Meta Predictive Matter


Master Thesis Project


ETH Zurich / Chair for Computer Aided Architectural Design


Prof. Ludger Hovestadt, Daniel Rohlek, Stella Azariadi


Generative Design, Texture Synthesis, Principal Component Analysis


Spring 2012



Description: ‘Meta predictive matter’ is a project that addresses the transformation of architectural ‘form’ into an ‘image’ and vice versa. The main focus lies on the interplay between the perception of space, form generation and its material manifestations and how this can inform the design process across multiple scales. The study investigates how to bring the programmability and polymorphism of digital forms into the physical world, by revising analytical design tools and matching them with computational tools for data analysis. This work emphasizes on the poly-semantic nature of design, by perceiving the design process as an open field of possibilities, which can react on the multiple layers of visual images that people experience as they traverse through buildings, cities and landscapes. The key point of this investigation is the creation of an integrated design system for the development of digital texture synthesis.
In order to achieve this, an abstraction of the perceived space is performed with the implementation of graphical maps where specific information is filtered out and fed into a digital elevation model that is simultaneously updated by performance analysis and texture-specific maps. ‘Meta predictive Matter’ explores the ability to intuit and predict the intractable behaviors of complex design systems and their repercussions to design eventualities and their material manifestations. It particularly focused on: a) gaining insight in the intractable design relationships that cannot be modeled using conventional associative methods, b) addressing both the opportunities in concept generation as well as the challenges in the translation of design to construction, c) investigating the combination of various ways of manufacturing for the development of custom fabrication processes.