Programmable Architecture

-Towards Human Interactive, Cybernetic Architecture-

Kensuke Hotta (B.Eng, M.Eng, Msc)
Architectural Association School of Architecture, 2013

プログラマブル アーキテクチャ


堀田憲祐, 英国建築協会建築学校 

1-1. Introduction

There are two different meanings associated with the word “architecture”. The first relates to “the style and design of a building or buildings” and the other to “the structure of a computer system and the way it works” from the Longman dictionary (Ed. Various, 2009). In this thesis, the former is described as ‘architecture (building)’, and is considered the “hardware”. The latter is described as ‘architecture (system) ’and is regarded as the “software”. This proposal will attempt to re-connect these two words because of the upsurge in computational methods, not only as an extension of the designer’s hand - such as drawing software - but also as an extension of the designer’s intelligent, brain-like functions (mind) - such as intelligent controllable tools.

     Programmable Architecture, as discussed in this research, proposes a new strategy for cybernetic architecture defined as a multi-scaled system that communicates with architectural components. The smallest elements are interconnected with humans by devices such as smartphones. Through both software and hardware, it has the ability to change its shapes during its interaction. (Fig1-1-1) The important thing is to ensure that the architectural space (building and system) is somehow controllable by human agents. 

     The core challenge is to make the hardware a lattice of discrete machines (which consist of self-contained components) that integrates numerous computers dedicated to sensing-calculating-actuating, each making its own decisions in order to produce an interactive interface. True adaptability or sustainability of architecture can result from such a system. 


 「アーキテクチャ」という言葉にはふたつの意味がある。ひとつは構造物や建築物のスタイルやデザインに関連するもの、そしてもうひとつは、コンピュータシステムの構造とその働きである。(ロングマン現代英英辞典, ,2009年版)  この論文では、前者は建築物(物理的建造物)を表し、これをハードウェアと呼ぶこととする。後者は建築システムと表現しソフトウェアと呼ぶこととする。この提案は、今日のコンピューテション手法の隆盛によって、この二つの言葉を再び接続することを試みる。それは、設計者の手の延長線上にある図面ソフトのようなものだけでなく、設計者の知的機能の延長としての役割を担う使い方を意味する。また、これはコンピュータが知的な協働のための道具となるような可能性を示唆する。



     The thesis makes a contribution to the below 3 points

- to the debate about cybernetic architecture, particularly real-time optimisation, and robotic architectural elements that can make real-time decisions and can learn.

- to methods of human interaction with learning algorithms in architecture

- to scientific testing through physical demonstration of a responsive roof structure.





Fig.1-1, 1, A diagram of proposed Architecture (both in building and system and those connections.) 

1-1-1: Definition of Original Words

Several key phrases are defined below. It’s worth noting that the word ‘Architecture’ is used, in this thesis, in its traditional context of building and construction, not as a reference to a structure of logic or as an algorithm in the computing field.

1-1-1: オリジナルの言葉の定義


An ‘Architectural Machine’ is an architectural structure (building) which incorporates some kind of mechanisms such as robotic features or mechanical engineering techniques. This contrasts with a traditional structure (building) which is a static object. In contrast to ‘Robotic Architecture’, this system does not have a metaphysical system such as a computer. 

「アーキテクチャル・マシーン」とは、ロボティックな特徴や機械工学技術のような何種類かのメカニズムを合体させた建築の構造(アーキテクチャル・ストラクチャ)である。これは静的なオブジェクトである伝統的な建築物とは対照的なものである。「ロボティック ・アーキテクチャ」に比べ、このシステムはコンピューターのような形而上のシステムを持たない。

‘Robotic Architecture’ is a cybernetic architectural system consisting of a building and its control system. This is a combination of a physical structure and a metaphysical system. As with a robot, there are various degrees of automation, creating a system that has different degrees of intelligence. This system is a kind of automata;  depending on its intelligence it may be more or less responsive. It is possible to achieve higher functions such as the distinction between right and wrong responses, and even to renew its own system (Autopoiesis). However, up to the present (2014), these systems lack in intelligence. 


