Computer vision assisted mmWave beamforming for UAV-to-vehicle links

This paper focuses on the beamforming algorithm for UAV-to-vehicle communications. To deal with high communication overhead caused by beam tracking in high mobility communication scenarios, we utilize the inherent vision functionality of UAV platforms and propose a vision-assisted beamforming framework. We propose to use a deep-learning-based network for vehicle detection. Based on the predicted positions of vehicles, we propose a lightweight beamforming algorithm to save beam tracking overhead. Experiments and simulations are implemented on the UAV detection and tracking (UAVDT) dataset, which shows that the proposed algorithm gains a significant performance on received signal-to-interference-plus-noise ratio (SINR).