Reinhard König

Previous Research

Since my first project on computer generated urban structures (Koenig & Bauriedel, 2004), I have worked intensively over the past ten years on the use of generative systems in architectural and urban applications. From 2005 to 2009 I explored generative and evaluative methods in greater depth in a dissertation (Koenig, 2010) at the University of Karlsruhe funded by a graduate scholarship from the State of Baden-Württemberg in Germany.

Thereafter I intensified my knowledge in urban simulation through a collaboration with the Centre of Regional Science at the Technical University of Vienna, in which we undertook a research project on the simulation of urban development in Vienna from 1888 to 2001 (Koenig & Mueller, 2011). Afterwards I worked intensively on two research projects funded by the German Research Foundation (DFG) at the Chair of Computer Science in Architecture (InfAR) at Bauhaus-Universität Weimar since 2009. The first project deals with computer-based methods for socially sustainable urban and regional planning (CoMStaR), which relates thematically to my dissertation. These projects used urban simulation methods to explore various correlations between the spatial structure of urban systems and the corresponding social organization of urban citizens (Koehler, Koenig, Kalisch, & Steinhoefel, 2009).

The focus of the second project was the development of a creative evolutionary design methodology for layout problems in architecture and urban design (Kremlas). As part of the Kremlas project, we conducted substantial work on the use of different EA and EMO for planning-related optimization techniques (Koenig, Schneider, & Knecht, 2012). The advantages of the methods we developed with respect to computing speed and flexibility of use were demonstrated in comparison to existing ones in Koenig and Knecht (2014).

I also contributed to the development of the “Framework for Enhancing Research in Architectural Design and Communication” (FREAC_3 ). This framework can be used for visualizing two- and three-dimensional geometric elements and also provides automatic versioning of all operations as well as spatially-distributed cooperation in real time (Schneider, Braunes, Thurow, & Koenig, 2011). Another more recent development in this vein is LUCI, developed at the ETH Zurich, which is a middleware system that can be used to combine various software tools and services as well as a GIS database via the web. The concept is similar to the recent proliferation of web processing services, but LUCI allows one to use any programming language to add a new service or tool.

In addition, I lead the Computational Planning Group (CPlan), which is published as an open-source project for using next generation computer tools for the production of architecture and urban design (Koenig, Standfest & Schmitt, 2014; Koenig, Treyer & Schmitt, 2013). Over the course of the project, we developed the CPlan planning synthesis application that uses evolutionary algorithms to optimize design proposals with respect to a number of design criteria.

The ongoing project “Analyzing trade-offs between the energy efficiency and social performance of urban morphology” (ESUM), which I also lead, investigates the relationships between the energy performance aspects of urban structures and the perception of a city by its inhabitants (König et al., 2014). Based on collected empirical data and our own computational models, we intend to find methods to derive the best possible solutions for spatial planning tasks that address both social and energy performance aspects.

Another ongoing project is Empower Shack, which is conducted by the Urban-ThinkTank (U-TT) Chair of Architecture and Urban Design at the ETH Zurichin collaboration with myself and several further international partners. The project aims to develop innovative and socially effective, yet affordable shacks for cities like Cape Town where the growth of informal settlements is uncontrollable. The Empower Shack project uses interdisciplinary approaches to explore the complexity of living conditions and address the economic, ecological, and security issues in such settlements. The U-TT is using analysis software developed by myself and the CPlan group to support the planning process. This project brings together many of my previous research topics in a practically applied planning situation on a real site.

Finally, I was part of the proposal writing team of the second phase of the interdisciplinary Future Cities Lab (FCL2) at the Singapore ETH center, and I am  Co-Principal Investigator (Co-PI) for the sub-projects cognitive design computing and citizen design science.

Planned Research

Figure 1: Diagram of the components of a framework of Cognitive Computing in Design.

1.   Methods of generating designs

As shown in Figure 1, there are several methods for generating geometry. The first is to synthesize geometry using evolutionary optimization methods, for example to generate floor plans, facades, building volumes, blocks, for land parceling and street networks. Evolutionary algorithms offer the possibility to vary restrictions and goals during the optimization process, making them an ideal choice in this context.

Geometry can also be generated by rule-based or procedural methods, or completely manually by individual designers. A further option is to use existing design repositories or catalogs. For the urban planning context, data from open street map can be used at various scales as a source of 2D geometry. 

