Compressive Sensing-Based Beam Alignment Schemes for Time-Varying Millimeter-Wave Channels

This paper considers the implementation of compressive sensing (CS) approaches for beam alignment (BA) in multiuser millimeter wave (mmWave) MIMO systems. We particularly consider wideband time-varying channels in the practical low SNR regime. We examine two different time scales for beam-switching in the BA training phase at both the base station (BS) and the user equipment (UE). We also compare different time scales for running the CS algorithm at the UE, with their corresponding overhead and complexity. We propose an overarching trial-based protocol that re-initializes the BA process at particular times. We also propose a new approach to designing the CS sensing matrix (SM), based on a deterministic construction. Rows of our proposed SM are Kronecker product decomposable, making it ideal for the BA problem. We show that when block-based beam switching is employed in combination with running the CS algorithm Every Epoch (CS-EE), our proposed SM gives superior performance compared to the other approaches. Moreover, our proposed overarching trial-based protocol enhances the performance even further. We also show that running the CS algorithm Every Block (CS-EB) outperforms CS-EE at the cost of higher complexity and overhead.

Automated summary

This paper investigates the implementation of compressive sensing (CS) approaches for beam alignment (BA) in multiuser millimeter wave (mmWave) MIMO systems. The study focuses on wideband time-varying channels in the low SNR regime and compares different time scales for beam-switching and running the CS algorithm. An overarching trial-based protocol and a new deterministic construction for designing the CS sensing matrix (SM) are proposed. The results demonstrate that the proposed SM and trial-based protocol outperform other approaches, while running the CS algorithm every block (CS-EB) has higher complexity and overhead than CS-EE.