Near-Field Sensing Enabled Predictive Beamforming: Fundamentals, Framework, and Opportunities

Hao Jiang, Zhaolin Wang, Zhaoyu Li, Hyundong Shin, Arumugam Nallanathan, Yuanwei LiuarXiv 2025 · DOI · arXiv

Abstract: The article proposes a novel near-field predictive beamforming framework for high-mobility wireless networks. Specifically, due to the spherical waves and non-uniform Doppler frequencies brought by the near-field region, the new ability of full-dimensional location and velocity sensing is characterized. Building on this foundation, the near-field predictive beamforming framework is proposed to proactively design beamformers for mobility users following arbitrary trajectories. Compared to the conventional far-field counterpart, the near-field predictive beamforming stands out due to: i) Prior-Knowledge-Free Prediction, and ii) Low-Complexity and Generalizable System Design. To realize these advantages, the implementation methods are discussed, followed by a case study confirming the benefits of the proposed framework. Finally, the article highlights promising research opportunities inspired by the proposed framework.

Introduction

In high-mobility wireless networks, such as Vehicle-to-Everything (V2X) or Unmanned Aerial Vehicle (UAV) communications, maintaining a sharp, directional beam on a moving target is a significant challenge. Traditional predictive beamforming relies on Integrated Sensing and Communication (ISAC) to track users. However, conventional systems operate in the far-field, where radio waves are assumed to be planar. This assumption limits sensing capabilities to angle and radial velocity, necessitating rigid "state evolution models" to guess user trajectories. These models often fail when users follow unexpected paths.

Near-Field Sensing Fundamentals

A paradigm shift is occurring as 6G networks move toward Extremely Large Antenna Arrays (ELAAs) and higher frequencies (mmWave/THz), significantly expanding the near-field region. Unlike the far-field, the near-field is characterized by spherical wave propagation and non-uniform Doppler shifts across the antenna array.

Conceptual geometry of near-field sensing and corresponding Cramér-Rao Bounds for distance, angle, and velocity components.

As illustrated in the figure above, the curvature of spherical waves allows the receiver to distinguish between signals at the same angle but different distances. Furthermore, because the antenna array is large relative to the user's distance, different parts of the array observe the user moving at slightly different relative velocities. This physical reality enables "full-dimensional" sensing, capturing distance, angle, radial velocity, and transverse velocity simultaneously. The Cramér-Rao Bound (CRB) analysis confirms that as a user moves into the far-field, distance and transverse velocity information "degenerate," proving that near-field sensing is uniquely capable of high-precision tracking.

The NFPB Framework

The proposed Near-Field Predictive Beamforming (NFPB) framework utilizes this full-dimensional sensing to predict a user's future position without any prior knowledge of their trajectory. The framework operates in a five-step loop:

  1. Reception: Receive echo signals from the user.
  2. Sensing: Extract full-dimensional status (angle, distance, and both velocity components).
  3. Prediction: Predict the next position using a simple kinematic model.
  4. Design: Design the beamformer to maximize gain and compensate for Doppler shifts.
  5. Transmission: Transmit data using the predicted beam.

This approach eliminates the need for traditional communication overheads, such as training pilots and dedicated uplink feedback, saving valuable airtime for data transmission.

Performance Evaluation

The framework was validated through a numerical case study simulating a downlink single-user setup with 256 antennas at 30 GHz. The simulation tracked a complex, arbitrary spiral-like trajectory to test the system's ability to operate without a pre-defined state evolution model.

Trajectory tracking results for a complex spiral path, comparing ground truth with NFPB tracking.

The results demonstrate that the NFPB framework can track both location and velocity components with high precision, closely matching the ground truth even for non-linear paths. This confirms that the physics of the near-field provides sufficient information for accurate prediction regardless of the movement type.

Opportunities and Future Work

While promising, several challenges remain for real-world deployment:

  • Computational Complexity: Real-time processing for ELAAs is demanding; Deep Learning (LSTMs or Transformers) may offer a solution for offloading complexity.
  • Non-Line-of-Sight (NLoS): Multi-path scattering in urban environments complicates sensing.
  • Hardware and Waveforms: Realizing these gains requires flexible antenna designs, Reconfigurable Intelligent Surfaces (RIS), and waveforms robust against non-uniform Doppler shifts.

By treating the near-field as a new sensing dimension rather than a source of interference, NFPB offers a robust, low-overhead solution for the high-mobility demands of 6G.

Summary updated

Abstract

The article proposes a novel near-field predictive beamforming framework for high-mobility wireless networks. Specifically, due to the spherical waves and non-uniform Doppler frequencies brought by the near-field region, the new ability of full-dimensional location and velocity sensing is characterized. Building on this foundation, the near-field predictive beamforming framework is proposed to proactively design beamformers for mobility users following arbitrary trajectories. Compared to the conventional far-field counterpart, the near-field predictive beamforming stands out due to: i) Prior-Knowledge-Free Prediction, and ii) Low-Complexity and Generalizable System Design. To realize these advantages, the implementation methods are discussed, followed by a case study confirming the benefits of the proposed framework. Finally, the article highlights promising research opportunities inspired by the proposed framework.
BibTeX
@misc{jiang2025near,
  author = {Hao Jiang and Zhaolin Wang and Zhaoyu Li and Hyundong Shin and Arumugam Nallanathan and Yuanwei Liu},
  title = {{Near-Field Sensing Enabled Predictive Beamforming: Fundamentals, Framework, and Opportunities}},
  year = {2025},
  eprint = {2506.09225},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2506.09225},
  howpublished = {arXiv},
}
Figure