Vision-Aided mmWave Beam and Blockage Prediction in Low-Light Environment

Vision-aided beam and blockage prediction schemes have attracted significant attention in millimeter wave (mmWave) communication systems as they can save training overhead and wireless resource waste compared to traditional methods. However, it is hard to maintain prediction accuracy in complex visual environments, especially in a low-light environment. To address this issue, we propose two methods based on curriculum training to enhance the performance of beam prediction and blockage prediction in low-light scenarios. Based on the real-world dataset, DeepSense 6G, the proposed approaches are validated to outperform the baseline algorithms.