Inverse design and uniformity optimization of diffractive waveguides for AR-HUD using neural networks

Surface relief grating waveguides are widely used in augmented reality (AR) devices such as AR glasses. However, the rapid design and optimization of such waveguides remain challenging, particularly in large-scale devices such as augmented reality head-up displays. We propose a method for optimizing the brightness uniformity of waveguide exit pupils using deep neural networks. The accuracy of the model is further enhanced through Bayesian optimization and particle swarm optimization. This approach accelerates computation by a factor of 40,000 while maintaining an accuracy of over 97%. A depth-first search algorithm is subsequently applied to optimize the uniformity of the waveguide exit pupil. Simulation results demonstrate that the proposed model effectively optimizes the exit pupil area and brightness uniformity, achieving a uniformity of 0.46 for the exit pupil, which represents a 44% improvement compared to previous methods.