Using Evolutionary Methods for Design Case Adaptation

Case-based reasoning (CBR) provides a methodology for directly using previous designs in the development of a new design. An aspect of CBR that is not well developed for designing is the combination and adaptation of previous designs. The difficulty with this aspect of case-based design is partly due to the extensive amounts of specialised knowledge needed to select the appropriate features of a previous design to include in the new design and the adaptation of these features to fit the context of a new design problem. In this paper we present a design process model that combines ideas from CBR and genetic algorithms (GA’s). The CBR paradigm provides a method for the overall process of case selection and adaptation. The GA paradigm provides a method for adapting design cases by combining and mutating their features until a set of new design requirements and constraints are satisfied. We have implemented the process model and illustrate the model for residential floor plan layout. We use a set of Frank Lloyd Wright prairie house layouts as the case base. The constraints used to determine whether new designs proposed by the process model are acceptable are taken from feng shui, the Chinese art of placement. This illustration not only clarifies how our process model for design through the evolutionary adaptation of cases works, but it also shows how knowledge sources with distinct origins can be used within the same design framework.