The adoption of extremely large antennas in high-frequency bands is driving the demand for accurate near-field localization.
This research proposes AI-aided subspace methods for near-field localization, addressing limitations of conventional techniques.
Specifically, NF-SubspaceNet uses deep learning to improve localization in challenging conditions, while DCD-MUSIC decouples angle and range estimation to reduce complexity.
Simulation results show that these methods outperform classical and existing deep-learning-based techniques for near-field localization.