Vehicle Network Beamforming Method Based on Multimodal Feature Fusion
Beamforming enhances the received signal power by transmitting signals in specific directions. However, in high-speed and dynamic vehicular network scenarios, frequent channel state updates and beam adjustments impose substantial system overhead. Furthermore, real-time alignment between the beam and...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
China Science Publishing & Media Ltd. (CSPM)
2025-08-01
|
| Series: | Leida xuebao |
| Subjects: | |
| Online Access: | https://radars.ac.cn/cn/article/doi/10.12000/JR24242 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Beamforming enhances the received signal power by transmitting signals in specific directions. However, in high-speed and dynamic vehicular network scenarios, frequent channel state updates and beam adjustments impose substantial system overhead. Furthermore, real-time alignment between the beam and user location becomes challenging, leading to potential misalignment that undermines communication stability. Obstructions and channel fading in complex road environments further constrain the effectiveness of beamforming. To address these challenges, this study proposes a multimodal feature fusion beamforming method based on a convolutional neural network and an attention mechanism model to achieve sensor-assisted high-reliability communication. Data heterogeneity is solved by customizing data conversion and standardization strategies for radar and lidar data collected by sensors. Three-dimensional convolutional residual blocks are employed to extract multimodal features, while the cross-attention mechanism integrates integrate these features for beamforming. Experimental results show that the proposed method achieves an average Top-3 accuracy of nearly 90% in high-speed environments, which is substantially improved compared with the single-modal beamforming scheme. |
|---|---|
| ISSN: | 2095-283X |