Millimeter-Wave Beam Prediction with Inverse Beamforming ML Model
Résumé
Next-generation wireless systems rely on millimeter-wave (mmWave) frequencies for high bandwidth, but their propagation is challenged by path loss and environmental blockages, affecting reliability. Beam management, including beam search for link establishment or recovery after link failures and beam steering for dynamic adjustments, is essential to maintaining robust communication in dynamic environments. Accurate and timely prediction of the Angle of Arrival (AoA) is critical for efficient beam steering, which ensures robust communication links in mmWave systems. Traditional methods focus on estimating the current AoA. While they are effective in static or low-mobility scenarios, these reactive approaches struggle in scenarios with high user mobility. Estimating the current AoA requires frequent adjustments, leading to increased latency and a higher risk of connection outages. In contrast, predicting the future AoA offers a proactive solution, enabling user equipment (UE) to anticipate the next beam direction. ensuring smoother transitions and reduced latency in dynamic environments. This work focuses on the beam steering phase at the UE side with limited resources, ensuring real-time operation within hardware and energy constraints. It aims to predict future AoA using prior ones and channel observations rather than estimating the current one. As a baseline for comparison, we introduce an EKF-based solution. Recent works have focused on using EKF techniques to address challenges such as beam tracking at both ends in line of sight conditions, and handling multipath channels dominated by a single line of sight path [2]. The EKF solution in this study serves primarily as a benchmark.
| Origine | Fichiers produits par l'(les) auteur(s) |
|---|---|
| Licence |