Centralized Cell-Free Massive MIMO with Low-Resolution Fronthaul

The increasingly new data-hungry applications in our digital society now might no longer be handled efficiently by the current cellular networks. Cell-free massive MIMO network comes to resolve the traditional way of deploying wireless networks by blurring the cell boundaries. The network comprises a large number of access points (APs) which connect the users to a central processing unit (CPU) via fronthauls for coherent transmission and reception. It is expected that this network can provide a uniformly high data rate per user and per unit area. In this thesis, we study a centralized approach to cell-free massive MIMO that can further exploit its potential with considering a practical issue of limited-capacity fronthauls. We develop different schemes as well as strategies that make the centralized approach feasible. Thereby, we propose the use of low-resolution fronthauls and analyse its performance by making use of Bussgang theorem. The first part of this thesis considers a cell-free network with single-antenna APs, where a coarse scalar uniform quantizer is devised as an interface to the fronthauls. In the second part of this thesis, we extend the network to the case of multi-antenna APs, where two different processing schemes at the APs are studied: individual processing and joint processing. For each part, two strategies for acquiring the channel state information (CSI) under low-resolution fronthaul constraint are developed: estimate-and-quantize (EQ) and quantize-and-estimate (QE). We analyse the performance of both strategies and take them into account for deriving the achievable rate of the systems. Moreover, the scalability of the centralized approach is also discussed in terms of fronthaul load and AP processing. In the last part, we propose the use of a lattice vector quantizer at multi-antenna APs for the high-mobility and high-density scenario, in which two procedures for constructing the lattice codebook are developed.