Distributed Beam Alignment in sub-THz D2D Networks

TL;DR

This paper introduces a novel, efficient method for distributed beam alignment in sub-THz D2D networks using compressed sensing and a unique pilot design, significantly reducing overhead and enabling scalable network-wide alignment.

Contribution

It proposes a new pilot sequence design and a sequential partitioning strategy that achieve logarithmic overhead scaling for beam alignment in large D2D networks.

Findings

Reduces pilot overhead compared to baseline methods.

Enables simultaneous multi-device alignment with orthogonal frequency pilots.

Channels become frequency flat after beamforming, suitable for single carrier transmission.

Abstract

Devices in a device-to-device (D2D) network operating in sub-THz frequencies require knowledge of the spatial channel that connects them to their peers. Acquiring such high dimensional channel state information entails large overhead, which drastically increases with the number of network devices. In this paper, we propose an accelerated method to achieve network-wide beam alignment in an efficient way. To this aim, we consider compressed sensing estimation enabled by a novel design of pilot sequences. Our designed pilots have constant envelope to alleviate hardware requirements at the transmitters, while they exhibit a "comb-like"' spectrum that flexibly allocates energy only on certain frequencies. This design enables multiple devices to transmit thier pilots concurrently while remaining orthogonal in frequency, achieving simultaneous alignment of multiple devices. Furthermore, we present a sequential partitioning strategy into transmitters and receivers that results in logarithmic scaling of the overhead with the number of devices, as opposed to the conventional linear scaling. Finally, we show via accurate modeling of the indoor propagation environment and ray tracing simulations that the resulting sub-THz channels after successful beamforming are approximately frequency flat, therefore suitable for efficient single carrier transmission without equalization. We compare our results against an "802.11ad inspired" baseline and show that our method is capable to greatly reduce the number of pilots required to achieve network-wide alignment.