‘Responsive Architecture’ is a primitive cybernetic architectural system which consists of a building and its control system which have a monotonous (or patterned) response system. This system is a simple closed-loop system that can respond to stimuli from humans or the environment. Its responses are programmed and fixed in advance creating a system with limited adaptability. If the stimulus levels or types change, an administrator, human most likely, needs to re-program the system. This system sometimes is incorporated into ‘Robotic Architecture’ (defined above). But also it can be used in other types of response systems such as a material-based system. 


‘Programmable Architecture’ is a cybernetic, responsive architecture (building) which consists of robotic architecture and a flexible control system which has an interaction system (interface) between its autonomous software and its human users. This ‘autonomous’ system can be controlled by various methodologies and algorithms, as well as accepting intervention or being overridden by its human users. This hybrid system has a flexible and intelligent ecology. ‘Programmable’ here is polysemic referring to the control system which is programmable by a computer script, program-able in terms of its architectural functions which can be altered by changing its shape, and finally referring to the fact that it is controllable by the end user who may wish to alter an existing program or initiate a new program. 


1-2. Research Motivation

Since this is not a conventional architectural field, it is difficult to find a specific definition for concepts such as time-based design and the meaning of ‘Programmable in architecture ’or even ‘Metabolism in architecture’. However, user participation is considered central to the architectural design concepts developed in this work. As a familiar example of this, a house renovation, including enlarging or reducing space/volume can be used to explain these concepts. Dwelling surveys carried out in West Africa in 2007 (Hotta, 2008) show that the local inhabitants proceed to modify their own dwellings (fig1-2-1). The local woman breaks part of a mud wall and patches it with wet mud. 



Fig.1-2,1;  A local woman plastering her house with wet mud. They do not have the architectural profession, but residents mend themselves. This picture is taken in Kion village, Lele Tribe, in Burkina Faso, 2007, at West African Survey, by Fujii Lab, The University of Tokyo. Author (K.Hotta) participated in this research trip.

     However this work, ‘renovation’ does not just mean renovating size by adding a room, but also the use of adaptability to provide an equivalent architectural function. Surprisingly, in the 1960s, Mr A.Isozaki, who is a Japanese architect, was already aware of this issue which placed him in a critical position with respect to the Japanese Metabolist Movement (see Section 2-2). The problem is: that the specialist’s design cannot help but be a teleological structure as opposed to a vernacular procedure. (Vernacular procedure is architect-less architectural design, the lay-person frequently makes and modifies their buildings without planning) The Metabolists (Isozaki, 1963.) tried to avoid those specialists’ subjective design methods by using autonomous and time-based design methods (Time-based design is the architectural system that has the ability to change after it is built), but they fell into the same trap. True autonomous design should be handled by the user. 


1-3. Research Field's Background


1-3-1. From the Field of Architecture

In recent years, the keyword 'adaptability' is increasingly used in the field of emergent architectural design. Two recent book titles are “Adaptive Ecologies"(Theodore, S., Frazer, J., Schumacher, P., 2013) and "Unconventional Computing: Design Methods for Adaptive Architecture"(Armstrong, R.(Au), Simone, F. (Ed)). Also in ACADIA (The Association for Computer-Aided Design in Architecture), the most famous conference in the computational design field, the title of its 2013 conference was 'Adaptive Architecture’.


 今日的な建築デザインの分野において、近年「アダプタビリティー≒適応性」というキーワードは、よく使われている。例えば、建築のデザイン方法論に関する文献で「アダプティブエコロジーズ≒適応適生態」(セオドア 2013)と、コンピュータープログラミングに関する文献で「アンコンベンショナル・コンピューティング:デザインメソッド・フォー・アダプティブアーキテクチャー≒型破りなコンピューティング:適応的アーキテクチャーの設計方法」(アームストロング)などが相次いで出版された。また、コンピューテーショナルデザイン分野で最も有名な学会の一つであるACADIA(建築におけるコンピューターエイデッドデザイン学会、北米)でも、2013年の学会カンファレンスのタイトルは「適応的建築」であった。

