Organizations: The Hong Kong University of Science and Technology (Guangzhou), Guangzhou, 511453, China · Southern University of Science and Technology, Shenzhen, China · The University of Hong Kong, HK SAR · The Hong Kong University of Science and Technology, HK SAR
Accurate and reliable near-field multi-source localization is essential for location-aware processing in future wireless systems. However, coherent multipath propagation complicates the direct recovery of line-of-sight (LoS) source positions, while existing estimation methods generally lack explicit control of the risk of source underestimation and positioning uncertainty. In this paper, we investigate reliable source enumeration and positioning in near-field coherent multipath environments. Specifically, we develop a conformal risk control (CRC) based source enumeration scheme that calibrates the penalty coefficient of an eigenvalue-based criterion to control the expected miss-detection ratio at a prescribed level. We further propose MUSIC-Net, a model-based deep learning framework informed by two-stage MUltiple SIgnal Classification (MUSIC), which learns the LoS-related subspace and surrogate distances through subspace-distance-based training, thereby enabling direct source positioning without estimating non-line-of-sight (NLoS) path parameters or performing subsequent path association. To quantify the uncertainty of the estimated positions, we develop a split conformal prediction (SCP) based calibration method that constructs angle-range predicted regions with finite-sample simultaneous coverage guarantees. Numerical results show that the proposed framework effectively controls the risk of source underestimation, while providing improved positioning accuracy with tighter predicted regions than benchmarks given the same target coverage levels.
Figures & tables
Fig. 1: An illustration of the system for near-field multi-source localization with multiple coherent NLoS paths.
Fig. 2: The overall protocol of the proposed reliable multi-source enumeration and localization framework. The plot in Stage 1 shows the MDR versus SNR for AIC, MDL, and the proposed CRC-calibrated scheme. The dashed line indicates the target MDR level αe=0.01 .
Fig. 3: Illustration of an injective mapping and the associated predicted regions with K=2 and K[R]=3 .
Fig. 4: The proposed two-stage MUSIC informed multi-source positioning based on MUSIC-Net.
Figure 5
Fig. 7: Violin plots of the empirical MDR and FAR for different source enumeration schemes with SNR=−3 dB and T=150 . “ ⋆ ” marks the empirical mean over 50 Monte-Carlo trials.
Fig. 8: Reliable source enumeration performance versus SNR with T=150 , αe=0.05 and ϵ=0.5 .
Fig. 9: MDR versus the number of observations T with SNR=−3 dB, αe=0.05 , and ϵ=0.5 . All four schemes yield zero empirical FAR over the tested range of T .
Fig. 10: ROC comparison at SNR=−3 dB and T=150 .
Fig. 11: Violin plots of the empirical coverage probability achieved under different target coverage probability 1−αp with SNR=5 dB, T=150 and αθ=αr=αp/2 . “ ⋆ ” denotes the empirical mean over 50 Monte-Carlo trials.
Fig. 12: Violin plots of the empirical coverage probabilities and corresponding efficiency metrics for position, DoA, and distance achieved by different positioning schemes with SNR=5 dB, T=150 , αθ=αr=αp/2=0.1 . “ ⋆ ” denotes the empirical mean over 50 Monte-Carlo trials.
Fig. 13: Comparison of different SCP score function designs with target coverage level 1−αp=0.8 , SNR=5 dB, and T=150 . “ ⋆ ” denotes the empirical mean over 50 Monte-Carlo trials.
Fig. 14: Comparison of different miscoverage allocation strategies in terms of the average set size A of Cp,k(R(n+1)) , the DoA quantile qθ,1−αθ , and the distance quantile qr,1−αr with SNR=5 dB, T=150 , 1−αp=0.85 and αp=αθ+αr . “ ⋆ ” denotes the empirical mean over 50 Monte-Carlo trials.
The Hong Kong University of Science and Technology (Guangzhou), Guangzhou, China · Southern University of Science and Technology, Shenzhen, China · The University of Hong Kong, HK SAR, China +1
Department of Computer and Electrical Engineering, Mid Sweden University, Sundsvall, Sweden · the school of electrical and data engineering, University of Technology Sydney, Sydney, Australia · the Department of Computer Science, KTH Royal Institute of Technology, Stockholm, Sweden