System Architecture Generation based on Le Pont du Gard
Doctoral Candidate, MIT Engineering Systems Division
Senior Research Scientist, MIT Engineering Systems Division and Department of Mechanical Engineering
MIT Professor in the Department of Aeronautics and Astronautics, Dual Professor in the Engineering Systems Division
Last modified: March 31, 2006
One of the primary problems inherent in the current approaches to system architecting is the limitation in the number of design proposals that results from time and budget constraints as well as from the existence of design paradigms. This study explores a bottom up algorithmic approach based on design rules intended to satisfy a specification. A shape grammar is developed for the specification, which is transcribed into cellular automata to computationally enumerate and generate a design language space of all possible solution system architectures satisfying the specification. One notable outcome of this approach is the emergence of new forms. In addition, creative diversity, greater efficiency and improved stability of forms were obtained. Computational time and space were managed using rules to generate only those architectures that were physically legitimate and satisfied the given requirements. Applying symmetry grouping rules further reduced the solution space to a size that was easily reviewable and decidable by stakeholders for selecting final system architectures.