# Site-specific online compressive beam codebook learning in mmWave vehicular communication

arXiv: https://arxiv.org/abs/2005.05485 · HTML: https://mlsanity.com/paper/2005.05485/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 11 May 2020.

## Built on by (most cited first, 5 of 5)

- [Real-time Outdoor Localization Using Radio Maps: A Deep Learning Approach](https://mlsanity.com/paper/2106.12556/) (2021)
- [LocUNet: Fast Urban Positioning Using Radio Maps and Deep Learning](https://mlsanity.com/paper/2202.00738/) (2022)
- [Massive MIMO Beam Management in Sub-6 GHz 5G NR](https://mlsanity.com/paper/2204.06064/) (2022)
- [Deep Learning-based Compressive Beam Alignment in mmWave Vehicular Systems](https://mlsanity.com/paper/2103.00125/) (2021)
- [Hierarchical ML Codebook Design for Extreme MIMO Beam Management](https://mlsanity.com/paper/2312.02178/) (2023)

## Abstract (arXiv)

Millimeter wave (mmWave) communication is one viable solution to support Gbps sensor data sharing in vehicular networks. The use of large antenna arrays at mmWave and high mobility in vehicular communication make it challenging to design fast beam alignment solutions. In this paper, we propose a novel framework that learns the channel angle-of-departure (AoD) statistics at a base station (BS) and uses this information to efficiently acquire channel measurements. Our framework integrates online learning for compressive sensing (CS) codebook learning and the optimized codebook is used for CS-based beam alignment. We formulate a CS matrix optimization problem based on the AoD statistics available at the BS. Furthermore, based on the CS channel measurements, we develop techniques to update and learn such channel AoD statistics at the BS. We use the upper confidence bound (UCB) algorithm to learn the AoD statistics and the CS matrix. Numerical results show that the CS matrix in the proposed framework provides faster beam alignment than standard CS matrix designs. Simulation results indicate that the proposed beam training technique can reduce overhead by 80% compared to exhaustive beam search, and 70% compared to standard CS solutions that do not exploit any AoD statistics.