Machine learning-assisted lens-loaded cavity response optimization for improved direction-of-arrival estimation.

Area of Science:

• Electrical Engineering
• Electromagnetics
• Signal Processing

Background:

• Millimeter-wave (mmWave) systems require accurate direction of arrival (DoA) estimation for applications like 5G/6G communication and radar.
• Traditional DoA techniques face challenges in complex environments and with limited aperture sizes.
• Oversized resonant cavities offer potential for novel antenna designs but require advanced control mechanisms.

Purpose of the Study:

• To develop and validate a novel millimeter-wave DoA estimation technique.
• To introduce a dynamic aperture concept within a lens-loaded cavity for enhanced DoA performance.
• To optimize the dynamic aperture using a machine learning-assisted evolutionary algorithm.

Main Methods:

• Utilizing a lens-loaded oversized mmWave cavity supporting quasi-random wave-chaotic radiation modes.
• Implementing a mechanically controlled mode-mixing mechanism to create a dynamic aperture.
• Employing a machine learning-assisted evolutionary algorithm for optimizing the dynamic aperture states.
• Verifying the concept through extensive simulations of various dynamic aperture configurations.

Main Results:

• The lens effectively confines radiation and improves the gain of radiation modes, enhancing DoA accuracy.
• A lens-loaded dynamic aperture was successfully realized and optimized using the proposed method.
• Simulations demonstrated a significant 25% improvement in the conditioning for DoA estimation.

Conclusions:

• The proposed lens-loaded dynamic aperture technique offers a promising approach for accurate millimeter-wave DoA estimation.
• Machine learning-assisted optimization is effective in tuning the dynamic aperture for improved performance.
• This work advances the capabilities of mmWave sensing and communication systems.