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

Area of Science:

• Optics and Photonics
• Computer Science
• Materials Science

Background:

• Surface relief grating waveguides are crucial components in augmented reality (AR) devices, enabling visual displays.
• Designing and optimizing these waveguides, especially for large-scale applications like AR head-up displays, presents significant computational challenges.
• Current methods struggle with rapid and efficient optimization, hindering the development of advanced AR systems.

Purpose of the Study:

• To propose and validate a novel method for optimizing the brightness uniformity of waveguide exit pupils.
• To accelerate the design and optimization process for surface relief grating waveguides used in AR devices.
• To enhance the visual performance of AR head-up displays through improved brightness uniformity.

Main Methods:

• Utilizing deep neural networks (DNNs) for waveguide design optimization.
• Integrating Bayesian optimization and particle swarm optimization to enhance DNN model accuracy.
• Applying a depth-first search algorithm for fine-tuning exit pupil uniformity.

Main Results:

• Achieved a computational acceleration factor of 40,000 compared to conventional methods.
• Maintained a high model accuracy exceeding 97% during optimization.
• Demonstrated a significant improvement in exit pupil brightness uniformity, reaching 0.46, a 44% enhancement over prior techniques.

Conclusions:

• The proposed deep neural network-based method offers a highly efficient and accurate solution for optimizing waveguide exit pupil brightness uniformity.
• This approach significantly reduces design time and computational cost for large-scale AR waveguide applications.
• The enhanced uniformity leads to improved visual quality and performance in augmented reality head-up displays.