The best paper for the 2011 ARCC Architectural Research Conference was awarded to Michael D. Gibson of Kansas State University for his paper “Integrating geometry and light: daylight solutions through performance-based algorithms“. The full paper is available here.

ABSTRACT: Designing spaces for daylight is a complex problem for architects, balancing geometry with the location of daylight sources. Conventional design practice approaches this balance one-dimensionally: common procedures, rules of thumb, and building codes lead designers to default to regularity when designing windows and skylights.

The problem of daylight can be restated, starting first from the basic performance goal of distributed, uniform light. In traditional vernacular architecture, it is common to observe intentional coincidences among windows and interior surfaces, illustrating that openings and interior geometry can be integrated to distribute light in a way that is also experientially dynamic: integration also understood by great architects of the past and present.

Parametric design – a method of working where pieces of a simulated model can be manipulated ad infinitum – provides a new way of studying the relationship between light and geometry in the producing desirable, uniform, lighting conditions. Taking parametric design a step further, it is possible to tie together parametric models and computer-based simulations to produce an algorithm that “finds‟ optimal configurations between openings and interior geometry. Such an algorithm reveals two possibilities. The first is that designers can systematically determine the best relationship among openings and interior
space. Secondly, the success of these algorithms offers objective proof that, in comparison to the default of regularized patterns of openings, a more organic (i.e. less artificially ordered) relationship between openings and interior indeed is better for producing uniform daylight.

Two parametric algorithms will be discussed in the paper: an optimization algorithm, leading to a given problem to a single solution, and an evolutionary algorithm, using the random generation of individual solutions to reach better fitting results. The workings of the algorithms as well as the interpretation of the results in the context of design for daylight are discussed.