Decoding Coherent Patterns from Arrayed Waveguides for Free-Space Optical Angle-of-Arrival Estimation.

Area of Science:

• Photonics
• Optical Sensing
• Artificial Intelligence

Background:

• Traditional Angle-of-Arrival (AOA) estimation methods face challenges in miniaturization, complexity, and reliability.
• Existing technologies struggle to meet the demands for compact and robust optical sensors.

Purpose of the Study:

• To present a novel free-space optical AOA estimation method overcoming limitations of traditional techniques.
• To leverage arrayed waveguide coherent mode decoding and deep learning for precise spatial angular information retrieval.

Main Methods:

• Utilizes AOA-related phase differences from incident light propagation and interference in an arrayed input waveguide.
• Forms multi-beam interference fringes at the slab waveguide output, sampled by an arrayed output waveguide.
• Employs a trained Convolutional Neural Network (CNN)-Attention Regressor for AOA estimation.

Main Results:

• Achieved Mean Absolute Error (MAE) of 0.0142° and Root Mean Square Error (RMSE) of 0.0193° over a 40° field of view.
• Demonstrated superior precision compared to conventional peak-linear calibration and other neural network architectures.
• Exhibited remarkable robustness against simulated phase noise and manufacturing tolerances.

Conclusions:

• The novel method effectively decodes spatial angular information of optical signals.
• Highlights the synergy between integrated photonics and deep learning for advanced optical sensing.
• Paves the way for highly integrated, robust, and high-performance on-chip optical sensors.