繁衍形體

Abstract

浮現子形這個現象在過去的十年間已經被大量的討論，特別是在設計初期的發展過程中，這個現象扮演著推進設計一個非常重要的角色。所以，在這個現象被提出之後，另外一群研究者從運算的角度來重新檢視這個現象，其中最知名的不外乎是形狀文法（Shape Grammar）、資料檢索模型 (Data-driven symbolic model)以及神經網路 (Connectionist networks) 等。這些方法最大的差異在於，前面兩種比較傳統的運算方式在對於產生很多形體的運算上很有效率，另外一個神經網路正相反，他提出了應該以尋找有效的隱含形體為主，而非是去尋找無限多的形。所以在本研究首要的目的便是希望可以提供一個方法，可以來同時處理多且有效的形的問題。 此外，在大量相關的研究之後，這些認知與運算的研究已經提供了非常充足的基礎來建構一個電腦輔助設計系統，是以能夠從認知的角度為出發，透過電腦運算的能力來加強設計者的能力。這樣一來，電腦媒材才可以不僅僅是被當作呈現的工具，而進一步從運算的能力來輔助設計者。所以如何能夠從這樣一個研究的基礎為背景，進一步提出適當的電腦輔助設計工具 (computer-aided design tool) 是在本文另外一個重要的問題。 所以，為了達到這樣的目標，在本研究中希望透過基因演算法，從一個演化的觀點來重新檢視這樣一個問題，最後，透過將這個方法整合到一個3D建模的環境，這樣一個結合理論研究與電腦輔助的工具將會更進一步推進到有效且實用的階段。

A cognitive phenomenon, emergent subshapes, has been examined extensively in the last decade. From the cognitive perspective, “emergent subshapes” plays the important role in developing the design, and this phenomenon happens very frequently during the early conceptual phase. Based on this, various computational approaches also arise to explore this cognitive phenomenon, and the most well known of them are symbolic and connectionist approaches. To compare these two methods, one distinct discrepancy is that symbolism is good at efficient performance where connectionism excels in providing valid recognition. In other words, one focuses on providing abundant shapes and the other tries to recognize those useful, though few, subshapes. Therefore, the predominate problem statement of this study is how to propose a suitable mechanism to generate not only numerous but also valid subshapes. Moreover, after abundant explorations in design thinking, cognitive and computational approaches are already sufficient enough to provide computer-aided design (CAD) tool with a very substantial foundation. New computer media also bring designers a promising vision by means of its powerful computing capacity; notwithstanding, this is no more than prospect. Therefore, how to take advantage of the computational capacity to assist designers from a cognitive perspective is also the other significant issue in this study. Particularly, in order to make the computer media get closely involved into the process of designing rather than merely serve as a presenting tool, this study aims to propose a mechanism that could show the potential of not only recognizing subshapes but also allowing designers to interact with throughout the early conceptual phase. To specify this, how to achieve this objective will start from exploring the cognitive perspective—emergent subshapes, and then implement this mechanism as a CAD system to assist designers. Furthermore, in this study, genetic algorithm is adopted to tackle this issue. By means of the nature-like selection, the relationship of shapes and subshapes was reexamined; thus, from this point of view, genetic algorithm provides another approach—neither symbolism nor connectionism—to support designers in evolving more unanticipated subshapes. Ultimately, through integrating this mechanism into the three-dimensional modeling application, the combination of theoretical researches with computer-aided design could go further into a more practical and constructive stage.