Model-Driven Deep Neural Network for Enhanced AoA Estimation Using 5G gNB

TL;DR

This work tackles AoA estimation for 5G NR gNBs under hardware impairments by reframing the problem as inverse spectrum reconstruction. It introduces a spectrum-calibrating MoD-DNN that couples a CNN-based calibrator with a sparsity-aware SCG solver in an iterative optimization loop. The method yields sharper spatial spectra and lower AoA RMSE than both model-driven and data-driven baselines, demonstrated through simulations and anechoic-chamber experiments. The approach offers robust, efficient AoA estimation with practical potential for enhanced 5G positioning and location-based services.

Abstract

High-accuracy positioning has become a fundamental enabler for intelligent connected devices. Nevertheless, the present wireless networks still rely on model-driven approaches to achieve positioning functionality, which are susceptible to performance degradation in practical scenarios, primarily due to hardware impairments. Integrating artificial intelligence into the positioning framework presents a promising solution to revolutionize the accuracy and robustness of location-based services. In this study, we address this challenge by reformulating the problem of angle-of-arrival (AoA) estimation into image reconstruction of spatial spectrum. To this end, we design a model-driven deep neural network (MoD-DNN), which can automatically calibrate the angular-dependent phase error. The proposed MoD-DNN approach employs an iterative optimization scheme between a convolutional neural network and a sparse conjugate gradient algorithm. Simulation and experimental results are presented to demonstrate the effectiveness of the proposed method in enhancing spectrum calibration and AoA estimation.

Figures (7)

• Figure 1: Framework of MoD-DNN for AoA estimation.
• Figure 2: Illustration of proposed iteration for MoD-DNN module. (a) Overall iteration in MoD-DNN module. (b) Alternative optimization between CNN-based calibrator and inverse problem-based SSR.
• Figure 3: Structure of CNN-based calibrator.
• Figure 4: Spatial spectrum results at different AoA. Red dashed lines denote the truth. (a) MUSIC, $-45^{\circ}$. (b) MUSIC, $45^{\circ}$. (c) DeepMUSIC, $-45^{\circ}$. (d) DeepMUSIC, $45^{\circ}$. (e) MoD-DNN, $-45^{\circ}$. (f) MoD-DNN, $45^{\circ}$.
• Figure 5: Performance comparison of different methods. (a) RMSE versus SNR. (b) RMSE versus degree of impairment $\rho$. (c) Standard deviation versus epoch.