RIS-Aided Joint Channel Estimation and Localization at mmWave Under Hardware Impairments: A Dictionary Learning-Based Approach

Reconfigurable intelligent surface (RIS)-aided millimeter wave (mmWave) wireless systems offer robustness to blockage and enhanced coverage. In this paper, we develop an algorithmic solution that shows how RISs can also enhance the positioning performance in a joint localization and communication setting, even when hardware impairments are considered. We propose a realistic system architecture that considers the clock offset between the transmitter and the receiver, impairments at transmit and receive arrays, and mutual coupling between the RIS elements. We formulate the estimation of the composite channel in a RIS-aided mmWave system as a multidimensional orthogonal matching pursuit problem, which can be solved with high accuracy and low complexity, even when operating with large antenna arrays as required at mmWave. In addition, we introduce a dictionary learning stage to calibrate the hardware impairments at the user array. To complete our design, we devise a localization scheme that exploits the estimated composite channel while accounting for the clock offset between the transmitter and the receiver. Numerical results show how RIS-aided mmWave systems can significantly improve the localization accuracy in a realistic 3D indoor scenario simulated by ray tracing.

I. Introduction

Multiple-input multiple-output (MIMO) communication utilizing large arrays and high bandwidths, such as those employed in millimeter wave (mmWave) bands, enables high data rates for wireless links while increasing angle and delay resolvability during channel parameter estimation. Moreover, the sparsity of the channels enhances the sensing performance of the communication waveform [2]. Despite these advantages, state-of-the-art solutions for mmWave localization still face challenges in achieving the required accuracy for specific use cases that depend on precise position information. Current approaches include ultra-dense deployments for beam-based coordinated measurements from multiple base stations (BSs) [3], [4], [5], deep networks exploiting power delay profiles or other channel parameters [6], [7], [8], [9], and two-stage methods that involve a sparse channel estimation phase and a subsequent stage that maps the estimated channel parameters to the position and orientation of the user by exploiting geometric relationships [10], [11], [12], [13], [14], [15], [16], [17].