Generating stylistically consistent product form designs using interactive evolutionary parametric shape grammars

Interactive grammar based design systems (IGBDS) are capable of generating large numbers of alternative designs. Prior to the application of IGBDS, a set of rules should be defined based on the theoretical theories and practical experiences in designing the objects, or by analysing the existing sets of objects. However, due to the time constraints in developing theoretical theories, gaining the practical skills of experts is of qualitative in nature and stylistically inconsistent designs to be analysed, it is difficult to systematically derive a set of shape grammar rules in generating designs which fulfils specific requirements. To address this problem, this paper presents an IGBDS enhanced by evolutionary computing to evolve a set of grammar rules for product form design. In this research, the forms of a product are analysed to derive shape features in the form of shape grammar (SG) rules. The rules are then encoded as the code scripts of a genetic algorithm (GA) representation in order to generate new shape grammar rules. The parameters in GA representation are modified by genetic programming (GP) which is regulated by control planning strategies named "GP-GA-SG". The "GP-GA-SG" control planning strategy first defines ways on how the control variables in GP should modify the variables in GA representation. The "GP-GA-SG" control planning strategy then defines ways on how the control variables in GA should modify the variables in SG representation. The SG rules modified by the GA variables define a new combination of shape features for alternative designs. In this way, traditional shape grammar is extended to an interactive context in which generative and evolutionary computing methods are combined. Both product component design as well as product configuration are supported in this framework. In this paper, we describe how this framework is formulated and discuss its potentials in supporting product design, with initial examples showing how the system is intended- - to work