Analysis and design of cell-free massive mimo systems under spatially correlated fading channels

• Autores: Alberto Álvarez Polegre
• Directores de la Tesis: Ana García Armada (dir. tes.)
• Lectura: En la Universidad Carlos III de Madrid ( España ) en 2021
• Idioma: español
• Tribunal Calificador de la Tesis: Rui Dinis Nascimento (presid.), M. Julia Fernández Getino García (secret.), Carmen Botella Mascarell (voc.)
• Programa de doctorado: Programa de Doctorado en Multimedia y Comunicaciones por la Universidad Carlos III de Madrid y la Universidad Rey Juan Carlos
• Materias:
  • Ciencias tecnológicas
    • Tecnología de las telecomunicaciones
    • Planificación urbana
      • Comunicaciones
• Enlaces
  • Tesis en acceso abierto en: e-Archivo
• Resumen

  • Wireless communications have become a key pillar in our modern society. It can be hard to think of a service that somehow does not rely on them. Particularly, mobile networks are one of the most necessary technologies in our daily life. This produces that the demand for data rates is by no means stopping from increasing. The cellular architecture is facing a crucial challenge under limited performance by interference and spectrum saturation. This involves cell-edge users experiencing poor performance due to the close vicinity of base stations (BSs) using the same carrier frequency. Based on a combination of the coordinated multi-point (CoMP) technique and traditional massive multiple-input multiple-output (MIMO) systems, cell-free (CF) massive MIMO networks have irrupted as a solution for avoiding inter-cell interference issues and for providing uniform service in large coverage areas. This thesis focuses on the analysis and design of CF massive MIMO networks assuming a spatially correlated fading model. A general-purpose channel model is provided and the whole network functioning is given in detail.

    Despite the many characteristics a CF massive MIMO system shares with conventional colocated massive MIMO its distributed nature brings along new issues that need to be carefully accounted for. In particular, the so-called channel hardening effect that postulates that the variance of the compound wireless channel experienced by a given user from a large number of transmit antennas tends to vanish, effectively making the channel deterministic. This critical assumption, which permeates most theoretical results of massive MIMO, has been well investigated and validated in centralized architectures, however, it has received little attention in the context of CF massive MIMO networks. Hardening in CF architectures is potentially compromised by the different large-scale gains each access point (AP) impinges on the transmitted signal to each user, a condition that is further stressed when not all APs transmit to all users as proposed in the user-centric (UC) variations of CF massive MIMO. In this document, the presence of channel hardening in this new architecture scheme is addressed using distributed and cooperative precoders and combiners and different power control strategies. It is shown that the line-of-sight (LOS) component, spatially correlated antennas, and clustering schemes have an impact on how the channel hardens. In addition, we examine the existent gap between the estimated achievable rate and the true network performance when channel hardening is compromised. Exact closed-form expressions for both a hardening metric and achievable downlink (DL) and uplink (UL) rates are given as well.

    We also look into the pilot contamination problem in the UL and DL with different degrees of cooperation between the APs. The optimum minimum mean-squared error (MMSE) processing can take advantage of large-scale fading coefficients for canceling the interference of pilot-sharing users and thus achieves asymptotically unbounded capacity. However, it is computationally demanding and can only be implemented in a fully centralized network. Here, sub-optimal schemes are derived that provide unbounded capacity with much lower complexity and using only local channel estimates but global channel statistics. This makes them suited for both centralized and distributed networks. In this latter case, the best performance is achieved with a generalized maximum ratio combiner that maximizes a capacity bound based on channel statistics only.