Joint Location Sensing and Channel Estimation for IRS-Aided mmWave ISAC Systems

Abstract

In this paper, we investigate a self-sensing intelligent reflecting surface (IRS) aided millimeter wave (mmWave) integrated sensing and communication (ISAC) system. Unlike the conventional purely passive IRS, the self-sensing IRS can effectively reduce the path loss of sensing-related links, thus rendering it advantageous in ISAC systems. Aiming to jointly sense the target/scatterer/user positions as well as estimate the sensing and communication (SAC) channels in the considered system, we propose a two-phase transmission scheme, where the coarse and refined sensing/channel estimation (CE) results are respectively obtained in the first phase (using scanning-based IRS reflection coefficients) and second phase (using optimized IRS reflection coefficients). For each phase, an angle-based sensing turbo variational Bayesian inference (AS-TVBI) algorithm, which combines the VBI, messaging passing and expectation-maximization (EM) methods, is developed to solve the considered joint location sensing and CE problem. The proposed algorithm effectively exploits the partial overlapping structured (POS) sparsity and 2-dimensional (2D) block sparsity inherent in the SAC channels to enhance the overall performance. Based on the estimation results from the first phase, we formulate a Cramér-Rao bound (CRB) minimization problem for optimizing IRS reflection coefficients, and through proper reformulations, a low-complexity manifold-based optimization algorithm is proposed to solve this problem. Simulation results are provided to verify the superiority of the proposed transmission scheme and associated algorithms.