透過您的圖書館登入
IP:54.226.210.133
  • 學位論文

活動與型態的雙生-多品類基因演算法之動態設計參數平衡實作

Design-Oriented Multi-Category Genetic Algorithm. Implementations of Dynamic Equilibrium of Design Parameters

指導教授 : 侯君昊

摘要


建築設計的複雜性源自於人類對空間不同的定義與追求,即便是相同機能的空間對應不同的使用者,皆會令結果有所不同,於是設計師時常必須賦予空間相當的泛用性。當前,因為空間或建築的設計參數整合之成熟(BIM),空間已經可以透過調整參數予以電腦先行模擬與調整,但仍依賴著設計者在施作前的設計選擇,本論文將探討多個空間參數在演算上動態調整與交互影響的可能,透過製造出空間調整的需求,思考真實空間因應活動或是使用者需求而改變的關係。 藉由重新檢視生態界運作的狀態,本論文試圖應用已存在的基因演算法,透過轉移物種間消長與資源分配的機制於設計參數的調用上,若能藉此重新思考設計參數在以往設計過程中的主從關係,變動的設計參數便開始對使用者產生價值。預想設計參數在基因演算法的調用下,能夠令空間與其產生之活動自我組織,除了能在其中容納不同的活動與事件,還能因應的環境與需求而持續自我調整,這即是設計參數對使用者再適應的過程。 本論文共計有兩個實驗階段,首先利用正弦波形在實驗第一階段測試改造後的基因演算法,並以之解釋演算法的詳細內容,接著將「露營」作為主題之設計譬喻加入實驗,企圖創造露營區內持續消長的動態配置,最後透過數據資料輔以圖面配置檢討演算法模型達成的目標。

並列摘要


Part of the variety of architectural design comes from the vary needs of spaces. Spaces with the same function can be way different according to users’ demands. Designers sometimes offer ambiguities of spatial programming which allow the re-definition from users. In the recent years, the integration of design parameters has been matured within BIM(Building Information Modeling). But the definitions of space still depend on users. Static parameters lead to static spatial programming, but users’ demands verify all the time. This research attempt to combine the design parameters into a self-evaluating mechanism through the observation of ecosystem. And this research introduces the term “environment” which represents the aggregation of design parameters. With the environment as parametric references, each design parameter has a fitness value in every GA cycle. This determines how important the design parameter could be under the circumstances of user. And the fitness values provide corrections to the environment as well. This research assumes the modified genetic algorithm will be able to provide users feedback instantly and create a self-organized procedure of spatial programming. The experiments of multi-category GA have 2 stages. At the first stage, the experiment invokes a phase-changing sinusoidal wave as a neutral goal and 2 series of random y-axis values for GA initialization. During the processes of GA routines, these series of y-values will converge to the goal wave gradually. With mathematical graphs and calculations as representation only, this implementation is designed just for the reliability testing. At the second stage, the design metaphor takes part of the programming process. This research uses the scenario “camping” as design metaphor. The configuration of camping site is composed by three elements (tent, vendor and camper), which also called “categories” in the later discussion. The configuration and the amounts of these three categories change constantly by adjusting the relationships between categories. During the procedure of the experiment, the micro successions between categories leads to a macro differences of environment, which also the objective this research attempts to achieve.

參考文獻


Delibasis, K., Asvestas, P.A., Matsopoulos, G.K.: 2010, “Multimodal genetic algorithms-based algorithm for automatic point correspondence”. Elsevier. Journal homepage: www.elsevier.com/locate/pr.
Tsutsui, S., Fujimoto, Y., and Ghosh, A.: 1997, “Forking GAs: GAs with Search Space Division Schemes” Evolutionary Computation, MIT Press, Vol. 5, No. 1, pp. 61-80.
Goldberg, David E.: 1989, Genetic Algorithms: In Search, Optimization & Machine Learning. [ISBN: 0 201 15767 5].
Holland, John H.: 1992, Adaptation in Natural and Artificial Systems: An Introductory Analysis with Applications to Biology, Control, and Artificial Intelligence. A Bradford Book (April 29, 1992) [ISBN: 0 262 58111 6].
參考論文:

延伸閱讀