Interactive evolutionary design systems allow individuals to explore solution spaces for design solutions that might never have been considered using more traditional means. Our approach first involves the "meta-design" of a parametric prototype which implicitly defines an immense set of possible designs. The computer initially presents the user with a randomly generated population of these designs. The user chooses the "best" ones based purely on subjective, personal evaluation criteria. The software uses this qualitative "fitness" information to select which designs will be recombined ("mated" and "mutated") to produce the next generation of new design solutions. This process can be repeated until the user converges on a satisfactory set of solutions. We have been investigating such systems by building them using existing digital content creation software such as Maya, Max/MSP/Jitter, Houdini, and RenderMan. While the great majority of interactive evolutionary design research systems have involved entirely custom software intended for breeding entities within one fairly narrow problem domain, eg, certain classes of buildings, cars, images, etc. Our current focus is the creation of generic tools within these environments to allow non-programmers who are familiar with these tools to create their own evolutionary design problem spaces within the scope of each software package (eg exploring 3D form, lighting, or motion with Maya, or real-time video filtering within Jitter ...
Tags: OSU, ACCAD, maya, jitter, evolutionary design, technology, computer graphics
