Channel-Training-Aided Target Sensing for Terahertz Integrated Sensing and Massive MIMO Communications

Abstract

Integrated sensing and massive multiple-input-multiple-output (MIMO) communication (mMIMO-ISAC) at terahertz (THz) bands can provide vast spatial degrees of freedom and abundant bandwidth resources. However, the employment of a massive number of antennas will pose prominent challenges to both target sensing and channel training in THz-mMIMO-ISAC. In this article, our goal is to integrate the target sensing functionality into the channel estimation stage and develop a channel-training-aided target sensing framework to facilitate the efficient resource sharing of THz-mMIMO-ISAC. Specifically, by exploiting the sparse characteristics of THz mMIMO channels, we build up the intrinsic connection between the channel parameters and the target parameters in angular, delay, and Doppler dimensions. Then, we propose a shared channel training pattern accommodating the hybrid architecture constraints of THz transceiver. Both the channel estimation and the target sensing can be formulated as two structured tensor decomposition problems and then concurrently addressed at the UE and BS sides, respectively. Next, we propose a tensor-based parameter estimation algorithm to acquire the target and channel parameters, where the associated angles of arrival/departure, time delays, Doppler shifts, and coefficients can be extracted from the estimated factor matrices. In addition, we present the detailed derivation of the Cramér-Rao bound (CRB) for the considered parameter estimation problem in THz-mMIMO-ISAC. Numerical results demonstrate that the proposed algorithm can achieve the target parameters estimation performance close to their corresponding CRB, and recover the high-dimensional THz mMIMO channels with substantially reduced training overhead.