Editorial: The next generation of lighting calculations

It’s been 100 years since Gibson and Tyndall formally proposed the values that were standardized a year later as the CIE spectral luminous efficiency function for photopic vision, V(λ). The 1920s was a decade when several new computing machines were introduced, but before the first introduction of modern electronic computers in the 1940s. Calculations involving the nascent lumen were likely done by hand and were then at the forefront of lighting science. By the 1960s, methods to calculate average horizontal illuminance, such as the zonal cavity method, were available and the subject of scientific research to understand error levels and extend applicability to new situations. With the increasing availability of personal computers in the 1980s and 1990s, new lighting software based on ray tracing models, radiosity models or a combination was introduced to ease the burden of computing lighting quantities, extend computational capabilities to a much wider range of applications and offer rendered images. The performance of lighting calculation software has continued to improve, and in the last several years, it has begun to include better modelling of spectral quantities as well. There’s still room to improve the speed, accuracy and utility of lighting calculations, and the five articles in this issue all relate to this aspect of lighting research. Yoshizawa et al. present a new model for spatial lighting calculations and visualizations, Photon Flow, which focuses on the light field itself rather than surface-bound quantities. Chen et al. propose a grayscale luminance function to determine average luminance and the spatial luminance coefficient. Tsesmelis et al. propose a neural network architecture, DeepLux, for predicting illuminance of indoor scenes in real time. The final two articles focus on spectral power distributions (SPDs). Lokesh et al. present work that also applies neural networks, but their use is to predict changes to SPDs over time. Finally, Royer et al. document a new algorithm for computing large sets of metamers for colour-mixed LED systems. In the coming years, it will be interesting to follow how these methods, and others that are sure to follow, will influence lighting practice and lighting outcomes. Artificial intelligence, machine learning and data science have become key elements of many scientific fields, but the lighting profession – which is not known for rapid change – may have to look outside itself to effectively integrate new metrics, methods, workflows and business models. As Leland Curtis suggests in this issue’s opinion, the future of lighting practice may be very different than what is common today, with a focus on delivering scalable digital technology products instead of projects. I wonder if Gibson and Tyndall, or any of the other researchers whose data contributed to the definition of the lumen, envisioned a future where calculation software could provide realtime, accurate, photorealistic rendering of light in a space, and the associated improvements to lighting design practice, lighting quality, energy efficiency – essentially all aspects of the lighting profession. Likewise, from our current viewpoint, it is difficult to imagine how the research of today can further improve the built environment in the next century. Near-term progress may seem slow, but I believe that the work being done in this decade, and highlighted in this issue, may one day be seen as a turning point from which a new generation of lighting tools emerged.