Fig 1-3-1,1: A diagram of Contemporary Architectural Concepts.Considering non-static architecture (building), terms related to emergent architecture have appeared which could be arranged in ranking order of complexity, from primitive to higher-level functionality: ‘time-based design’, ‘event-based design’, ‘responsive design’, ‘interactive design‘, ‘adaptive design‘, ‘intelligent design‘ (Sherbini and Krawczyk, 2004). In recent times, in addition, new keywords have also appeared (Sterk, 2009a) associated with concepts such as ‘cybernetic machine’, ‘kinetic architecture’, ‘user participation’, and ‘discrete model‘.
図.1-3-1,1:今日的、建築コンセプトの図解静的ではない建築物に関するキーワードを、簡単なものから複雑なものの順で以下に挙げる。「タイムベースド・デザイン≒時間軸を考慮した設計」、「イベントベース・デザイン≒出来事を起点にした設計」、「レスポンシブデザイン≒応答可能な設計」、「インタラクティブデザイン≒相互作用可能な設計」、「アダプティブデザイン≒適応性のある設計」、「インテリジェントデザイン≒知能的設計」(シェルビニ等2004)である。近年では加えて、「サイバネティック・マシーン」、「キネティック アーキテクチャ」、「ユーザー参加型」、「離散モデル」など、方法に関する新しいキーワードもまた現れている。(スターク、2009)

     There are many approaches to making architecture (both buildings and systems) adaptive and sustainable. (Fig1-3,1) The most contemporary architecture uses scientific approaches based on mathematics, physics, and computational tools and through experimentation is able to achieve higher levels of adaptability against ever-changing circumstances. However, in considering the keyword ‘adaptability’ in architectural design, it is critical to discuss what adaptability is for. Also, there is a problem that the high 'adaptability' is used for the better thing in this context without description. It is problematic to speak of high ‘adaptability’ without a description of the nature and benefits of adaptability. This leads to the question - why is adaptability needed in the design process. Making this logic clear is a critical aspect of design research. 



     In this research, it was necessary to narrow the approach to certain aspects of the design. In this thesis, environmental adaptability is the centre of discussion. The aim is for architecture with higher environmental adaptability, and to develop an architectural design methodology that will address the shape, systems, and devices that constitute the building. In addressing adaptability a morphological method is taken though there are thousands of methodologies that could result in an adaptable architecture. The idea of shape-adaptive architecture, which consists of dynamic components, has been gaining popularity (Schumacher, 2010a). Component-based design methods, inspired by biology, have been present in German architectural design as exemplified by Frei Otto. However, because of the development of computational approaches, it has been re-interpreted within the contemporary period through the Emergent Technology and Design Programme at the Architectural Association (Hensel, M. Menges, A., Weinstock, M., 2010). After a decade its popularity has also been enhanced with contributions from the theoretical side. (Hensel, M. and Menges A., 2007)

 この研究では、設計(デザイン)行為の特定部分に限定してアプローチする必要がある。そこでこの論文ではビルドエンバイロンメント(主に建築環境工学)への適応性をケーススタディーとして議論の中心とすることと設定する。そこでの目的は、構築物がそれを包含する環境に対してより高い適応性をもつことであり、そのような建築デザインの方法論を確立することである。ここではとくに以下三点;形、システム、建物を構成するデバイスに関する設計方法に集中し議論する。適応可能な建築のための方法は沢山あるが、ここでは適応性を発揮するために形態学的な方法が採用された。例えば、動的コンポーネントで構成される形状・適応型建築のアイデアが建築デザインの界隈で人気を集めている(シューマッハー、2010)。生物学に触発されたコンポーネント(構成要素)ベースの設計手法は、フライオットーによって例示されているようにドイツの建築設計のながれにある。また、AAスクールのエマージェント・テクノロジー・アンド・デザイン・プログラム(ヘンセル、メンゲス、ウェインストック、2010)などのアカデミズムに代表されるように近年のコンピューテーショナルなアプローチの発展によって、現代的に再解釈された。 10年後、その流れは理論的な側面からますます強化されることとなった。(マイケルヘンセル、アキムメンゲス、2007)

1-3-2. From the Control Engineering Field

The ‘Control Engineering’ field is an interdisciplinary field seeking stable behaviour with a cybernetic aspect, in various systems. 'Control Theory’ in particular deals with dynamic systems from a mathematical point of view. All theory consists of 3 aspects namely 'the representational model', 'the analytic methodology' and 'the control design'. Recently, Control Engineering methodologies have expanded significantly alongside advancements in Control Theory. Because of technological advancements, 'Post Modern Control Theory ' and 'Intelligent Control Theory' have become increasingly important since the 1980s in contrast to 'Classical Control Theory' and 'Modern Control Theory’.



