Enhancing LIDAR-based mm Wave Beam Selection with AugMix Augmented CNNs.

In the task of millimeter wave (mmWave) beam selection, challenges abound due to the narrow beamwidth of mm Wave signals and the high mobility of users. Crucially, the integration of light detection and ranging (LIDAR) data offers a promising avenue to enhance the precision of beam selection. In this study, we harness LIDAR data in tandem with augment and mix (AugMix), a data processing technique, to advance mm Wave beam selection. By incorporating AugMix into the training of a convolutional neural network (CNN) model, we enhance model robustness and reduce uncertainty. AugMix, coupled with LIDAR, creates an augmented data manifold richer than those from standard techniques, bolstering the model's generalization capabilities. Our methodology details the synergistic use of AugMix data augmentation and LIDAR data for a CNN-based beam selection model. Performance evaluations showcased its superiority in accuracy and throughput compared to benchmark models. Our research findings contribute to enhancing the robustness and accuracy of beam selection in mm Wave communication systems.