Architectural Design Optimization: Not an Usual Optimization Process

Abstract

Whereas optimization dowry is vastly and efficiently applied to many areas, both on academic and practice level, it had a poor influence on the architectural design practice. The complexity of such problems goes beyond the large number of constraints, decision variables and objective functions and beyond the difficulty of an accurate quantification of the customer’s intentions. Not all the objectives are known in advance, the design progresses by incorporating constraints and objectives in stages, there is a continuous co-evolution between the problem formulation and the solution space and so on. All these characteristics made up of architectural design an interesting and challenging field of study through optimization glasses. In search of a better design (achieved by an architect), the constant seems to be a continuous travelling on different type problem-formulation spaces, which are vast, complex and significantly interdependent one with each other. The perspectives of classical optimization and of architectural design prove to be very different; therefore special approaches, such as machine learning and other artificial intelligence techniques are more appropriate to tackle architectural design, even if the results obtained so far are still limited in performance. The paper integrates the course of action made by researchers and practitioners in finding adequate approaches for modeling and solving architectural design optimization. By that, it constitutes an interesting learning experience with unusual optimization contexts.