Abstract:Millimeter wave (mmWave) communication adopts narrow beams to combat severe propagation loss, resulting in unbearable training duration and overhead. To address this problem, this paper proposes a beam training scheme based on a multimodal fusion network. We investigate the feature fusion process of mmWave channel power leakage (CPL) and sub-6GHz channel state information (CSI) to provide richer feature information. Compared with the uni-modal counterparts (mmWave only or sub-6GHz only), the complementarity and interactivity between different modalities make the proposed network more accurate and robust. Specifically, we propose a multimodal fusion network based on the instantaneous signal received by the mmWave wide beam and sub-6GHz CSI to estimate the optimal narrow beam. In addition, the additional mmWave narrow beam decided by the predicted probabilities is employed to calibrate beam direction further. Moreover, two optimization strategies are presented based on the proposed multimodal fusion network. Firstly, to improve the robustness to noise in wide beam training, the long and short memory (LSTM) network is employed to track the motion direction of the user and calibrate the mmWave beam. And the wide beam for training is selected according to the previous prediction probability to reduce the training overhead. Secondly, since beam training focuses on partially specific information, to focus on its related features, we introduce the self-attention mechanism to improve the global consistency of the network. Numerical results illustrate that the proposed scheme can achieve significantly more accurate results with less overhead than traditional and existing deep learning-based methods.

Beam Prediction for Mmwave Massive MIMO Using Adjustable Feature Fusion Learning.

Enhancing Mmwave Beam Prediction Through Deep Learning with Sub-6 GHz Channel Estimate

Multi-modal Fusion for Millimeter-Wave Communication Systems: A Spatio-Temporal Enabled Approach

Deep Learning Assisted Mmwave Beam Prediction with Prior Low-frequency Information

A Deep Learning-Based Spatial Domain Beam Prediction in Massive MIMO System

Deep Learning Based Joint Beam Selection and Precoding Design for Mmwave Systems with Lens Arrays

Multimodal Fusion Assisted mmWave Beam Training in Dual-Model Networks

Machine Learning for Beam Alignment in Millimeter Wave Massive MIMO

Deep Learning Assisted mmWave Beam Prediction for Heterogeneous Networks: A Dual-Band Fusion Approach

FusionNet: Enhanced Beam Prediction for Mmwave Communications Using Sub-6 GHz Channel and a Few Pilots

FusionNet: Enhanced Beam Prediction for mmWave Communications Using Sub-6GHz Channel and A Few Pilots

Incremental Collaborative Beam Alignment for Millimeter Wave Cell-Free MIMO Systems

Advancing Multi-Modal Beam Prediction with Multipath-Like Data Augmentation and Efficient Fusion Mechanism

Fast Beam Switching Based on Machine Learning for MmWave Massive MIMO Systems

Low-Overhead Beam Selection for mmWave Massive MIMO Systems by Deep Learning

DNN-Aided Based Beamforming for FDD Millimeter-Wave Massive MIMO Systems under Multipath

Multimodal Deep Learning Empowered Millimeter-Wave Beam Prediction

Hybrid Beamforming/Combining for Millimeter Wave MIMO: A Machine Learning Approach

Deep Learning for Beam Training in Millimeter Wave Massive MIMO Systems

A Two-stage Prediction-based Beam Selection Algorithm in MmWave Massive MIMO Systems