26 Part II Research 29 4 Conceptual Definition 33 4.1 Systematisation 33 4.1.1 Elements 34 4.1.2 Components 34 4.1.3 Functional Assemblies 35 4.1.4 Construction Assemblies 36 4.2 Building System 38 4.3 Data Structure 40 5 Developed Methods 43 5.1 Organisational Methods 43 5.1.1 Co-Design 43 5.2 Computational Methods 47 5.2.1 Column Arrangement 48 5.2.2 Slab Segmentation 52 5.2.3 Reinforcement Placement 54 6 Research Outcomes 59 Part III Conclusion 63 7 Discussion 65 7.1 Relevance 65 7.2 Conceptual Definition 66 7.3 Organisational Methods 66 7.4 Computational Methods 67 7.5 Research Outcomes 68 8 Conclusion 71 Part IV Articles 75 9 Stakeholder Review 77 10 Segmentation Methods 109 11 Co-Design Methods 129 12 Arrangement of Reinforcement 151 xvi Contents 13 Segmentation Analysis 175 14 Multi-Storey Design Methods 187 15 Interactive Column Optimisation 199 16 Architectural Applications 213 Part V Backmatter 227 Glossary 229 A Supporting Publications with Contribution by the Author 233 A.1 Publication 1 – Buildings Review 234 A.2 Publication 2 – Building System Comparison 236 A.3 Publication 3 – Feedback-Based Column Placement 237 A.4 Publication 4 – Building System Development 239 B Supporting Thesis Projects Advised by the Author 241 B.1 Thesis 1 – Reinforcement Structural Integration 241 B.2 Thesis 2 – Interactive Column Design 241 C Project Credit List 243 C.1 Project 1 – Studio 2019/2020 243 C.2 Project 2 – Studio 2020/2021 244 Bibliography 247 Image Credits 263 Curriculum Vitae 265 xvii List of Figures 1.1 Graphical overview of dissertation, including problem statement, aim, objectives, all included publications, and result. . . . . . . 4 2.1 Timber stakeholders analysis methods and results. From [151] by Orozco (2023). . . . . . . . . . . . . . . . . . . . . . . . . . . 14 2.2 350 Buildings analysis methods and categories. . . . . . . . . . 16 4.1 Graphical overview of the multi-level, multi-agent design method: (a) the field of moment forces in the slab; (b) the agent simulation for placing structural reinforcement; (c) the field of shear forces in the slab; (d) the final segmented slab, with integrated support members in fields and at joints; (e) the results of the structural verification simulation. . . . . . . . . . . . . . . . . . . . . . . 32 4.2 Systematisation of timber building elements. . . . . . . . . . . . 34 4.3 Systematisation of timber building components. . . . . . . . . . 35 4.4 Combinatorial functional assembly types. . . . . . . . . . . . . 36 4.5 Example of a building system broken into functional assemblies. 37 4.6 Example of a building system broken into construction assemblies. 38 4.7 Five categories of building systems, based on reinforcement location. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 4.8 Hollow slab system build-ups. From [101] by Krtschil (2023). . 40 xix List of Figures 5.1 Diagrammatic overview of developed methods, from (a) a systematisation of elements, components, and functional assemblies, through (b) an example of the developed building system composed of construction assemblies [188], and representative diagrams of the computational approaches used in its design: (c) column arrangement [169; 196], (d) slab segmentation [149], and (e) web distribution [148]. Adapted from [188; 169; 196; 149; 148] . . . . . . . . . . . . . . . . . . . . . 42 5.2 Co-design framework for multi-direction, multi-storey timber buildings: (a) material systems use different wood species based on their properties; (b) building systems are multi-directional to allow for long column-free spans; (c) design methods use agent-based models to self-organise components; (d) engineering methods validate the building as it is being designed; (e) fabrication processes assemble components at high tolerances; (f) construction processes use robots to combine construction assemblies on-site. 44 5.3 Schematic rendering of prototype building . . . . . . . . . . . . 45 5.4 Simplified engineering and agent-based design simulation process diagram. From [152] by Orozco (2023). . . . . . . . . . . . . . 46 5.5 Process Workflow Diagram . . . . . . . . . . . . . . . . . . . . 48 5.6 Examples of potential column agent behaviours: (a) maximum, minimum, and target span distances. (b) continuous range of column movement freedom, from fixed to free. (c) target zones. (d) maximum, minimum, and target distances from the edge of the slab environment. (e) collinearity between column agents. (f) columns exclusion zones. . . . . . . . . . . . . . . . . . . . . . 49 5.7 Diagram showing the data flow , including inputs, variables, algorithms, solvers, and outputs, of the developed System Dynamics method. From [169] by Sahin (2023). . . . . . . . . . 50 5.8 Column agent system properties and behaviours. From [196] by Udaykumar (2023). . . . . . . . . . . . . . . . . . . . . . . . . 51 5.9 Computational column arrangement methods applied to Tamedia Office Building plan by Shigeru Ban Architects. . . . . . . . . . 52 5.10 Segmentation Methods, compared on material use, structural performance, and fabrication. . . . . . . . . . . . . . . . . . . . 53 xx List of Figures 5.11 Example floor plans for the comparison of slab segmentation methods. (a) – (d) fictive floor plans with regular outlines and regular column layouts. (e) – (f) fictive floor plans with regular outlines and irregular column layouts. (i) – (j) fictive floor plans with irregular outlines and regular column layouts. (k) Tamedia Office Building in Zurich, CH, by Shigeru Ban Architects (2013). (l) Bjergsted Financial Park in Stavanger, NO, by Helen & Hard and Saaha (2019). (m) Chicago Horizon Pavilion in Chicago, USA, by Ultramoderne (2015). (n), (p) fictive floor plans with irregular outlines and irregular column layouts. (o) Triodos Bank in Driebergen, NL, by RAU (2019). From [101] by Krtschil (2023). 54 5.12 Results of internal structural slab reinforcement methods, using 0.7 m staring web length. From [148] by Orozco (2022). . . . . 56 6.1 Research Outcomes: (a) Campus Lab Prototype Building, (b) – (e) Architectural Case Studies. From [188] by Svatoš-Ražnjević (2023). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 xxi Abstract Timber, the most widely used natural material in the construction industry, can se- quester carbon as it grows and store it in the built environment. However, due to the logistically mandated orthogonal shape of raw and conventionally prefabricated Engineered Wood Products and the limited single spans they can achieve, most contemporary timber buildings currently only have one of a restricted set of uses, with little possibility for reuse. This limits timber’s utility in urban environments, which often require filling in irregular sites and extending existing structures. It also limits the building industry’s potential environmental contributions. These restric- tions could be overcome by using interdisciplinary organisational and computational design methods. This research investigates new ways to design wood buildings through improved cross-discipline collaboration. First, it evaluates co-design as a means of integrating different disciplines throughout the design process by applying it to the design of a prototype building. Then, it proposes agent-based methods for procedurally and interactively negotiating between conflicting sets of optimisation criteria. Agent- based simulations were developed for the placement of columns, the subdivision of floor plates into fabricable and transportable slab segments, and the reinforcement of these segments with internal members. These developments built upon a framework for timber building systems and their data structures. These methods were then tested on the design and fabrication of timber buildings across a range of scales and resolutions. This research demonstrates that innovative computational and organisational methods can results in an increased design space for multi-storey timber buildings. This results in an expanded palette of building types, and an increased contribution by the building sector to the global environmental and humanitarian housing crises. xxiii Zusammenfassung Holz, das am häufigsten in der Bauindustrie verwendete natürliche Material, kann während seines Wachstums Kohlenstoff binden und in der bebauten Umwelt speich- ern. Aufgrund der logistisch bedingten orthogonalen Form von rohen und konven- tionell vorgefertigten Holzwerkstoffen sowie der begrenzten Spannweiten, die sie erreichen können, sind die meisten modernen Holzgebäude derzeit nur für eine be- grenzte Anzahl von Verwendungszwecken geeignet und bieten wenig Möglichkeiten zur Wiederverwendung. Dies schränkt den Nutzen von Holz in städtischen Umge- bungen ein, wo häufig unregelmäßige Flächen aufgefüllt und bestehende Strukturen erweitert werden müssen, und begrenzt somit auch den potenziellen Beitrag des Baugewerbes zum Umweltschutz. Diese Einschränkungen könnten jedoch durch den Einsatz interdisziplinärer organisatorischer und rechnerischer Entwurfsmeth- oden überwunden werden. In dieser Forschungsarbeit werden neue Ansätze zur Gestaltung von Holzge- bäuden durch verbesserte interdisziplinäre Zusammenarbeit erforscht. Zunächst wird Co-Design als Mittel zur Integration verschiedener Disziplinen während des gesamten Entwurfsprozesses bewertet, indem es auf den Entwurf eines Gebäude- prototyps angewendet wird. Anschließend werden agentenbasierte Methoden für die prozedurale und interaktive Aushandlung zwischen widersprüchlichen Opti- mierungskriterien vorgeschlagen. Es wurden agentenbasierte Simulationen für die Platzierung von Stützen, die Unterteilung von Bodenplatten in fabrizierbare und transportierbare Deckensegmente und die Verstärkung dieser Segmente mit internen Elementen entwickelt, basierend auf einem Rahmenwerk für Holzbausysteme und deren Datenstrukturen. Diese Methoden wurden dann für den Entwurf und die Her- stellung von Holzgebäuden in verschiedenen Maßstäben und Auflösungen getestet. Die Forschung zeigt, dass innovative Berechnungs- und Organisationsmethoden zu einem größeren Gestaltungsspielraum für mehrstöckige Holzgebäude führen und xxv Zusammenfassung somit eine erweiterte Palette von Gebäudetypen ermöglichen, was wiederum einen größeren Beitrag des Bausektors zur Bewältigung der globalen ökologischen und humanitären Wohnungsnot leisten könnte. xxvi PART I Introduction 1 Figure 1.1: Graphical overview of dissertation, including problem statement, aim, objectives, all included publications, and result. 