     A brief history of this theory is described here. The first so-called Automatic Control Systems were developed before Christ. For example in Alexandria, Egypt there was a water clock which had a feedback control system. Automata such as dancing figures became popular in Europe in the 17th and 18th centuries for entertainment. These systems were the typical 'Open-loop Control' systems, which repeated the same task over and over. 

     In contrast, the 'Closed-loop' control device was first invented by C. Drebbel around 1620, as a temperature regulator for a furnace. J Watt is famous for inventing the centrifugal fly-ball governor for steam engines in 1788. J.C. Maxwell first described control systems with differential equations in his paper "On Governors". (Maxwell, J.C.,1868) Edward John Routh and Adolf Hurwitz analyzed system stability using differential equations in 1877. This resulted in the Routh–Hurwitz theorem. This demonstrated the importance of mathematical models and started mathematical system theory though not in a convincing way.

     Around World War 2, mechanical applications of control devices became mainstream being used in flight control, fire control, guidance systems, sidewinder missiles, ship stabilizers, and even electronics. This technical competition shifted into the Space Race during the Cold War. Building on progress in stochastic, robust, adaptive and optimal control methods, this theory made significant progress. 


 対して、「閉回路」制御装置は、暖炉の温度調節器として、1620年頃にC.ドレベルによって最初に発明された。ジェームス ワットは1788年に、蒸気機関のための遠心力を使ったフライボール調整器を発明したことで有名である。J.C.マクスウェルは、彼のペーパー「オン・ガバナーズ」(マクスウェル、J.C.、1868年)で、微分方程式を使用した制御システムについて最初に説明した。エドワード・ジョン・ラウスとアドルフ・フルビッツは、1877年に微分方程式を使用してシステムの安定性を分析した。この結果が、ラウス-フルビッツの定理である。このようにして、制御理論の中でも数学的モデルの重要性が認識され、完全な方法ではないにしても、数学的システム理論が始まった。


     Referred to as 'Classical Control Theory', it was formally organized in the 1950s representing the dominance of the closed-loop system over open-loop systems. following Classical Control Theory, 1) in the time domain differential equations are used, 2) in the complex-s domain the Laplace transform is used and 3) in the frequency domain one uses a transformation from the complex-s domain.

     In this way, single input and output (SISO) are dealt with in a linear system called the 'Transfer Function'. When the frequency domain approach is taken, the Laplace transform is frequently used on the variables.'PID control' or short proportional-integral-derivative is one representative example of this. These theories are still at the centre of the field of Industry.



     'Modern Control Theory' is distinguished from Classical Control Theory through its use of time-domain 'State Space' representation (also known as the "time-domain approach") in contrast to 'the frequency domain analysis of the ‘Classical Control Theory’. The sets of inputs and outputs are represented as a first-order differential equation. Unlike the frequency domain approach, the use of the state space representation is not limited to systems with linear components and zero initial conditions. “State space" refers to the space whose axes are the state variables. The state of the system can be represented as a vector within that space. (Donald M Wiberg, 1971). Because of this, it can be applied to more complex problems. In the 1960s, optimized output feedback was popular in research. In the 1970s, systems with a combination of sensors and optimized regulators became the centre of academia. It resulted in a variety of regulators. 


     In terms of recent development, there are various types of methodologies referred to as 'Post-modern Control Theory'. The most common examples are Adaptive control, Hierarchical control, Optimal control, Predictive Control (MPC) and Linear-Quadratic-Gaussian control (LQG), Robust Control, Stochastic Control, Energy-Shaping control and Self-Organized Criticality control. Every control system must guarantee first the stability of the closed-loop behaviour. For linear systems, this can be obtained by directly placing the poles. Non-linear control systems use specific theories (normally based on Aleksandr Lyapunov's Theory) to ensure stability without regard to the inner dynamics of the system. The ability to address different specifications varies according to the model considered and the control strategy is chosen. 

