Black-Box Optimization of Lighting Simulation in Architectural Design

This paper deals with an application of optimization in architectural design. Formally, we consider the problem of optimizing a function that can only be evaluated through an expensive oracle. We assume that the analytical expression of the function is unknown and first-order information is not available. This situation frequently occurs when each function evaluation relies on the output of a complex and time-consuming simulation. In the literature, this is called a

black-box

optimization problem with costly evaluation. This paper presents a black-box problem from architectural design: we aim to find the values of the design variables that yield optimal lighting conditions inside a building. The building façade is described as a parametric model whose parameters are the design variables.We tackle this problem by adapting the Radial Basis Function (RBF) method originally proposed by Gutmann (2001). Experiments indicate that our open-source implementation is competitive with commercial software for black-box optimization, and that it can be a valuable decision-support tool for complex problems requiring time-consuming simulations. The usefulness of this approach goes beyond the specific application in architectural design.