Beam Prediction in mmWave V2I Communications with Quantum Transformer Networks

In this paper, we introduce hybrid quantum-transformer networks (QTNs) for predicting the optimal beam in vehicle-to-infrastructure (V2I) systems operating in the millimeter-wave (mmWave) band. Large antenna arrays and narrow beams are essential in high-frequency V2I communications, but they also introduce substantial beam training overhead. To address this, We leverage the 60 GHz channel data collected at the base station and the GPS positions from the user equipment to train deep learning models that intelligently predict the optimal beam for maximizing signal power, evaluating their effectiveness in real-world V2I scenarios using the DeepSense 6G dataset and comparing the results with existing methods. The experimental outcomes reveal a distance-based accuracy (DBA) score of 0.8832 for positional data. These QTN models demonstrate both low complexity and high performance, underscoring their potential in addressing the challenges of beam management in mmWave V2I systems.