Accommodating Change and Open-Ended Search in Design Optimization
David A. Newton · Proceedings of the International Conference on Computer-Aided Architectural Design Research in Asia · 2018
Many real-world architectural multi-objective problems (MOPs) are dynamic and may have objectives, decision variables, and constraints that change during the optimization process. These problems are known as dynamic MOPs (DMOPs). Dynamic multi-objective evolutionary algorithms (DMOEAs) have emerged in the fields of optimization, operations research, and computer science as one way to address the challenges posed by DMOPs. DMOEAs offer new capacities for exploration and interaction with the designer, but they have not yet been studied in the field of architecture. This research addresses these issues through the development of a unique interactive DMOEA-based design tool for the conceptual design phase. We propose a new modification to the popular nondominated sorting genetic algorithm II (NSGA-II), that we call the dynamic progressive for architecture NSGA-II (DPA-NSGA-II). We show that DPA-NSGA-II outperforms NSGA-II in finding novel solutions.