Projet de recherche doctoral numero :6110

Description

Date depot: 27 juin 2019
Titre: Hybrid Beamforming Techniques for Massive MIMO
Directeur de thèse: Dirk SLOCK (Eurecom)
Domaine scientifique: Sciences et technologies de l'information et de la communication
Thématique CNRS : Non defini

Resumé: As Multi-Input Multi-Output (MIMO) systems allow spatial multiplexing, Massive MIMO systems employ large numbers of antennas at the base station to increase the spectral efficiency of the system and possibly simplify beamforming techniques. With a large number of antennas though, it may not be feasible to have as many RF chains as the antennas due to the increased cost of the number of RF chains required (which includes Analog-to-Digital and Digital-to-Analog converters (ADCs/DACs), power amplifiers and low noise amplifiers). So signal processing techniques called hybrid beamforming have been developed to take care of the case where the number of RF chains is less than the number of antennas. For a multi-cell multi-user MIMO, hybrid digital and analog beamformers need to be derived based on the maximization of weighted sum rate which is formulated as a weighted sum MSE (WSMSE) problem. The analog precoder is assumed to have all elements with unit modulus (or only phase elements) and each of the elements in this matrix is optimized iteratively in an alternating optimization fashion, as also the digital beamformers and auxiliary quantities (receivers and weights). A hybrid beamforming design based on the Weighted Sum Rate (WSR) criterion needs to be considered which is simplified using the Difference of Convex Functions (DC) approach. The advantage of the DC approach is that the iterative algorithm is expected to converge faster. For the analog beamformer, each phase shifting element is derived based on the WSR formulation. To overcome the issue of local optima, a deterministic annealing based approach is proposed for the analog phasors design starting from the optimal unconstrained analog beamformer matrix.



Doctorant.e: Kurisummoottil Thomas Christo