Alter EVO: an interactive evolutionary computation tool for instant architecture processing.
Résumé
This paper briefly describes a bio-mimetic principle used to solve a simple spatial allocation problem. We define an initial set of bidimensional figures by specifying their respective dimensions. An initial spreadout creates 144 random solutions, each with a multiple set of optional possibilities: the inner evaluation function points the best pattern combination in terms of compacity and/or non overlapping occurences. Henceforth user can choose - following the evaluation function or not - to rebuild a fresh set of combinatory patterns within a mutation range preset: according to the visual result of the prior refinement process, user can indeed choose to "gently nudge" the present combinatory set (simply by flipping an occasional random bit in the chromosome) or - on the contrary - to "smash badly" its phenotype by a full recasting of its genotype. The purpose of this experimental implementation is to substantially reduce computation or evaluation time to produce plausible solutions, according to initial spatial constraints and considering that human attention tends to rapidly decrease its efficiency.
Origine | Fichiers produits par l'(les) auteur(s) |
---|