1 Overview Timber is a renewable building material that has the potential to store captured carbon. It is comparatively softer and lower density than other common construction materials, and is easy to work with industrially. Building with timber has the poten- tial to make a difference in our warming world, especially because the architecture, engineering, and construction (AEC) industry is such a large contributor of global carbon emissions. Timber’s positive effect can be amplified if a higher proportion of global build- ings are built from timber. With so many people around the world moving to cities and requiring places to live and work, there is further emphasis on the need for responsible construction. However, the proliferation of timber construction is not possible with existing building technology. Most contemporary timber buildings have one of a restricted set of uses. This is due to the limited spans discrete wooden components can achieve when they come directly from standard production, and also to the orthogonal shape of raw and conventionally prefabricated timber elements, which is preferable for international logistics. These restrictions limit the design of timber buildings to a few specific programmes, with little possibility for future adaptive reuse. In order to address the global climate crisis and house the world’s growing urban population, the AEC industry should expand the types of buildings timber is used for, and ensure they are easier to reuse in the future. Increasing the variety of buildings that can be designed in timber will require new design approaches. These methods will need to be interdisciplinary because designing timber buildings requires a deeper integration of different disciplines than other more common building materials. The wide range of required expertise 5 1 Overview results in methods that must optimise complementary and contradictory design ob- jectives. Co-design invovles various stakeholder perspectives and feedback in the development of designs and methods. Agent-based methods are known for their abil- ity to negotiate complexity, interactively producing bottom-up design explorations through emergence. Together, the application of a co-design approach to timber construction, supported by an agent-based modelling (ABM), could be a solution. 1.1 Aim This dissertation investigates new ways of designing timber buildings that may ex- pand their use in a materially responsible and architecturally innovative way, by improving disciplinary collaboration. Its first objective is to evaluate an organisa- tional method that integrates different disciplines into the same design process. Its second objective is to develop modelling methods for architectural part-to-whole and whole-to-part relationships that interface with a variety of multidisciplinary constraints. To achieve the first objective co-design was applied to the development of a timber prototype building. To achieve the second objective, agent-based design tools for multi-storey timber design were developed. These methods integrate constraints from across the AEC industry, such as structural analysis, building physics, life cycle analysis, and fabrication. Conceptual definitions for a common building system and data structure were required to investigate alternative design approaches, and were subsequently developed. The outcomes of both objectives were validated through the design of a series of multi-storey case studies. The organisational methods are evaluated on how well co-design facilitated interdisciplinary interaction, including the quality and efficiency of design decisions. The computational methods are evaluated on how well the agent-based simulations integrated multidisciplinary constraints, including the architectural and environmental performance of their results. The results of the research demonstrate that greater design freedom is achieved through the application of co-design and agent-based methods to multi-storey wood building systems. 1.2 Structure of Dissertation Chapter 1 presents an overview of the dissertation, the problem it addresses, its aims and research objectives, the methods employed to address them, and the evaluation criteria applied. Chapter 2 presents the relevance of timber construction in an 6 1.2 Structure of Dissertation increasingly warm and populated world, and the shortcomings of current timber construction practices to address it, with the support of the publication attached in Chapter 9. Chapter 3 describes the current state of organisational and computational methods in the AEC industry. It also describes current building system and data modelling practices. Chapter 4 explains the conceptual definition necessary to support research into alternate design approaches. It includes a conceptual, generalised definition of tim- ber building systems and their data structure, as well as the definition of an example building system. Chapter 5 contains the work done to address the dissertation’s two objectives. Section 5.1 describes the application of co-design to the design of a pro- totype building with the support of the publication attached in Chapter 11. Section 5.2 describes the development of agent-based design tools for different multi-storey timber functional assemblies. It begins with an overview of the multi-level, multi- agent design system with support from the publication attached in Chapter 14. It is then broken up into 3 sections. Section 5.2.1 investigates the placing of columns in multi-storey buildings, with the support of the publication attached in Chapter 15. Section 5.2.2 investigates the computational segmentation of floor slabs, supported by the publications attached in Chapters 10 and 13. Section 5.2.3 investigates the arrangement of reinforcement within a hollow timber floor, with the support of the publication attached in Chapter 12. Chapter 6 showcases the increased design free- dom afforded by using co-design and agent-based models to design timber buildings using a series of conceptual case studies that appear in the publication attached in Chapter 16. The accomplishments and limitations of this work are discussed in Chapter 7. Chapter 8 summarises this dissertation’s findings, and proposes future avenues of investigation. This is a cumulative dissertation composed of eight scientific peer-reviewed papers, each representing a specific contribution to the research aim. They are found in Chapters 9-15. 7 2 Relevance Some of the greatest contemporary challenges could be addressed by building more with timber. These include the construction industry’s high consumption of re- sources and its stagnant productivity [36], as well as increased urbanisation [198] caused by global population growth [201]. Wood stores CO2 as it grows [38] and building with it substantially reduces greenhouse gas emissions [75; 108]. Timber structures can be processed off-site and assembled on-site more efficiently than con- crete or steel, resulting in faster and quieter construction. Therefore, timber is well suited for the construction of multi-storey buildings, especially in urban contexts [107]. Mass timber construction has the potential to address environmental and sustainability concerns in the building sector, while also improving its productivity. Although multi-storey timber buildings (MsTBs) have become not only taller but more prevalent since the turn of the millennium [104; 40; 103], they still only make up a fraction of the AEC industry’s output [175]. This is because of how they are designed, which limits their applicability. The timber industry is fragmented and constrained by established design and fabrication methods, resulting in buildings that depend heavily on their grids and are restricted to specific typologies. Inter- disciplinary organisational and computational design methods could extend timber construction’s applicability. 2.1 Environment Increased use of timber in construction could contribute to the AEC industry’s environmental objectives. 9 2 Relevance In order to limit global warming to the levels stated in the 2015 Paris Agree- ment, humanity must dramatically reduce greenhouse gas (GHG) emission. The AEC industry is one of the world’s largest polluters [4; 200; 83; 26]. Life Cycle As- sessment (LCA) is a method for comparing the environmental impact of buildings, from the sourcing of their raw materials, through construction, use, and demolition, until those materials are disposed of [136]. Although older buildings expend most of the carbon in their life spans from heating and cooling [220], the majority of carbon emissions in new buildings come from embodied carbon released during their construction and disposal [164]. Concrete and steel, the two most common building materials of the past century, are largely to blame [58]. Concrete is the most polluting building material in the world: the more than 4 billion tonnes of cement produced annually contribute approximately 8% of global CO2 emissions [111]. The quantity of embodied carbon in a structure is strongly coupled with the materials from which that structure is made [136]. Because approximately 80% of the carbon storage capacity of a building is in its structural assembly, it is this load bearing system that should be replaced with wood [38]. Timber is the most common biological material in the construction industry [23] with great mechanical and thermal properties [30]. Substituting timber in place of other materials in buildings has the potential to greatly the reduce embodied carbon [202; 177] and GHG emissions from the construction industry [78]. This is even without considering timber’s ability to sequester carbon [41]. Large amounts of timber substitution are considered possible without adversely affecting agricultural production [135] or global markets [141; 199]. Although there is some debate as to the availability of wood for construction [145], there are currently more trees on earth than at any point since 1982 [182], despite increasing wood use for construction and energy production. Furthermore, a recent study of 65 countries showed that 66% of them had the capacity to harvest more timber than they currently do, based on how much they grow each year [53], and that this increase in production would cover a scenario in which 90% of all buildings are built with timber [38]. Increasing timber’s use in construction is therefore not only possible, but de- sirable. Increasing the proportion of global construction made with timber would therefore address the construction industry’s high resource consumption. 