GeneRT: A Generative Raytracing Tool for the rapid approximation of internal luminous conditions

Abstract

Over the past years, building research community has been exploiting simulation tools to estimate internal luminous conditions. However, this approach inherited many challenges related to the increased computational complexities and processing durations that can disrupt the workflow of design process. Previous attempts presented more efficient 2D forward raytracing tools to replace heavy simulations. Still, they did not consider orientations, materials and geometries of study spaces and surrounding contexts, without supporting their results by statistical evaluations. This research addresses previous limitations by introducing GeneRT as a rapid proxy to sophisticated simulations. Building upon 3D forward raytracing, GeneRT is developed in Python on top of Grasshopper, unprecedently accounting for daylight components, sky models and different factors that affect daylighting, without requiring complicated techniques or hardwares. The research describes the calculation procedures of GeneRT, statistically comparing its raytracing results against 18 simulation scenarios to evaluate the tool's accuracy and efficiency. Validation tests confirmed the ability of GeneRT to estimate simulated illuminance values, considering different geographical locations, external obstructions and daylighting systems, with acceptable normalized root mean square errors of −0.64–24.00%, 8.35–12.57% and 3.32–24.15%, respectively. GeneRT could obtain these results faster than simulations as 38.59–75.07%, 36.53–69.86% and 52.67–65.23%, respectively. Supported by 65,700 verification tests, the results also showed strong associations between GeneRT and simulation outputs for 70% of occupied hours. GeneRT shall redefine the practice of daylighting design, providing architects with flexibility to explore and make informed decisions during the early stages of design using the tool's interactive visual and numerical feedbacks.