More is More: The No Free Lunch Theorem in Architecture

  • YEAR
    2020
  • AUTHORS
    Pereira, Inês
    Leitão, António
  • CATEGORIES
    2020 Conference Papers
    Conference Papers
    Digital Architecture, BIM & City Information Modelling (CIM)

Extract

The growing interest in sustainability and environmental design promoted the development of tools that respond to the architects’ demand for efficient ways to evaluate building performance and optimize their designs. To optimize a design, a parametric model of that design is iteratively instantiated and evaluated by an optimization algorithm that searches for the solutions that best fulfil a set of performance objectives. However, according to Wolpert´s No-Free-Lunch Theorem, different optimization problems are best served by different optimization algorithms and, thus, architects should test multiple ones. Unfortunately, this is not a straightforward task for the currently available algorithmic design tools because different optimization algorithms have different requirements, forcing architects to spend considerable time and effort to adapt their parametric models and simulation tools, to configure the optimization algorithms, and to visualize the optimization results. This paper addresses the beforementioned problems by presenting the integration of a wide range of optimization within the same algorithmic design tool. We discuss the use of this tool in a case study that demonstrates the usefulness of multiple optimization algorithms to avoid the potential non-optimality that emerges from using just one of them.

Keywords: Architectural optimization, No free lunch theorem, Parallelization

PDF

Download
To top