10 2.2 Productivity 2.2 Productivity Increasing the use of timber in construction could be part of alleviating the con- struction industry’s stagnant productivity. Timber can be used in construction in two main ways: either as raw or nominal lumber [28], or as a source of fibre for Engineered Wood Products (EWPs) [143]. Stick frame construction uses nominal lumber and is common in North American domestic construction. Almost all other timber buildings worldwide use EWPs and are designed for prefabrication. EWPs combine wood fibre elements and include everything from fibreboard to adhesively bonded structural I-Beams [134]. En- gineered Timber Products (ETPs) are a subset of EWPs specifically for structural use. Although there are mechanically bonded ETPs, such as nail laminated timber (NLT), and adhesive free products, such as dowel laminated timber (DLT), they are less performative and available then glue-bonded products. Cross laminated timber (CLT) and glue laminated timber (Glulam) are some of the common options on the market, while laminated veneer lumber (LVL) is perhaps the most performative [81]. ETPs give wood the ability to customise its structural performance, regardless of scale or typology, simply by using fabrication techniques [132] and material engineering [71]. Timber is relatively low density, easy to machine, and suitable for prefabrication. Studies indicate that increasing prefabrication in construction can significantly boost efficiency in the building sector [6]. This is important because the construction industry has the least digitisation, even behind forestry and agriculture [12]. It also offers numerous other benefits over conventional construction methods, such as reduced labour, time, cost, and environmental impacts [80] and improved efficiency and quality of construction [168; 123]. Prefabrication can be automated to further increase productivity. Robotic timber construction, which refers to the additive manufacturing of wooden structures with industrial robots, offers various advantages, including the fully autonomous construction of non-standard timber structures [216]. There are several approaches to prefabrication based on where the work is done: (i) off-site, which enables flexible automation [207] (ii) on-site, which can take advantage of flexible fabrication platform [209] (iii) in-situ, on the construction site [208] Many of the advantages of prefabrication come from shifting labour from complex 11 2 Relevance and distributed construction sites to temperature-controlled and situated production halls. Therefore, off-site prefabrication is often the preferred approach. Svatoš-Ražnjević et al. [187] classified timber buildings’ structures into three categories: (a) linear elements, like columns and beams (b) planar elements, such as load bearing walls and floor cassettes (c) volumetric modules, boxes which can even come finished and furnished These categories define what is being prefabricated. Prefabricated construction as- semblies, called modules, are assembled from components and then combined into a building [105]. Module sizes are restricted by logistics and transport constraints [82]. Although the MsTB industry worldwide has invested heavily in their devel- opment and marketing, volumetric modular design strategies are only optimal for 12.5% of all buildings [16]. Furthermore, volumetric modules must be optimised to fit on trucks instead of for environmental or structural performance, or for human comfort and spatial qualities [140]. Some approaches to modularity attempt to over- come this maximum size restriction by adopting the rigid standardisation of small, universal, modular construction units [39]. This "discrete architecture" is outside the scope of this research’s interest. Using timber as a construction material allows for increased prefabrication and automation. These are two ways of improving the low levels of productivity and digitisation in the building industry. 2.3 Urbanisation The widespread adoption of timber construction will be necessary to address the global population crisis sustainably. The world is urbanising at an incredible pace [1], and housing is needed where people already gather to live and work: cities. Although most population growth will occur in Asia and Africa, due to migration, the populations of all countries will increase [72]. Whereas some countries, such as Germany, have robust markets for timber construction [211], others, such as India and those in west Africa, do not yet have sustainable forest management practices in place [198]. This presents an opportunity to leverage lessons learnt in developed markets and apply them elsewhere. So far, increased urbanisation has forced European cities to expand 12 2.4 Construction Industry and disperse [89]. Urban sprawl has adverse environmental and social effects that should be avoided if possible [116]. More compact, efficient, and sustainable urban development models that can accommodate population growth and reduce the environmental footprint of cities are needed. Timber construction can address some of the challenges of urbanisation, especially if used to design flexible buildings. Timber’s high strength-to-weight ratio and low density [57], make it ideal for gravity-load-resisting systems of tall buildings [161]. Because timber is so light- weight compared to other construction materials, more sites are viable for timber construction, including those with low load bearing soils and the vertical extension of existing buildings [46; 58]. This can increase the density and diversity of urban areas, as well as provide additional sources of revenue for building owners. Altern- atively, less foundation material, usually concrete, could be used for buildings of comparable size, reducing material consumption and GHG generation from planned structures. There are many benefits to living in wood buildings, including cognitive, emo- tional, and even behavioural [215]. The biolphilic effects of living in wooden buildings are said to provide the inhabitants with a better quality of life [91]. Not only do patients in wood environments recover faster than in others[54], they feel better there as well [181]. The physical and affective properties of timber make it ideal for urban construc- tion. Timber construction can offer a sustainable and innovative solution to the challenges of urbanisation and population growth. 2.4 Construction Industry Increasing the number and variety of timber buildings requires new design ap- proaches. However, the timber industry is not currently suited to collaborative and integrative design methods. Orozco et al. [151], attached in Chapter 9, carried out a detailed comparative study on the almost 650 construction stakeholders who participated in the construc- tion of 300 buildings built between 2000 and 2021 [186]. This study showed the interconnectivity between global stakeholders and produced a new classification of these stakeholders based on their perceived level of expertise in timber construction. Figure 2.1 shows the inputs and results of the analysis methodology. By analysing network graphs of the MsTB construction industry over the last 20 years, Orozco et al. [151] found that the timber construction industry is fragmented 13 2 Relevance Figure 2.1: Timber stakeholders analysis methods and results. From [151] by Orozco (2023). and local in nature, restricting innovation. These graphs were made using the relationships between stakeholders and their projects. Although some global regions, such as North America, have a strong timber construction tradition, others do not. There are significant differences in timber construction adoption rates throughout Europe. Much of the use of timber is centred in the DACH region, which consists of Germany, Austria, and Switzerland [151], and is therefore currently a research and development hub for timber construction techniques and technologies. Scandinavia has seen much growth in the application of timber for multi-storey buildings since governments started lifting restrictions on timber construction in the 1990s [165]. There have even been positive trends in southwestern Europe (SUDOE), which consists of Portugal, Spain, and Southern France [14], despite their relative lack of forests and subsequent lack of a history of domestic timber construction. Fortunately, as in other global regions, including Australia and Oceania, the climate imperative has led to increased timber construction [187]. Adoption rates are related to the location of timber expertise. A stakeholder’s level of expertise in timber construction is a qualitative and interpretive metric that is influenced by their projects, their engagement with new products, and their level of digitisation [151]. Orozco et al. [151] found that fabricators often have higher levels of expertise. This may be because in order to manufacture construction as- semblies, they are required to have the most multidisciplinary processes, or because the manufacturing process is often digital, and they have greater experience with computational tools. Regardless, they lead the industry by example, with more 14 2.5 Building Qualities projects and more connections to other stakeholders. The disconnected global nature of the timber construction industry reveals a need for more transdisciplinary organisational design methods. The importance of stakeholders with high levels of expertise supports the need for multidisciplinary collaboration and for greater adoption of advanced computational methods. Or- ganisational and computational design methods that integrate different disciplines could help the timber construction industry increase the range of buildings that can be built with timber. 2.5 Building Qualities The most publicised timber buildings from the last twenty years are representative of the forefront of timber construction, yet they still succumb to the same restricted uses and structural systems. New design approaches are needed to change this status quo. Svatoš-Ražnjević et al. [187], attached in Section A.1, collected and analysed 350 buildings published between the years 2000 and 2022 three storeys or taller. Projects were classified according to their structure, material, program, massing, and ordering system to identify their architectural qualities. Figure 2.2 shows the detailed categories of this classification. Svatoš-Ražnjević et al. [187] found that most buildings have a strong grid regard- less of building system, and that most massings had almost no variation in section. However, not all building sites are rectangular, nor are the grids underlying these rectangular buildings all the same, even though most buildings are rectangular for reasons that include the ease of combining spaces [184]. Although often derided as merely expressive intention, irregular building forms are a response to uneven programmatic needs and site conditions [77]. The rigidity and regularity of standard timber buildings render them unable to adjust to the complex boundary conditions of restricted urban environments, where land is at a premium. Strong grids are a reflec- tion of timber’s anisotropic nature, which means that its properties vary depending on the direction of the wood grain. It is therefore difficult to design and distribute slab stiffness in buildings [212], which may reduce the variety of buildings to which simple timber slabs are applicable. Anisotropy, however, creates an opportunity for structural optimisation at a material level, especially with computational methods [127; 129]. Timber is often considered suitable only for certain types of buildings, such 15 2 Relevance Figure 2.2: 350 Buildings analysis methods and categories. as schools or residential buildings [187], an idea reinforced by examples found in timber design manuals [55; 204]. This may be