Automatic generation of architectural floor plans requires many constraints and rules to be considered, based on the user's preferences regarding the structure, functionality, geometry and connectivity of the rooms. In this work, we present the implementation of a system that allows automatic generation of floor plans using evolutionary algorithms and techniques inspired by shape grammars. Depending on the user's constraints, we define several criteria whose scores represent the fitness score. We compare different ways of calculating fitness scores; furthermore, we explore the effect of specific mutations defined for this problem, and variable crossover and mutation rates.
We also involve the user in the generation process, where they can change the layout of the floor plan during the evolution in an interactive way. The implemented system can achieve the optimization of all objectives for a given scenario in an interactive or non-interactive way, however, it is not guaranteed that every floor plan will be acceptable.
|