2.   Evaluation and indexing

The basis for this part of the CCD framework is a geometry schema that permits the evaluation of the provided geometry. Ideally, this includes 2D urban data and 3D digital building representations. Based on this schema, the geometry can be evaluated using various analysis and simulation methods.

Some criteria are easily quantifiable, others require using statistical and machine learning methods for estimations. Additional analysis of the interdependencies between the criteria could reveal how certain design aspects influence the social context. Accordingly, empirical studies on the social implications of design should be integrated here.

The evaluation results are used at various scales as indexes to characterize a design by as many criteria as possible. This makes it possible to populate a search space with the corresponding design variants, which can be used for the parts of the CCD concept that are described in the following.

3.   User interaction

A primary objective of the CCD approach is to seamlessly integrate human cognition and design knowledge in a computer-based design process. One issue in this context is how to formulate the design or planning problem: the planners’ often vague qualitative requirements need to be translated into a precisely quantifiable constellation of criteria that can then be used to query the solution space. For a planner, the most natural means is a visual user interface rather than a numerical one. Another challenge is that the priorities of the criteria depend on many factors that vary according to the context of the project and what planners have come into contact with up to now: the set of good solutions may be too large and difficult to analyze, and the program must therefore model the context to narrow down the search domain.

Rather than focusing on finding optimal designs, we use CCD as an efficient technique for exploring the design space (the search and solution space in Figure 1). The aim is to produce an open CCD framework to efficiently search for compromise solutions to complex planning problems using an intuitive user interface for both formulating and representing planning problems and presenting solutions at various stages of the design process. The CCD system constitutes a design space exploration method that enables planners to deal with problems in a completely new, data-informed and evidence-based way.

4.   Learning and data analysis

The CCD system described here can be used as a basis for estimating the priorities of a planner with respect to certain criteria. To this end, a model is created to train a machine using the data collected from observing how users work with the CCD system. This model involves applying clustering algorithms to the first stage and machine-learned ranking to the second. Finally, we will implement a design routine that, on the one hand, proposes additional design alternatives to a planner and, on the other, evaluates the user’s choices of design variants in order to learn from them and better adapt to the user’s needs.


Koehler, H., Koenig, R., Kalisch, D., & Steinhoefel, J. (2009). Computer-based methods for a socially sustainable urban and regional planning. International Journal of Sustainability, Technology and Humanism, (4), 115–124. doi:10.5821/sth.v0i4.1058

Koenig, R. (2010). Simulation und Visualisierung der Dynamik räumlicher Prozesse: Eine computergestützte Untersuchung zu den Wechselwirkungen sozialräumlicher Organisation und den baulichen Strukturen städtischer Gesellschaften. (J. Klüver, Ed.)Modellbildungen und Simulationen in den Sozialwissenschaften (VS Researc.). Wiesbaden: VS Verlag. Retrieved from

Koenig, R., & Bauriedel, C. (2004). Computer-generated Urban Structures. In Generative Art Conference. Milan. Retrieved from

Koenig, R., & Knecht, K. (2014). Comparing two evolutionary algorithm based methods for layout generation: Dense packing versus subdivision. Artificial Intelligence for Engineering Design, Analysis and Manufacturing, 28(03), 285–299. doi:10.1017/S0890060414000237

Koenig, R., & Mueller, D. (2011). Cellular-automata-based simulation of the settlement development in Vienna. In A. Salcido (Ed.), Cellular Automata: Simplicity behind Complexity (pp. 23–46). Vienna: INTECH. Retrieved from

Koenig, R., Schneider, S., & Knecht, K. (2012). KREMLAS: Entwicklung einer kreativen evolutionären Entwurfsmethode für Layoutprobleme in Architektur und Städtebau. (R. Koenig, D. Donath, & F. Petzold, Eds.) (Koenig, R.). Weimar: Verlag der Bauhaus-Universität Weimar. Retrieved from

König, R., Schneider, S., Hijazi, I., Li, X., Bielik, M., Schmitt, G., & Dirk, D. (2014). Using geo statistical analysis to detect similarities in emotional responses of urban walkers to urban space. In Sixth International Conference on Design Computing and Cognition (DCC14) (p. 1). London.

Schneider, S., Braunes, J., Thurow, T., & Koenig, R. (2011). Design Versioning: Problems and possible solutions for the automatic management of distributed design processes. In CAAD Futures: Designing Together. Liege, Belgium.