Answers are grounded in this paper and cite their sources.
Reliable wireless coverage in curved passenger subway tunnels requires spatially resolved propagation predictions for coverage assessment. Fine simulations are computationally expensive, while coarse grids introduce numerical discrepancies and sparse sampling. This article presents amplitude-corrected implicit reconstruction (ACIN) for three-dimensional received-power reconstruction. A calibration branch corrects normalized coarse-grid power, while cross-sectional and axial implicit neural representations (INRs), termed SectionINR and AxialINR, respectively, recover unsampled details. Joint training at 0.9, 2.4, and 3.5 GHz yields receiver-trace root mean square errors (RMSEs) of 1.3231, 1.5366, and 1.6004 dB against fine-grid vector parabolic equation (VPE) solutions. The baselines comprise a convolutional neural network (CNN), a U-shaped convolutional encoder–decoder network (U-Net), and SRFormer, a super-resolution transformer using permuted self-attention. ACIN reduces the RMSE by 56.89–58.70% relative to SRFormer and achieves 14.59×–28.42× speedups over fine-grid simulation, including the coarse solution. In the Madrid application, the same model is evaluated at 0.92 and 2.4 GHz and two bend radii without fine-tuning. The RMSEs range from 3.1732 to 3.8462 dB against fine-grid simulations and from 4.6585 to 9.4808 dB against the measured receiver power. These results support efficient power reconstruction for coverage assessment in the evaluated curved-tunnel environments.
No takes yet. Share an insight, caveat, or question.
Yang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: