Cover Image

Spatial configuration: Semi-automatic methods for layout generation in practice

Lucy Helme, Christian Derix, Åsmund Izaki


Despite five decades of research into layout automation that have produced a wide variety of methods for automatically solving layout problems, no automated layout programme is used as standard within the architectural profession. Therefore scope still exists to better formulate these methods in terms of their relevance to architects. In a traditional workflow, architects design by generating ideas and testing them in an iterative process through which a solution gradually emerges. The knowledge he or she has from years of training and experience is fed into the project through this design-evaluation loop. By automating the entire process, many explicit criteria can be successfully built into the model and optimised. However, these tools sit outside of the architect’s normal workflow, and the integration of his or her non-quantifiable assumptions during the design process is lost.

The Computational Design + Research group (CDR) of Aedas|R&D has developed semi-automated methods in which the designer is central to the process, known in other fields as ‘human-in-the-loop’. We see these methods as hybrids, where the design drivers that can be encoded as rules are built into the resulting tools, and the remaining criteria can be evaluated and optimised by the designer through interaction with the tool in a participatory process. Live evaluation and feedback are therefore fundamental to the tools’ success; the consequences and implications of both the encoded design drivers and of decisions taken while using the tool must be transparent. The problem of developing these methods is then to strike the balance between automation and interaction. If semi-automatic models are the ideal, what should be automated, and where should the designer be able to intervene? What algorithms can be used to aid arrangement while still allowing exploration of the design space? These questions have been explored through the development of a series of computational layout methods over the last few years. This paper presents a selection of these methods as applied to building configurations between 2008 and 2011.


Layout generation, spatial analysis, semi-automatic planning, interaction design, spatial configuration, simulation, heuristics.

Full Text: JOSS_2014_P35-49