Fast evaluation of waveguide AR displays with an open-source GPU-based ray-tracing method

Conventional simulation methods for waveguide augmented reality (AR) displays rely on CPU-based commercial software and precomputed bidirectional scattering distribution function (BSDF) data, which are computationally expensive and inefficient for large-scale design optimization. To overcome this limitation, we developed a fast and open-source GPU-based ray-tracing algorithm that performs parallel computation using a look-up table (LUT) constructed from rigorous coupled-wave analysis (RCWA) results. A full-color diffractive waveguide design is used as the validation model, and the simulated results closely match those obtained from LightTools using BSDF input. Benchmark tests show that the proposed GPU algorithm achieves 7.63 × 10⁶ rays/s on an NVIDIA RTX 4080, representing a 57× speed improvement compared with an Intel Core i7-12700 K CPU. System performance is further analyzed in terms of efficiency, color uniformity, FoV uniformity, and eyebox uniformity. These results demonstrate that the proposed algorithm provides accurate optical predictions while significantly reducing the computation time, enabling rapid and scalable evaluation of diffractive waveguide-based AR displays.