Shape Evolution

Shape Evolution is a generic design tool for supporting the initial (concept design) stages of architectural design. It aims to inspire the architect towards more innovative solutions to design problems by offering unanticipated, evolved designs that both respond to the architect’s stylistic agenda and satisfy the functional requirements of the brief.

Shape Evolution combines a shape grammar, used as the vehicle for aesthetics and style, with a genetic algorithm that optimises designs with respect to their functional performance. The designer inputs the shape grammar and the optimisation criteria and is presented with a collection of appropriate design solutions.

The key interface between the shape grammar and the genetic algorithm is a string that encodes the sequence by which shape grammar rules have been applied to generate a given design. This string, which uniquely identifies each design, is used as the genotype for the genetic algorithm. This allows the genetic algorithm to operate by modifying the sequence of rules that generated a design, not the geometry of a design directly. Consequently, modified designs are valid in the design language defined by the shape grammar and retain the stylistic characteristics chosen by the designer.

A simple computer implementation of Shape Evolution for the design of an apartment building has been developed for testing.

The use of Shape Evolution as a tool during early design stages has a number of advantages:

The high level of separation between input and output means that resulting designs may have unanticipated properties. This has the potential to inspire the user towards innovative solutions.
The genetic algorithm “weeds out” solutions which are not relevant to the design’s context, defined through the genetic algorithm’s evaluation criteria. This frees up the designer to focus on the more creative tasks.
Generated designs already conform to the designer’s aesthetic requirements to a large degree, thanks to the use of shape grammars for their generation.
A large number of designs can be quickly evaluated and the most successful ones presented to the designer.

This project was supervised by Professor Alan Day.

Downloads

A poster on this work was presented at the Artificial Intelligence in Design 2002 conference, which was held in Cambridge in July 2002. You can download the poster abstract [PDF, 312 KB], the poster itself [PDF, 576 KB], and a VRML file [WRL, 600 KB] of the example design shown on the poster.

A paper [PDF, 912 KB] relating to this work was published in June 2007.
Chouchoulas, O. and Day, A. (2007). Design Exploration Using a Shape Grammar with a Genetic Algorithm. Open House International, 32 (2), pp. 26-35.

Also available is the thesis itself [PDF, 13.68 MB].

Orestes Chouchoulas

I was born in Athens, Greece in 1976. I received a BSc (Hons) in General Architectural Studies in 1999, an MArch degree in June 2004, and a PhD degree in December 2004, all from the Department of Architecture and Civil Engineering at the University of Bath. My PhD research, within the Centre for Advanced Studies in Architecture, concerned the development of Shape Evolution, an inspirational design tool combining shape grammars and genetic algorithms. I have lectured and tutored in architectural design and the use of digital technologies in architecture at the University of Bath, Oxford Brookes University, Central Saint Martins, and the School of Art, Architecture and Design at the London Metropolitan University. I have worked in large and small architectural practices with a focus on residential architecture. I direct multi-disciplinary studio Bureau Bureau where we develop and implement visual communication strategies. Previously I co-founded Dark Matter Laboratories to research and design system-aware institutional infrastructure for more equitable cities and organisations.