Abstract
Most evolutionary computation (EC) applications in design fields either assume simplified, static, performance-oriented procedures for design or focus on well-defined sub-problems, to be able to impose problem-solving and optimization schemes on design tasks, which render known EC techniques directly applicable. However, in most design situations, well-defined and static problems are not given, but must be constructed from messy situations, and the definition of a problem takes place during the solution process. Thus, evolutionary design requires contextual and dynamic problem definition and evaluation procedures, which has not yet been realized through EC. This study sets out for a critical reappraisal of EC for design, and proposes a conceptual framework as a research tool for the exploration of dynamic evolutionary design. After a critical review of EC in design, the article discusses its claims with reference to design theory, outlines the framework, and examines dynamic evolutionary strategies and required intelligent technologies. Although tackling a practical task, or solving the problem of dynamic evolution are not aimed in this study, an experimental application based on the framework will be presented in detail, to exemplify a mapping between the rather abstract concepts of the framework and the operators of a specific evolutionary algorithm.
