High-resolution data-driven models of daylight redirection components

Daylight Redirecting Components (DRCs) guide daylight to zones with insufficient
daylight exposure. They reduce energy demand for lighting, heating and cooling, and improve visual and thermal comfort. The data-driven model in Radiance is a general means to model DRCs in daylight simulation. Rather than internal optical
mechanisms, their resulting Bidirectional Scattering Distribution Function (BSDF) is
replicated. We present models of two DRCs that are generated from measurements. The impact of three necessary steps in the generation of data-driven models from measured BSDF shall be evaluated:
1. interpolation between measurements at sparse sets of incident directions,
2. extrapolation for directions that cannot be measured, and
3. application of a directional basis of given directional resolution.

It is shown that data-driven models can provide a realistic representation of both DRCs. The sensitivity to effects from interpolation differ for the two DRCs due to the different complexity of their BSDFs. Due to the irregularity of the measured BSDFs, extrapolation is not reliable and fails for both tested DRCs. Different measurement and modeling protocols should be applied to different classes systems, rather than aiming at a common low-resolution discretization.