partially explained by timber’s poor acoustic behaviour especially at low frequencies [138] and with impact sound transmission [31], which can affect the comfort and satisfaction of occupants. Fur- thermore, wood’s natural inconsistencies make it difficult to predict the acoustic performance of timber buildings with simulation software [88]. These acoustic and structural simulation issues could be addressed through the integration of more mul- tidisciplinary approaches, and with the application of more robust computational tools. Incorrect building layout, resulting from restrictive column grids, is often quoted as a deciding factor in demolishing or re-building existing buildings instead of re- using them [29] alongside low-quality construction and low performing building 16 2.5 Building Qualities fabric. Buildings are often demolished prematurely because either they are "func- tionally obsolete" (the desired use for the plot of land does not fit into the existing building) [29], or they are no longer perceived to be valuable [95]. "Value loss in the building industry," when a building loses value because it can no longer be used for its intended use [38], is a leading cause of building demolition, which is parallel to functional obsolescence [106]. This is a problem because construction is the most carbon intensive part of a building’s life. Therefore, reusing buildings or mixing uses within a single building can save a significant amount of embodied carbon and energy. Reusing existing buildings has a wide variety of benefits, including: reduced disruption; lower material, transport, and energy consumption; less material waste; and reduced resource consumption [29]. Although many stakeholders emphasise the importance of sustainable adaptive reuse of buildings, most example projects are warehouses transformed into offices and apartments [29]. These are examples of what Arge & Landstad [5] defines as "flexible" buildings, with a homogeneity of spaces and a "generality" of space. The problem with large open spaces is that they often do not have enough building systems to be effectively partitioned or used [93]. This problem can be addressed by designing flexible buildings. Design for flexibility is the process of influencing design factors that determine the ability of the building to adapt to change [188]. Process flexibility, is the capacity of a system or a (design) process to adapt to change and disruption [188]. Because the "longevity of buildings is often determined by how well they can absorb new services technology" [24], interchanging and removing components is important for building use conver- sion [213]. Design for flexibility and design process flexibility are both needed to make adaptable buildings [67]. Incorporating various transdisciplinary viewpoints into the building design process, and negotiating between them using agent-based methods may be a solution. As timber building systems become more mature and efficient, spatial design innovation, such as flexibility, will be the key driver for the differentiation and marketing of timber buildings in the future [104]. This can be achieved through the use of transdisciplinary organisational and computational design methods. 17 3 State of the Art The disconnect between what it is technically possible to build, and what actually gets built is an institutional problem in architectural design, which uses many conflicting and incomplete design practices, ranging from organisational to computational. One way to address this problem is to use co-design, a collaborative design process that involves multiple stakeholders and disciplines in the early phases of design. The application of co-design in the conceptual phases of architectural design stands to make a marked difference in how and which disciplines can play a role in affecting a building’s outcome. ABM is a computational design approach that simulates the behaviour and interactions of autonomous agents in complex systems. When building components are modelled as agents, they can interpolate between diverse design criteria and self-organise into performative arrangements. Both co-design and ABM rely on pre-established concepts of building systems and data structures. Co-design and ABM can complement each other in addressing the challenges and opportunities of timber construction, the limited use of timber buildings. 3.1 Design Approach Buildings are designed using both organisational and computational methods. Com- putational methods are the manual, technical tasks that produce a building. These include drawings, modelling, and data management. Organisational methods organ- ise the building and its conception. These include team compositions, contracts, and collaborative methods. New design approaches will be needed to increase the variety of buildings that can be designed in timber. 19 3 State of the Art Despite the fact that many people, including most Americans, already live in wooded houses [61], timber is not widely used for high-rise or non-residential buildings around the world [76]. Part of the global construction industry’s hesitation towards timber construction may be caused by its familiarity with concrete and steel, which have become institutionally locked [122]. Another may be due to perceived but unfounded concerns with the material itself. Xia et al. [218] suggest that non- technical obstacles to the wider adoption of MsTBs include the design community’s limited awareness of emerging and available timber technologies, along with a perceived yet unjustified risk of fire. This lack of awareness undermines the fact that many of what were previously considered the material’s greatest weaknesses, such as fire performance, fire resistance, and sound transmission, have been thoroughly explored and can now be considered to be mostly resolved [110; 51]. Espinoza et al. [51] interviewed stakeholders, mostly researchers, engineers, and educators, to deduce that structural performance, connections, and moisture performance were the most pressing areas of study. Lehmann & Kremer [110] produced a list of topics by their mention in the discussion sections of recently published literature. This list included: • Collaboration; Education and Awareness; Workforce Skills • Technical Innovation and New Products • Building Physics Improving the Durability (moisture), Dimensional Stability, and Longevity of Wooden Structures • Architectural and Structural System Design for Disassembly and Modularity Both these sets of concerns could be addressed through improved design and simula- tion methods, such as those investigated in this research. The following subsections describe the current state of organisational and computational methods in the AEC industry. 3.1.1 Organisational Design Strategies Organisational design processes describe how work systems are designed. They are often linear, and retain disciplinary silos during teamwork. These issues could be addressed through the implementation of co-design. Architectural design is an inherently collaborative act, with experts from differ- ent fields coming together to work on a single project. This is commonly a linear 20 3.1 Design Approach process, where decisions are made and information is passed down a "digital chain" from conceptual design to construction and facility management [47]. Changes are costly if the need becomes apparent further down the chain. One of the challenges of architectural design is to integrate the perspectives and preferences of various stake- holders, such as clients, engineers, contractors, and users, in the conceptual design phase of a project. The concept of the "extended digital chain", where the design is analysed at every hand-over, aims to increase participation across disciplines, create smaller feedback loops, and reduce the need for costly changes [189]. Regardless, "various stakeholders have identified a need for a better integrative design strategy" [79]. A design process like co-design, that includes interdisciplinary criteria not as a check, but as a driver for each design phase, would address these shortcomings. Organisational strategies are the tools the AEC industry uses to coordinate and collaborate with its experts contractually and operationally. Alternative strategies often come from other fields, such as business, manufacturing, aerospace, or auto- motive. Some, like Multidisciplinary Design Optimisation (MDO) methods are seen as a way to increase the agency of different design stakeholders to affect building performance and cost in the conceptual design phase of a project [65]. MDO, how- ever, simply reinforces disciplinary boundaries by breaking down design problems and distributing them [56]. Others aim to move beyond multidisciplinary work, where different specialists work on a problem without challenging their disciplin- ary boundaries, to more interdisciplinary work, with common methodologies and reciprocal interaction between disciplines [37]. Two prevalent approaches in con- temporary project delivery are Integrated Project Delivery (IPD) and Lean Project Delivery (LPD) . IPD has been formally codified by the American Institute of Ar- chitects (AIA) and fosters collaboration and integration among project participants [92]. LPD does the same, but applies lean principles and operational systems to do so [11], which originate from the idea of "flow" in production [131]. Some of the benefits of IPD include early collaboration, shared risks and benefits, integrated leadership, and a shared software platform [160]. Subjective studies place IPD above competing project delivery methods, but quantitative studies are not so definitive [7]. Regardless, both IPD and LDP have the same overarching goals [114]: 1. genuine collaboration throughout design, planning, and execution 2. increased relatedness among all project participants 3. commitment networks 21 3 State of the Art 4. project, not piece, optimisation 5. tightly coupling action with learning Regardless of the operational strategy used, engineers are seldom given equal pos- itions, even in the most prestigious projects [120]. Better organisational strategies, such as co-design, aim instead for transdisciplinarity, involving experts and non- experts to transcend disciplinary boundaries. Co-design is an alternative and novel organisational strategy to overcome dis- ciplinary boundaries [94]. It uses Direct Feedback, in the form of numerical data passed from one digital design process to another, and Curated Feedback, exchanged orally between collaborators [207]. Direct Feedback makes co-design an example of process flexibility, where a design process can adjust to accommodate changes in inputs during the design process [42]. Unlike IPD and LPD, that focus on project delivery, co-design focuses on technology development. It has been used to simul- taneously develop digital methods and buildings [66; 15], especially those that have deep and direct connections between the fabrication and performance of the built artefact [130]. The development of digital methods sets co-design apart from other organisational methods. Organisational methods, including IPD, can be improved through the use of digital tools [158]. These are often used for "computerisation", which uses digital tools to perform previously non-digital tasks, with the goal of achieving greater efficiency through automation. Alternatively, co-design’s focus on simultaneous digital method development allows for the digitalisation of building design. Digitalisation is the use of digital technologies to produce fundamentally new outcomes that were not previously possible or conceivable [128]. Co-design and digitalisation, through "the linking of methods, processes and systems, also lays the foundation for behavior-based, sensor-driven, cyberphysical manufacturing and construction" [94], and therefore an improved design process for timber buildings. Co-design offers an alternative to the linear and multidisciplinary organisational strategies used in the AEC industry today. 