An algorithmic approach to system architecting using shape grammar-cellular automata

Abstract

This thesis expands upon the understanding of the fundamentals of system architecting in order to more effectively apply this process to engineering systems. The universal concern about the system architecting process is that the needs and wants of the stakeholders are not being fully satisfied, primarily because too few design alternatives are created and ambiguity exists in the information required. At the same time, it is noted that nature offers a superb example of system architecting and therefore should be considered as a guide for the engineering of systems. Key features of nature's architecting processes include self-generation, diversity, emergence, least action (balance of kinetic and potential energy), system-of-systems organization, and selection for stability. Currently, no human-friendly method appears to exist that addresses the problems in the field of system architecture while at the same time emulating nature's processes. By adapting nature's self-generative approach, a systematic means is offered to more rigorously conduct system architecting and better satisfy stakeholders. After reviewing generative design methods, an algorithmic methodology is developed to generate a space of architectural solutions satisfying a given specification, local constraints, and physical laws. This approach combines a visually oriented human design interface (shape grammar) that provides an intuitive design language with a machine (cellular automata) to execute the system architecture's production set (algorithm). The manual output of the flexible shape grammar, the set of design rules, is transcribed into cellular automata neighborhoods as a sequenced production set that may include other simple programs (such as combinatoric instructions).

(cont.) The resulting catalog of system architectures can be unmanageably large, so selection criteria (e.g., stability, matching interfaces, least action) are defined by the architect to narrow the solution space for stakeholder review. The shape grammar-cellular automata algorithmic approach was demonstrated across several domains of study. This methodology improves on the design's clarification and the number of design alternatives produced, which should result in greater stakeholder satisfaction. Of additional significance, this approach has shown value both in the study of the system architecting process, leading to the proposal of normative principles for system architecture, and in the modeling of systems for better understanding.