Beyond heuristics: A novel design space model for generative space planning in architecture

AbstractThis paper proposes a novel design space model which can be used in applications of generative space planning in architecture. The model is based on a novel data structure which allows fast subdivision and merge operations on planar regions in a floor plan. It is controlled by a relatively small set of input parameters and evaluated for performance using a set of congestion metrics which allows it to be optimized by a metaheuristic such as a genetic algorithm (GA). The paper also presents a set of guidelines and methods for analyzing and visualizing the quality of the model through low-resolution sampling of the design space. The model and analysis methods are demonstrated through an application in the design of an exhibit hall layout. The paper concludes by speculating on the potential of such models to disrupt the architectural profession by allowing designers to break free of common ‘heuristics’ or rules of thumb and explore a wider range of design options than would be possible using traditional methods.