3.1.2 Computational Design Strategies Computational design strategies increase the digitisation of the AEC industry. There are many computational methods than can and have been used to design buildings. Some are interactive, some are fast, and some can handle diverse criteria. However, only ABM can satisfy all the requirements for transdisciplinary design. Parametric modelling is a common technique in architectural design [153]. Most 22 3.1 Design Approach applications of parametric modelling are design procedures, in which parameterized components can be constantly changed and manipulated to explore design spaces [13]. Unlike ABM, parametric modelling is completely deterministic, and requires external validation and simulation. Generative design is used to explore the design solution spaces afforded by parametric models [98]. In generative design, designers choose from design solu- tions generated from within the solution space. However, these solutions need to be manually recompiled or changed if they do not meet the desired requirements. Conversely, ABM simulations can be interactive and adjusted live if the process is not yielding desirable results. Currently, AI tools in architecture are increasingly proposed in an attempt to simplify the design process [35]. However, for most designers, AI tools and other computer science techniques, such as Multi-objective Optimisation (MOO) [217], are black boxes that only expose you to one solution without allowing designer interaction. These methods often yield optimal mathematical solutions, but do not account for additional data, knowledge, and sensitivities of the designer. Optimisa- tion such as this is useful for specific technical design studies. However, it is not suited to larger collaborative design as it can remove agency from designers. ABM is a method that simulates the behaviours and interactions of individual agents within complex adaptive systems, [117] codified in engineering standards [206]. ABM’s defining quality in the design realm is its ability to harness self- organisation (as part of a swarm intelligence approach [43]) and emergence for the exploration of complex design models. Agent-based Models (ABMs) are built on the fundamental modelling construct of agents and their behaviours [118]. An agent’s behaviours affect its own properties, those of the other agents in the agent system, and their shared environment. Other terms for autonomous self-organising systems that use individual agents exist, including Multi-agent Systems (MAS) [155] and Complex Adaptive Systems [34]. According to Carmichael [33], MASs are typically more complex than CASs, as they involve higher levels of communication, coordin- ation, and cooperation between agents. Agent-based simulations are increasingly popular in pedagogic settings [119], where it is often contrasted with Equation- based Modelling (EBM)[205] approaches such as System Dynamics. Whereas EBM is best suited to capturing the equilibrium of systems with central interactions and continuous dynamics, ABM is capable of demonstrating emergence and self- organisation in complex systems. This is different from Discrete-event Simulations (DES) [147], whose interactions occur at semantically significant moments [197]. 23 3 State of the Art Dynamic relaxation is a common alternative technique for self-organising building components [32]. Some implementations of dynamic relaxation use projections onto the zero-energy state of each constraint to improve the convergence rate [22]. Dynamic relaxation has been used to co-design column arrangements by Sahin et al. [169]. However, this approach offered fewer goal options and customisation when compared to ABMs with the same objective, such as those by Udaykumar et al. [196]. ABM has been used extensively in architecture, where it can reduce the se- mantic difference between what is modelled and how it is modelled. The work of this dissertation conceptualises the elements of its spatial systems as agents [185]. Previous examples of agent systems that are used to organise building components in such a way include Groenewolt et al. [70]. Their work served as a starting point for the methods described in Section 5.2.2. The work of Pantazis & Gerber [156] has a similar aim to that of this dissertation, in that it uses an ABM informed by interdisciplinary concerns. In their case, they are mostly concerned with structural and environmental performance, as their ABMs model building facades, whereas the methods in Section 5.2 focus on structural and fabrication performance in the design of building structures. The approach of Gerber & Pantazis [63] to moving the fab- rication information upstream involves simulating the robotic construction process and assigning it a constructability score based on estimated fabrication duration and number of potential collisions. The results of the constructability function were then passed back as parameters to the "generative agent" that instantiates their system, thus reducing the probability that difficult-to-construct panels are generated the next time the agent system is run. Conversely, the methods in Section 5.2.2 receive their fabrication feedback through both Direct and Curated Feedback. Gerber & Pantazis [64] use Kangroo [157], Karamba [159], and Ladybug [166] to inform their MAS, which applied solar path information to their existing structural geometry. By ana- lysing each of the joints between their structural members locally and the whole structure globally, they could locally adjust the length and thickness of elements to improve solar and structural performance. However, they are still lacking methods for feeding back structural analysis results as input parameters for future iterations of the agent simulation, such as those described by Sahin et al. [169] or by Orozco et al. [150] in Section 5.2. Hamidavi et al. [74] propose a method to place columns within the walls of an early-phase architectural design. However, their method requires a fixed spatial separation between the columns, which limits the design flexibility. Moreover, while their method allows for the structure to be optimised, it does not 24 3.2 Timber Construction provide structural feedback to the architectural design as Udaykumar et al. [196] do in Section 5.2.1. ABMs are powerful simulations that can negotiate transdisciplinary constraints while modelling architectural part-to-whole and whole-to-part relationships through emergence. 3.2 Timber Construction Effective co-design using ABM relies on building systems and data structures that can also handle the transdisciplinary requirements of both organisational and com- putational design approaches. The building system must exhibit process adaptation to incorporate constant developments from all project partners. It must be supported by a data structure robust enough to ensure the same. 3.2.1 Building Systems The design of contemporary mass timber buildings is based on the development of their building systems. Building systems describe the materials, components, and connections that make up a structural load bearing system. It is in the design of a building system that decisions, such as which of the many available floor or vertical support systems, will be used in the building to be designed [99]. Some architect engineers, such as Michael Green, often develop new building systems for their buildings [69]. This allows them to maximise the structural, architectural, and environmental potential of each building and its unique site. Other companies have developed a single building system and deploy it across multiple projects, such as the "Platforms for Life" by Krieg & Lang [97]. In many of these building systems, non-wood materials are used to solve technical challenges or when greater design freedom is needed [187]. Building systems that embody aspects of both process flexibility and design flexibility exist on the market, most notably the TS3 system [222; 221] and the Rothoblaas system [121]. The development of these building systems focused on the joints between components. The Rothoblaas system focuses on steel connections between plates and between plates and columns [125]. The TS3 system uses only mass timber and glue [60]. Although originally only for solid floor systems, they have recently been looking at applying the TS3 system to hollow floor systems [59]. The use of steel parts in the Rothoblaas system undermines the advantages of timber construction, as steel has different thermal, moisture, and fire properties 25 3 State of the Art than wood and can cause problems such as thermal bridging, corrosion, and fire spread. A major limitation of the TS3 timber building system is its reliance on hard- to-maintain and expensive butt-joint bonding technology that requires specialised equipment and skills, as well as compatible timber elements that are not widely produced or distributed. However, TS3 has a computational design tool that helps in the specification and design of buildings with their system [195]. Unfortunately, both the building system and the design tool lack the capacity for design flexibility. In contrast, the system described by Mureşan et al. [139] has a linked structural system and design process that is not restricted by the dimensions or configuration of floor modules or column arrangements. The system also allows for the adjustment the slab density according to structural criteria. However, performance and CO2 emission concerns limit the applicability of the system to temporary and high reuse scenarios [52]. While there are many high performing timber building systems on the market, their properties are often not conductive to process or design flexibility. This contributes to inapplicability to many building types. The systematisation of timber buildings described in Section 4.1 supports the development of a flexible building system for ABM and co-design in Section 4.2. 