 次に「ポストモダン制御理論」と呼ばれるさまざまなタイプの方法が近年発展中である。最も一般的な例は、適応制御、階層制御、最適制御、予測制御(MPC)および線形二次ガウス制御(LQG)、ロバスト制御、確率的制御、エネルギー整形制御、および自己組織化臨界制御である。すべての制御システムは、まず、閉回路動作の安定性を保証しなければならない。線形システムの場合は極を直接配置することで実現でき、非線形制御システムは、特定の理論(通常はAleksandr Lyapunovの理論に基づく)を使用して、システムの内部ダイナミクスに関係なく安定性を確保する。それぞれの仕様へ対応する能力は、検討するモデルと選択した制御戦略によって異なる。

     'Intelligent Control Theory' uses various AI (Artificial Intelligence) computing approaches such as neural networks, Bayesian probability, fuzzy logic, machine learning, evolutionary computation, Intelligent agents (Cognitive/Conscious control) and genetic algorithms to control a dynamic system. These methodologies have become popular not only because of computational advancement but also because of the development of algorithmic software. A feature of these methodologies is that when the model or controllers are constructed, they don’t require a specific physical character giving them great versatility. 


1-4. Aim and Objectives

The aim of this thesis is to demonstrate that the architectural fabric made by programmable architecture (PA) can reconfigure space in order to control its environment. This set space is ephemeral but can support the various activities that compose the human lifestyle. Physical architecture interferes with environmental elements such as light (illumination), sound (volume and frequency), air (direction, speed and heat) etc. The programmable architecture will control those elements by changing their physical form. By changing form, this architecture will make different types of layered spaces. In this proposal, the space is underneath a canopy.

1-4. 狙いと目的


     This approach to affordance is based on the hypothesis that environmental conditions can induce people to take specific actions. 

     For example, is ‘reading in the library’-how is it possible to reconfigure the space and induce the action? Though human action is unpredictable, it is possible to prepare circumstances which encourage the desired behaviour. In this instance, there is a comfortable environmental range of conditions for reading. For instance, there should be no rain, it shouldn’t be too dark for visibility, it should be silent, there should be a calm wind or no wind, extreme ranges of temperature need to be avoided etc.



     Thus this architecture (a combination of building and system) will encourage human activity. To support this argument, a new paradigm for spatial reconfiguration policy needs to be defined. Rather than planning based on spatial functions, PA is designed as a system which can fulfil the pluripotent functions of day-to-day living and working. Any architecture (building) has functions; functions are a way of using space in this context. In architectural design, traditionally most of the architecture (building) is planned, drawn and built based on this principle. However, in this proposal, the functions are ephemeral, even after being built the building’s functions are ever-changing in the same way as the events in a plaza. Users are searching for an appropriate space which has an optimal environment for a specific action. When this action is complete, they are encouraged to engage in a new activity or simply leave this space. However, this may sometimes be inconvenient for stable users. So In the proposed PA system, the user can control the character of space in a more advanced way creating spaces environmentally appropriate for their activities.


     This thesis’ objective is to design and test an architectural fabric (a combination of physical structure and system) which can follow an objective function more tightly than optimal or non-optimal conventional static structural fabrics. Here the objective function is based on an environmental factor, the amount of illumination. (Ideally, this experiment should be done with each of the above environmental factors, but those are omitted because of limited time and space) As in all control theories, the system's stability is measured as the difference between the desired value and the measured (sensor) value. The more stable system has a smaller difference over a predetermined period of time. (In this experiment one day was used). This proposed architectural system will test whether it is able to fulfil the chosen environmental requirements over the period of one day. This result will lead to determining the sustainability of the structure with regards to its functional flexibility as well as its energy consumption over the lifespan of the building. 

