On the algorithmic design of architectural configurations

The algorithmic design of an architectural configuration depends on one's ability to compare the merit of different solutions quantitatively, in order to decide which design is superior. We propose a merit function that accounts for the amount of view from a given unit and its accessibility to other units. The amount of view is measured as the size of the spatial region visible from the center of the unit, and accessibility is quantified as the proximity of the unit to all other units in the configuration. We show how to compute the merit of a given configuration of units efficiently and to use this in an algorithm that generates configurations by greedily adding and removing units with the aim of maximizing the merit of the resulting configuration, either in a planar or a spatial setting. We demonstrate that this method reproduces familiar architectural designs (such as streets, plazas, and skylines) and sheds light on the forces that drive their creation.