Deep Learning Prediction of Beam Coherence Time for Near-FieldTeraHertz Networks

TL;DR

This work addresses beam management in mobile THz networks by accounting for near-field, spherical-wave propagation. It introduces a beam coherence time tailored to NF THz channels and trains a lightweight feedforward neural network to predict TB from a small set of mobility and system features, enabling beam updates with substantially reduced overhead. A large synthetic dataset with Gauss-Markov mobility and TB computed at $f_c\in\{142,280\}$ GHz provides offline labels for training, yielding a predictor that closely matches ground-truth TB and an ideal upper bound with zero overhead. Across pedestrian, bicycle, and vehicle mobility—and across the 142–280 GHz range—the approach yields higher effective rates with low inference latency, demonstrating practical viability for NF THz networks. Future work includes extending to multi-user scenarios and jointly optimizing TB prediction with beamforming.

Abstract

Large multiple antenna arrays coupled with accurate beamforming are essential in terahertz (THz) communications to ensure link reliability. However, as the number of antennas increases, beam alignment (focusing) and beam tracking in mobile networks incur prohibitive overhead. Additionally, the near-field region expands both with the size of antenna arrays and the carrier frequency, calling for adjustments in the beamforming to account for spherical wavefront instead of the conventional planar wave assumption. In this letter, we introduce a novel beam coherence time for mobile THz networks, to drastically reduce the rate of beam updates. Then, we propose a deep learning model, relying on a simple feedforward neural network with a time-dependent input, to predict the beam coherence time and adjust the beamforming on the fly with minimal overhead. Our numerical results demonstrate the effectiveness of the proposed approach by enabling higher data rates while reducing the overhead, especially at high (i.e., vehicular) mobility.

Figures (5)

• Figure 1: THz link between a large array AP and a UE.
• Figure 2: Diagram of the prediction model.
• Figure 3: Effective rate achieved using different beam durations for various mobility types. Updating the beam every $T_B$ yields a better tradeoff between rate performance and beam overhead.
• Figure 4: Average beam-update duration for different UE mobility profiles. $T_B$ remains larger than $T_C$, yielding less overhead. When the speed increases, $T_B$ decreases as expected.
• Figure 5: Average effective rate for different THz frequencies $f_c$. The prediction model effectively adapts to $f_c$ variations.