Evolutionary Algorithms in Urban Planning

The functions supported by commercial CAD software are drawing, construction and presentation. Until now, no programssupporting the creative part of architectural and urban problem solving are on the market. The grand hopes of symbolic AI ofprogramming creative architectural and urban design have been disappointed. In the meantime, methods called New AI are available.Among these methods, evolutionary algorithms are particularly promising for solving design problems. The paper presents anapproach to town panning and architectural problem solving that combines an evolutionary strategy (ES), a genetic algorithm (GA)and a Particle System. The problem that remains incapable of being solved algorithmically has to do with the fact that in architectureand urbanizm form as well as function count. Because function relates to comfort, easiness of use, and aesthetics as well, it ishopeless to fully specify the fitness function of architecture. The approach presented circumvents a full specification through dividinglabor between the software and its user. The fitness function of town plans is defined in terms only of proportions of the shapes, areasand buildings to be accommodated and topological relations between them. The rest is left to the human designer who interactivelyintervenes in the evolution game as displayed on the screen