Leveraging Location Information for RIS-Aided mmWave MIMO Communications

Abstract

Location information offered by external positioning systems, e.g., satellite navigation, can be used as prior information in the process of beam alignment and channel parameter estimation for reconfigurable intelligent surface (RIS)aided millimeter wave (mmWave) multiple-input multiple-output networks. Benefiting from the availability of such prior information, albeit imperfect, the beam alignment and channel parameter estimation processes can be significantly accelerated with less candidate beams explored at all the terminals. We propose a practical channel parameter estimation method via atomic norm minimization, which outperforms the standard beam alignment in terms of both the mean square error and the effective spectrum efficiency for the same training overhead.

Reconfigurable Intelligent Surface Aided Multi-User Communications: State-of-the-Art Techniques and Open Issues

“…Content may change prior to final publication. Derivation of exact end-to-end SNR expressions for the RISaided system and AF relay system [68] Provide a two-stage positioning with dual RIS in mmWave MIMO systems [69] Introduce the cross-entropy method to maximize the sum-rate performance of RIS-aided indoor THz transmission [70] Discuss the coverage of an RIS-assisted large-scale mmWave cellular network using stochastic geometry [71] Present a robust technique for mitigating quantization errors in low resolution mmWave RISs under plane-wave illumination using random phase delays [72] Investigate an IRS-aided multi-user THz MIMO system with OFDMA, where the sparse radio frequency chain antenna structure is adopted for reducing the power consumption [73] Formulate the closed-form expression of uplink achievable rate expression of RIS-aided mmWave systems [74] Study the enhancement of network reliability and connectivity of mmWave RISs-aided communication system in the presence of random blockages [75] Discuss Geometric Mean Decomposition (GMD)-based beamforming for RIS-aided wideband hybrid mmWave MIMO systems [76] Study the ranging performance of RIS-aided mmWave systems in the presence of blockages [77] Present a design of a practical beam alignment and channel parameter estimation method for RIS-aided mmWave MIMO networks via atomic norm minimization [78] Discuss a holographic version of an RIS and its use in THz mMIMO systems [79] Present the optimal RIS placement in mmWave networks reduction, channel acquisition, beamforming design, and user localization strategies [67].…”

“…To design a practical beam alignment and channel parameter estimation method for single-user RIS-aided mmWave MIMO networks, the results in [77] show that the proposed method outperforms the standard beam alignment in terms of mean square error and spectrum efficiency. To reduce the pilot overhead, Compressive Sensing (CS)-based algorithm is developed that exploits the dual sparsity of THz MIMO channels in both the angular and delay domain [78].…”

Active Sensing for Communications by Learning

“…For example, [18] adopts a scheme with random phase shifts at the RIS over multiple pilot transmissions. To further improve the performance, [19] proposes to leverage location information to learn a better design for the RIS reflection coefficients in the channel estimation phase, but it is not a data-driven approach. In this paper, we make a case that a data-driven approach can result in a significantly better design of the reflection coefficients in the uplink channel estimation phase.…”

“…Further, for the downlink precoder design problem in which the output of the final DNN is also a beamforming vector, we set the activation function σ (v) R (•) to be one of the normalization functions in (18) or (19), depending on the beamforming constraint. Finally, for the AoA estimation problem where the output is the estimated AoA vector, we employ a linear activation layer for σ…”

“…To use the existing deep learning libraries that only support real-value operations, we set y t := [ (ỹ t ), (ỹ t )] as the input vector to the deep active learning unit, w t := w t , w t as the output of the DNN in the active learning unit, and v := ṽ , ṽ as the final control vector designed by the proposed framework. To ensure that the RIS reflection coefficients in the sensing phase and the downlink data transmission phase satisfy the unit modulus constraint, we use an activation function similar to (19) for σ…”