An Efficient Nocturnal Scenarios Beamforming Based on Multi-Modal Enhanced by Object Detection

Jiali Nie1; Yuanhao Cui2; Tiankuo Yu1; Junsheng Mu1; Weijie Yuan2; Xiaojun Jing1

10.1109/GCWkshps58843.2023.10464587

2023-12-08

979-8-3503-7022-5

2023 IEEE Globecom Workshops (GC Wkshps)

4-8 Dec. 2023

Kuala Lumpur, Malaysia

The progress in integrated sensing and communication(ISAC) technologies has facilitated the application of sensing data for beamforming, resulting in a reduction of training overhead. Nevertheless, the diminished visibility during nocturnal scenarios poses a significant impact on beamforming performance. In this research, we proposed a machine-learning approach that relies on object detection and multimodal fusion to achieve efficient beamforming prediction by leveraging visual and positional data collected from nighttime vehicle communication scenarios. Experimental findings reveal that our developed model achieves the top-1 accuracy exceeding 60% and top-5 accuracy approaching 100%, all the while substantially mitigating the training overhead.

其他

1.Beijing University of Posts and Telecommunications
2.Southern University of Science and Technology

Jiali Nie,Yuanhao Cui,Tiankuo Yu,et al. An Efficient Nocturnal Scenarios Beamforming Based on Multi-Modal Enhanced by Object Detection[C],2023.