Decentralised Distributed Massive MIMO

In this thesis, decentralised distributed massive multiple-input multiple-output (DD-MaMIMO) is considered for providing high spectral efficiency (SE) per user. In the DD-MaMIMO system, a large number of access points (APs) within a coordination region are connected to an edge processing unit (EPU) via fronthaul links, serving the users within a service region. Initially, we investigate a DD-MaMIMO system with perfect fronthaul links and assume that the processing takes place in the EPU. To demonstrate the improved SE, we compare our proposed architecture to cell-free MaMIMO. Furthermore, we discuss the scalability of DD-MaMIMO and give its definition. Secondly, we extend our research to the limited-capacity fronthaul links which is essential in practice. To model the limited-capacity fronthaul links, we adopt the Bussgang decomposition to express the quantisation. We propose two strategies for obtaining channel state information (CSI): estimate-and-quantise (EQ) and quantise-and-estimate (QE). Particularly, in the QE scheme, we derive the closed-form expressions of Bussgang decomposition coefficients for the non-Gaussian distribution input of the quantiser, as the elements of pilots follow complex Gaussian distribution. Both CSI acquisition strategies are analysed with respect to the mean square error (MSE) of channel estimation. Finally, we explore the processing which happens at the AP which is the local estimation in DD-MaMIMO. Here, two approaches are exploited for data decoding at the EPU: simply averaging decoding and large scale fading decoding. We further compare the local estimation scheme with the decentralised processing scheme. The scalability is also discussed as the channel estimation and data detection happens at the AP.