3.2.2 Data Strategies Operational strategies rely on digital structures. The linear "digital chain" design processes can be rooted in the design tools they use [158]. The most common digital paradigm in the construction industry today is that of Building Information Modelling (BIM) [85]. BIMs consist of geometric models coupled with semantic data. This combination allows for, among other things, the digital representation of real-world objects, with up-to-date information, known as "digital twins" [62]. Despite its utility and ubiquity, BIM can struggle with complexity in the design process. This difficulty is due in large part to the ontological model underlying BIM’s central data schema: IFC (Industry Foundation Classes)[87]. However, the problems also came from software: Computer-aided design (CAD) platforms like Autodesk Revit [10] and ArchiCAD [68], and their unique interpretations of IFC. The most faithful implementation of the IFC standard is xBIM, but it still reflects one developer’s interpretation as an open source project [115]. In response to these challenges, some researchers have explored the use of parametric and adaptable BIM models during the design process [20]. Alternate data strategies are necessary, especially to transfer data between disciplines as part of co-design. 26 3.2 Timber Construction Information exchange in the AEC domain is challenging due to the participation of multiple disciplines, organisations, and formats that often have incompatible data and standards [171]. A thorough study of data interoperability in AEC by Elshani et al. [49] demonstrated the weaknesses of data schemas such as IFC, especially from the perspective of highly interdisciplinary design strategies such as co-design. Elshani et al. [49] argue that BHoM, the Buildings and Habitats object Model [17], an open source platform for data federation, could serve as a foundation on which to build a model with different disciplinary ontologies that all connect and relate to each other. Hierarchical graph-based data structures are proposed as a suitable solution to represent and manage complex and heterogeneous information [179]. Current research suggests that graph-based modelling is a resource-efficient and transparent way of integrating different knowledge scopes within the same digital model and semantic information [50; 124]. This suggests that graph-based data structures implemented in data ontologies such as BHoM have the potential to improve data interoperability and enable more collaborative and interdisciplinary design processes in the AEC domain. A robust, graph-based data model could support the expansion of uses of tim- ber buildings by supporting collaborative organisational methods and interactive, transdisciplinary computational methods. The systematisation of timber buildings described in Section 4.1 lays out the framework of the data structure for ABM in co-design outlined in Section 4.3. 27 PART II Research 29 Figure 4.1: Graphical overview of the multi-level, multi-agent design method: (a) the field of moment forces in the slab; (b) the agent simulation for placing structural reinforcement; (c) the field of shear forces in the slab; (d) the final segmented slab, with integrated support members in fields and at joints; (e) the results of the structural verification simulation. 4 Conceptual Definition Increasing the variety of buildings to which timber construction is applied using co-design and ABM methods requires building systems and data structures that are built for interdisciplinary design. A thorough understanding of timber buildings is a prerequisite for their development. Section 4.1 describes the systematisation of timber buildings into their elements, components, and assemblies. Sections 4.2 and 4.3 explain the building system and data structure derived from this systematisation. 4.1 Systematisation This is the process of organising and standardising the information, components, and methods involved in the design, construction, or management of a project or system. This research used the comprehensive systematisation of the components within timber buildings, ranging from linear or sheet-based lumber to complex column-through-plate connections with integrated reinforcement as its conceptual definition. This resulted in the formal definition of the categories, properties, and relationships between the pieces of a timber building, with an emphasis on design and fabrication efficiency. This systematisation can be used to identify and categorise the constituent assemblies and components of all timber building systems. This serves as a theoretical framework for the practice of building system design, including its underlying data structure. On the basis of this analysis, it was evident that the placement of columns, the subdivision of floor plates, and the arrangement of reinforcement would benefit the most from multidisciplinary computational modelling methods. 33 4 Conceptual Definition 4.1.1 Elements This systematisation builds on the vocabulary of technical product terms [84]. Ele- ments (bestandteilen) are the constituent parts of components (bauteilen) and the smallest pieces of the systematisation. Because a component "cannot be physically divided into smaller parts without losing its character" [84, p. 2.14], elements are generic planar or linear wood elements, such as sheets of veneer, or the individual lamella of CLT (Figure 4.2). Figure 4.2: Systematisation of timber building elements. 4.1.2 Components Components are the individual pieces from which the timber building is made. They are the pieces that fabricators order from manufacturers. They are made from at least one element and, like elements, they are either linear or planar. A piece of solid roundwood could be a component, as could a piece of nominal lumber. Components also include sheets of CLT, made from multiple linear elements, its lamellae, or blocks of LVL, made from many planar elements, the veneer layers. Components are differentiated by being vertically or horizontally orientated (Figure 4.3). This orientation roughly aligns them with the base structural architectural pieces of buildings: • horizontal plates (HP) are analogous to floors and roofs • horizontal linear (HL) are analogous to beams • vertical plates (VP) are analogous to walls • vertical linear (VL) are analogous to columns Although this analogy helps in building system decision making, not all components used in a building are architectural pieces. Functional assemblies can sometimes re- quire the inclusion of additional components which are not analogous to architectural pieces. 34 4.1 Systematisation Figure 4.3: Systematisation of timber building components. 4.1.3 Functional Assemblies A single component can serve multiple purposes within a building. Depending on how it is designed and detailed, a single sheet of CLT used as an interior partition may interface with the floor below it, a beam above it, an adjacent column, and another intersecting wall. It is therefore said to be part of multiple functional assem- blies (funktionsbaugruppen). Functional assemblies are the pieces that make up a building system, and are the building blocks of building system conceptualisation. Like structural members, the joints are crucial for the structural performance, the fabrication efficiency, and the architectural expression of the building [113]. Walls and floors are therefore functional assemblies, as are the connections between com- ponents. The definition of functional assemblies aids in the synthesis of building system ideas. The four component classifications: HP, HL, VP, and VL; make a return, this time standing in for their analogous architectural pieces. Although any number of these may be present in a given building system, the vast majority of building systems include a floor assembly and a vertical load bearing assembly. Based on the four types of architectural functional assemblies, there are ten types of joint functional assemblies, shown diagrammatically in Figure 4.4. Figure 4.5 shows a diagrammatic section of the structural elements of a concep- tual building system. This building system uses columns (VL, blue) and ribbed floor plates, consisting of a top layer of CLT (HP, yellow) and beams below (HL, purple). The non-load bearing internal partition walls (VP) are not shown. In this scenario, 35 4 Conceptual Definition Figure 4.4: Combinatorial functional assembly types. the column serves as a representative example of functional assemblies. The column functional assembly consists of a single VL component, a piece of softwood glulam. The column functional assembly is also part of three other functional assemblies: column-to-beam connection (HL-VL), column-to-plate connection (HP-VL), and column-to-column connection (VL-VL). The column-to-column joint must transfer vertical forces without crushing the perpendicular fibres of the floor plate. They are connected using two additional non-analogous components: an intermediate pin (VL) and pyramid-shaped inserts (VL), both out of beech LVL, as described by Tapia & Aicher [192]. The columns and beams are sufficiently coupled by a glued half-lap joint. These would usually be CNC milled into the column functional assembly it- self, but because of its addition they are now milled into the pyramid-shaped insert. The column-to-plate connection requires two additional, non-analogous compon- ents: a top insert (HP) that "reinforces the tensile stressed region of the continuous CLT plate" and a bottom insert (HP) that both stiffens the CLT where it meets the column and prevents high rolling shear concentrations there, as described by [191]. 4.1.4 Construction Assemblies Although components pertain to functional assemblies during building system design, they eventually need to arrive on site and be assembled into a building. Fab- rication takes individual components and combines them into construction assem- blies (konstruktionsbaugruppen). Construction assemblies are the physical pieces that are put together on site to make a building. While a single component may be part of several functional assemblies, it can only be part of a single construction assembly. Construction assemblies consist of at least one component. In manual construction practices such as stick-frame construction, each piece of nominal lum- ber is a construction assembly by itself, although many pieces together form a wall 36 4.1 Systematisation Figure 4.5: Example of a building system broken into functional assemblies. functional assembly. Prefabrication is central to the concept of construction as- semblies. Construction assemblies are a way of thinking of and simplifying what needs to be produced as part of fabrication planning and how it can be brought to site through logistics. The construction assembly that a component should be part of is decided by the tasks its fabrication requires, the skills that the fabrication setup requires to achieve those tasks, and the ability of the resulting construction assembly to get to the construction site. The concept of construction assemblies brings fabrication considerations into the definition of the building system and the architectural design, ensuring the feasibility of all design decisions. Figure 4.6 shows the same diagrammatic section as Figure 4.5, but divided into construction assemblies. In this example, there are three types of structural assemblies: column, column plate, and spanning plate construction assemblies. There are also extra construction assemblies consisting of single components that are used to join the larger construction assemblies during on-site assembly. All column construction assemblies are identical and contain all components related to the column-to-column functional assembly. Column plate and spanning plate construction assemblies are semantically self-similar, containing the same glulam beams and CLT sheets, but differ by whether they contain a column-to-plate joint. They also have the potential to be geometrically unique based on the design of the building. Mass-customisation of building components is not an issue if CNC processes such as robotic timber construction are used in prefabrication. 