 この論文の目標地点は、形状最適化された静的なキャノピー構造物よりも、変動する目的関数によりぴたりと追従する動的なシステムをもつキャノピー構造物(物理的構造物とシステムの両方)を設計/デザインし、その性能をテストすることである。ここでの目的関数は環境要因、特に照度に基づいている。 (理想的には、この実験は上記の環境要因のそれぞれで行われるべきでだが、時間とスペースが限られているため、それらは省略した)すべての制御理論において、そのシステムの安定性(※正しい用語ではないが直訳)は、目標値と計測値(センサー値)の差として測定される。より安定したシステムは、ある期間内で差を小さくする力が強い。(この実験で、使用された期間は便宜的に1日の設定である。)この提案された建築システムは、一日を通して、選択された環境的条件を満たすことができるかどうかをテストする。この結果は、建築的機能の柔軟性によってもたらされるもの、構造物のライフスパンに渡るエネルギーの消費量の両方によって、将来にわたってのこの構造物のサスティナビリティーの延長につながるであろう。 

1-5. Thesis Overview

In chapter 2, several seminal literatures is reviewed in the context of proceeding among architectural theories and computational theories especially control theory, and other emergent fields on temporal design methods while addressing their limitations. The temporal design methods presented by the likes of Metabolists and Cedric Price offered alternative approaches to the static forms of architecture, where predominant discussions were inclined to address adaptability solely in the phase of planning as a representation of frozen time. When materialized, the resultant static architecture had already lost most of its flexibility and sustainability. The current computational design methods, including parametricism, also typify these issues. Yet true adaptability in architecture necessitates both dynamic hardware and software with the potential for continually renewable forms capable of all possible variations for the changing demands and conditions, without having to resort to the one supposedly optimal solution.



     In chapter 3, the hypothesis and following research questions are introduced. then, Programmable Architecture (PA) introduces a new strategy for robotic architecture as an intelligent system, consisting of both autonomous and subservient schemes that maintain constant homeostasis within its contained environment. Transmissions between genetic algorithms (GA) and user input prompt this hybrid system to output the consequent, ever-changing physical form. The contained environmental conditions are maintained more effectively than in the previous models where GA and user input operate disparately.


     In chapter 4, a concrete research approach and physical design will be introduced. The hardware for PA is an accumulation of self-sufficient machines that are dedicated to the actions of sensing-calculating-actuating. Each local machine makes its own simple decisions, which collectively turn into a larger problem-solving machine, or architectural robot, that simultaneously takes central orders into account. This robot manifests itself as a self-supporting skin structure, in which numerous machines are embedded. As a case study for this thesis, the machine that is organized with tensegrity components of variable forms is proposed. A model of this machine is built at a one-to-one scale and tested via the electrically controlled and wirelessly connected microcomputer chip called Arduino. 


     In chapter 5, data and analytical methods will be discussed. Referencing several previous experiments done by the author, the meaning of experiments in the following chapters is explained. Based on the illumination which functionally required, fluctuating objective functions are set. The difference between a measured value and desired value is compared and discussed how can it be minimized. Also, GA will be explained, its procedure and feature including positive point and negative points.


     In chapter 6, the first experiment done by pure GA with grasshopper is shown. Causes, of why GA does not work effectively, are assumed and tested in different settings, though the results are unsavoury. Hence this failure will be highlighting typical GA's shortcomings: protracted calculation time, adaptability for fluctuating objective functions, which represents the ideal condition at any given time, and the ability for ad hoc responses when the system experiences usage overload or environmental irregularities. 


     In chapter 7, reflecting the previous chapter, a hybrid system consisting of automated GA and manual user inputs is examined. The software for PA consists of operating systems and control systems. The latter subsists on a combination of automatic responses and user manipulations for a faster and higher degree of adaptation. Utilizing the versatility of GA, this model applies multiple user input to partially substitute its purely random mutations, thus resolving GA’s shortcomings. Incorporating the anonymous user input, the system can respond rationally to actual conditions unanticipated by GA. Therefore user input simultaneously controls the system locally to reflect individual preferences and contribute to the global optimization and increased efficiency for the system as a whole. 


     In chapter 8, the discussion and conclusion are described. Through the example of proposed hardware and software, this research provides a case of defining system performance measurements using original indicators, which aids in evaluating the ability of the system to react to environmental changes. The outcomes of this proposed hybrid system will be compared with a computationally static model, as in the case of parametrically optimized form, in addition to testing different arrangements, i.e. varying ratios of user input versus random mutations, within the dynamic model.