Self-organizing computation a framework for generative approaches to architectural design

Abstract

This thesis proposes a conceptual framework for applications of self-organizing logics in generative design systems. The methods introduced in this thesis are in an abstract and conceptual form that explores one possible future direction of computational design strategy. In order to explain the potential of this problem-solving direction, general aspects of what our contemporary practice in architecture and urban design is facing will be discussed in response to the increasing complexity in our culture. However, the main focus of this thesis is not on providing immediate solution methods to resolve any specific professional problems in contemporary architecture. Rather, the thesis investigates the emergent characteristics of this method that can potentially evolve new design solutions over time, and shows how tools employing the method can be used for design collaboration with humans, rather than simply as passive evaluation and analysis tools. The thesis foresees important potential for this new design direction inspired by self-organizing computation (SOC) and speculates regarding its potential areas of application in architecture and urban design. In recent years, many scientists have started to gain the advantages of self-organizing systems in nature through their computational models in areas such as telecommunication networks and robotics. Main advantages of such systems are robustness, flexibility, adaptability, concurrency, and distributedness. One of the unique characteristics of SOC is its non-reliance on any external knowledge. As with many conventional computational methods in architecture, imposing existing design patterns or transformation sequences is beneficial when one wants to efficiently derive what appear to be the subjects of our recognitions. However, reliance on a pre-existing template might preclude the possibility of discovering what original inputs naturally turn into. SOC is a computational approach that brings out the strengths of the dynamic mechanisms of self-organizing systems: structures appear at the global level of a system from interactions among its lower-level components. In order to computationally implement the mechanisms, the system’s constituent units (subunits) and the rules that define their interactions (behaviors) need to be described. The system expects emergence of global-scale spatial structures from the locally defined interactions of its own components.