37 4 Conceptual Definition Figure 4.6: Example of a building system broken into construction assemblies. This systematisation includes all pieces of timber buildings, their joints, and their networked relationships. Functional assemblies help define the qualities and components in a building system for ABM supported co-design. The relationship between function assemblies and components is the foundation for a graph-based data model. 4.2 Building System A robust systematisation of timber buildings enables the development of a building system suited to feedback-driven design and fabrication upon which to test the efficacy of agent-based methods. This building system was developed by the author within the auspices of the Cluster of Excellence Integrative Computational Design and Construction for Architecture (IntCDC) as a core component of Research Project 3, "Computational Design, Engineering and Development of Digitally Fabricated Multi-Storey Wood Building System". The mono-material wood building system was designed to achieve flexible spans and to accommodate non-orthogonal column configurations with a flat underside of slab. The systematisation enables a system that embodies design flexibility and is adaptable to a wide range of building types, shapes, occupancy loads, and uses. The developed building system is mono-material, using only wood as the main structural material. Because they are composed of "new components and solutions that reduce the use of non-renewable natural resources," mono-material building 38 4.2 Building System systems have several advantages, such as "reducing waste production in the building industry" by reducing the variety of material sources [109]. The mono-material system was achieved by mixing hardwood and softwood elements thereby using each species for its relevant properties and optimising the performance and efficiency of the structure [109]. Selectively using hardwood for specific components, such as joints, makes up for the weaknesses of the surrounding softwood. Deciduous woods, such as European beech (Fagus sylvatica) or Pedunculate oak (Quercus robur), have 1.5 to 3 times greater strength, respectively, than coniferous wood like Norway spruce (Picea abeis). This results in better load resistance and smaller deformations [214]. The recent development of highly performative all-wood joints for point-supported slabs, including column-to-slab [190] and slab-to-slab [193] joints, enable this material mixture. Kaufmann et al. [90] categorise timber building systems into massive, ribbed, and hollow, based on where their reinforcement is located. Figure 4.7 shows the five categories of floor system (HP) investigated in this dissertation. (a) Solid (b) Ribbed (c) Hollow (d) Integrated (e) Reinforced Figure 4.7: Five categories of building systems, based on reinforcement location. The hollow and solid systems were chosen for further study because their flat underside of slab is conducive to the easy reposition of interior partitions. The question of using solid or hollow construction is not uncommon in computational design projects [96]. A prevalent misconception that influences the use of solid CLT plates in flooring systems is the notion that timber buildings should maximise their wood content to enhance their carbon sequestration performance in life cycle analyses [2]. A more sustainable approach to addressing the climate crisis is to minimise the amount of timber used in each building, rather than to maximise it [146]. Krtschil et al. [100] compared these two systems on the basis of their structural performance and prefabrication potential. Like Bissig & Frangi [18], they found that a hollow system was comparatively performative to a solid one, but with less on-site material, in spite of losing more material during the prefabrication process. The hollow system was chosen for further study because of its potential for future improvement. Figure 4.8 shows three potential hollow slab system build ups. Greater slab depth increases 39 4 Conceptual Definition (a) Disconnected solid slab. (b) Conventional solid CLT depth. (c) Increased depth for greater span- ning potential. Figure 4.8: Hollow slab system build-ups. From [101] by Krtschil (2023). acoustic decoupling, and decreases structural deformation and material use. The middle system (Figure 4.8b) was chosen because it is able to meet the deflection and loading requirements of Eurocode 5 [45], the European standard for the design of timber structures, while providing increased architectural possibilities in terms of spatial flexibility, geometric complexity, and aesthetic expression. Krtschil et al. [102] provide a more thorough, technical description of the building system and the destructive tests conducted in its testing. This is attached in Appendix A.4. The proposed building system, with its mono-material joints and hollow slab, addresses the limited spans and orthogonal restrictions impeding timber’s wider adoption as a construction material. The design flexibility of its columns, slab shapes, and internal reinforcement make it ideal for testing this dissertation’s second objective, to develop agent-based design tools for multi-storey timber construction. 4.3 Data Structure The relationships between elements, components, and both functional and construc- tion assemblies defined by the systematisation form the basis for a graph-based data model. As described in Section 3.2.2, IFC, as the backbone of BIM, is the most common data structure for the AEC industry. IFC struggles with "generating, representing, and managing the data needed for digital fabrication processes" [176] because it uses architectural pieces, such as walls, doors, and ceilings, as its smal- lest building blocks. It has pre-assigned features and types for every architectural piece, which may support manual design methodologies but constrict computational ones. Conversely, this dissertation’s timber building systematisation is built using timber elements as its ontological foundation. All component can carry semantic in- formation about its location, use, and required fabrication tasks regardless of which functional assemblies they are part of. Top-level assemblies or their constituent components can modify each other’s shared and unique properties. A horizontal linear component could be a beam, a joint, or part of a hollow slab, but still carry 40 4.3 Data Structure the same information. The data management approach described by Elshani et al. [50], which combines semantic webs and BHoM, enables such a data structure. Most importantly, it demonstrates an alternate method for geometric and semantic building data that can still be communicated with other stakeholders through a more open approach to BIM. A data management approach that combines semantic webs and BHoM can enable more interoperable and collaborative workflows that can integrate different disciplines and stakeholders in the timber building project. This data structure ensures that feedback, direct or curated, from any stakeholder prolif- erates through the digital model, enabling this dissertation’s first objective, to apply co-design to the development of MsTBs. 41 Figure 5.1: Diagrammatic overview of developed methods, from (a) a systematisation of elements, components, and functional assemblies, through (b) an example of the developed building system composed of construction assemblies [188], and representative diagrams of the computational ap- proaches used in its design: (c) column arrangement [169; 196], (d) slab segmentation [149], and (e) web distribution [148]. Adapted from [188; 169; 196; 149; 148] 5 Developed Methods This dissertation seeks to find new ways of designing timber buildings that can improve their material responsibility and architectural innovation by fostering col- laboration among disciplines. The first objective is to assess an organisational method that brings different disciplines together in the same design process. To achieve the first objective, co-design is employed to develop a timber prototype building. The second objective is to create modelling methods that can deal with various multidisciplinary constraints and connect architectural parts and wholes. To achieve the second objective, agent-based tools for MsTB design that incorporate constraints from across the AEC industry were developed. This chapter contains the work done to achieve these two objectives. 5.1 Organisational Methods Co-design is an approach that integrates many design disciplines into the conceptual design process, including architectural design and construction planning. The co- design process is split into six areas which must be developed simultaneously: "[design] and engineering methods, fabrication and construction processes, and materials and building systems" [94]. Figure 5.2 shows a diagrammatic overview of the application of co-design to MsTBs. 