Design Exploration Using A Shape Grammar With A Genetic Algorithm

• Published in: Open house international 2007-06, Vol.32 (2), p.26-35
• Main Authors: Chouchoulas, Orestes, Day, Alan
• Format: Article
• Language: English
• ISSN: 0168-2601
• DOI: 10.1108/OHI-02-2007-B0004

Although the idea of linking a shape grammar to a genetic algorithm is not new, this paper proposes a novel way of combining these two elements in order to provide a tool that can be used for design exploration. Using a shape grammar for design generation provides a way of creating a range of potential solutions to a design problem which fit with the designer's stylistic agenda. A genetic algorithm can then be used to take these designs and develop them into a much richer set of solutions which can still be recognised as part of the same family. By setting quantifiable targets for design performance, the genetic algorithm can evolve new designs which exhibit the best features of previous generations. The designer is then presented with a wide range of high scoring solutions and can choose which of these to take forward and develop in the conventional manner. The novelty of the proposed approach is in the use of a shape code, which describes the steps that the shape grammar has taken to create each design. The genetic algorithm works on this shape code by applying crossover and mutation in order to create a range of designs that can be tested. The fittest are then selected in order to provide the genetic material for the next generation. A prototype version of such a program, called Shape Evolution, has been developed. In order to test Shape Evolution it has been used to design a range of apartment buildings which are required to meet certain performance criteria.