5.1.1 Co-Design To achieve the dissertation’s first objective, co-design was applied to the development of a prototype timber building. This application of co-design to building system 43 5 Developed Methods Figure 5.2: Co-design framework for multi-direction, multi-storey timber buildings: (a) material sys- tems use different wood species based on their properties; (b) building systems are multi-directional to allow for long column-free spans; (c) design methods use agent-based models to self-organise components; (d) engineering methods validate the building as it is being designed; (e) fabrication processes assemble components at high tolerances; (f) construction processes use robots to combine construction assemblies on-site. development and building design produced a 37 m2 prototype building called the Campus Lab. Figure 5.3 shows a schematic rendering of the Campus Lab. The Campus Lab is composed of two identical floor plates that can serve as either floor or roof, to demonstrate the building system’s applicability to multi-storey construction. It is raised from the ground to demonstrate the fundamental functional assembly for multi-storey construction: the column-to-column (VL-VL) joint. Many alternative functional assemblies were developed with input from structural design, timber engineering, and fabrication planning during the definition of its building system based on the systematisation (Section 4.1). It contains two different column-to-plate connections (HP-VL), one stiff in moment and the other not, not only to resist lateral forces, but to demonstrate that functional assemblies can be varied without affecting 44 5.1 Organisational Methods the robotic timber construction process. The need for a unified design method with integrated structural and agent simulations was highlighted during the design of the Campus Lab, as the feedback loop between disciplines was too slow before the direct feedback link was developed. Figure 5.3: Schematic rendering of prototype building A hybrid softwood and hardwood material system improves the slab’s spanning length and structural performance. The Campus Lab uses hardwood throughout its slabs, to resist bending and punching shear, as well as for durability in its joints. The building system consists of a hollow slab supported by columns with tailored discrete internal reinforcement. The architectural qualities of this building system, such as a flat ceiling, open corners, integrated services and non-orthogonal column placement, enable adaptability and increase architectural possibilities. The Campus Lab’s design methods included the Sorted Delaunay slab segment- ation method (Figure 5.10d) and the structurally informed ABM web placement method (Figure 5.12c). Figure 5.4 shows a simplified design process diagram of the interactive process described in detail by Orozco et al. [150]. Due to the small 45 5 Developed Methods scale of the building prototype, the columns were placed manually to demonstrate varying span and cantilever conditions that would be challenging for conventional timber construction systems. Figure 5.4: Simplified engineering and agent-based design simulation process diagram. From [152] by Orozco (2023). The engineering methods used a global model to simulate the overall behaviour of the prototype building, linked to a finite element model for the detailed simulation of joints like that described by Krtschil et al. [102] in Appendix A.4. This is an example of multi-scale modelling, in which the finite element model for a single joint is connected to and informs the macro system model [194]. The global model was read and written to by both structural and design methods as part of a multi-level, multi-agent design process. The fabrication process and its constraints guided the design of the prototype building. Direct and curated feedback permitted structural and timber engineers to engage with fabrication. Together, they could verify the design of all components from their unique technical perspectives. The fabrication process employed novel techniques for human-machine collaboration, including Augmented Reality (AR) to help distribute tasks between human and robot actors [3; 219] and machine vision for workpiece calibration. The construction process was simplified by leveraging construction assemblies. The prototype building has only two types of construction assemblies: column construction assemblies and plate construction assemblies, corresponding to the VL and HP component classifications. These construction assemblies were designed to facilitate self-alignment and vertical assembly directions. Thanks to the variability afforded by mass customisation, the fabrication process prefabricated two unique column construction assemblies and four unique plate construction assemblies using the same tasks, skills, and robotic subroutines. Thanks to co-design, parts of the prototype building’s early design phase could be influenced by sociocultural aspects. The building services, location, and amenities 46 5.2 Computational Methods were informed through community engagement. This case study demonstrated the applicability co-design methods to MsTB con- struction. The project successfully produced a building with complex geometry and a high degree of prefabrication, accomplishing the dissertation’s second object- ive of connecting design and fabrication methods into a single digital model and framework. The co-design process achieved the first objective of this dissertation by enabling the integration of structural, architectural, and fabrication constraints, as well as the participation of multiple stakeholders. This is necessary to increase the variety of buildings that can be designed in timber. 5.2 Computational Methods The ABMs developed to address this dissertation’s second objective must accom- modate different functional assemblies and construction methods. The building components that are most common between all possible building system imple- mentations, and simultaneously the most suitable for automation, are the columns, floor plates, and horizontal reinforcement. These components have a significant impact on the structural performance, spatial organisation, and aesthetic expression of the building. However, in order to achieve a holistic and integrated design, this process cannot be isolated from other aspects of the project. Instead, it requires a multi-level, multi-agent design method that can communicate with the building model at different levels of detail and abstraction. Figure 5.5 shows an overview of the design methods and how they communicate with the building model. The design tools that are developed in this work are based on the principles of exploration and interaction rather than strict optimisation, while still automating parts of the design process. They take previous advancements in the integrated design and analysis of complex structures, including plate shells and lattices, and apply them to timber construction [180; 183]. They use techniques that are slower and older than some of the techniques described in Section 3.1.2, but are more transparent and flexible than black-box optimisation and AI tools. They also use their intelligence to support the designer, not to supplant them. The objective of these tools is to allow the designer to modify the massing and have the system (the entire building system, from column placement to joint articulation) adapt accordingly, while also enabling the designer to manipulate elements manually without recomputing the entire model or being constrained by the system. A condensed overview of the multi-level, multi-agent design method is described by Orozco et al. [150], attached 47 5 Developed Methods Figure 5.5: Process Workflow Diagram in Chapter 11 These tools provide real-time feedback and direct design integration, which give the designer a much higher level of control than generative design. The design tools are built on top of the ABxM Core, which is a framework that makes designing and using agent-based models in architecture more accessible [142]. 5.2.1 Column Arrangement Point-wise structural supports, like columns, are better suited for a reconfigurable architecture with a long building life span than load bearing walls. Load bearing walls put an immovable partition within the building and even when made of wood require relatively serious intervention to alter. Columns can be integrated inside any future partition system, regardless of its building physics requirements, and can also be used intentionally within the rooms of a building. In conventional architectural design workflows, designers set and choose a structural grid, which is then applied as consistently as possible over the building site. This grid then governs many aspects of the design, from the shape and size of rooms, and therefore the current and potential use of the building, to the rhythm of the facade, and therefore its expression. Regular grids work well on rectangular sites. However, 48 5.2 Computational Methods many urban sites have irregular shapes or are infill projects that require adaptation to the existing context. A regular grid is not suitable for these sites. This research proposes a network-based approach to structure instead of a grid-based one. A non-orthogonal approach is materially responsible because it reduces the need for extra supports or structural transfers. It also makes for better spaces, by removing the awkward transition zones in them and allowing for the freer placement of non- structural partitions. Timber buildings with these qualities would be more suitable for expanded use. (a) Column Span Distance (b) Column Movement Freedom (c) Target Zones (d) Edge of Slab (e) Minimise Displacement (f) Exclusion Zones Figure 5.6: Examples of potential column agent behaviours: (a) maximum, minimum, and target span distances. (b) continuous range of column movement freedom, from fixed to free. (c) target zones. (d) maximum, minimum, and target distances from the edge of the slab environment. (e) collinearity between column agents. (f) columns exclusion zones. The arrangement of columns into a non-orthogonal network is a complex task that accounts for structural considerations and architectural intention, such as space planning and building services, while allowing direct input from a designer. Existing strategies for the computational arrangement of columns are based on mathematical optimisation and are therefore not interactive or iterative [126]. Scheurer [170] demonstrated that the use of ABM to self-organise columns is a viable way to optimise irregular spatial structures. This problem seemed uniquely suited for discrete-time-based, interactive agent simulations, such as System Dynamics and ABM. 49 5 Developed Methods Figure 5.7: Diagram showing the data flow , including inputs, variables, algorithms, solvers, and outputs, of the developed System Dynamics method. From [169] by Sahin (2023). A collection of behaviours that could govern column agents was defined. These included exclusion zones, where the designer would not want columns, for example to create a large welcoming atrium space, or to avoid an unsupported area of the foundation or substructure (Figure 5.6f). Conversely, target zones are areas where a designer would rather a support be located, for example, adjacent to a service riser or along spatial subdivisions (Figure 5.6c). Some behaviours are structurally informed, with columns always moving towards the nearest area of greatest displacement in the slab, as calculated every iteration of the agent system (Figure 5.6e). Some functional assemblies bring their own requirements, such as a minimum distance from the edge of a floor because of the size of the punching shear resistant reinforcem