Lattice Network Coding in Distributed Massive MIMO Systems

In this thesis, the uplink of distributed massive MIMO where a large number of distributed access point antennas simultaneously serve a relatively smaller number of users is considered. Lattice network coding (LNC), which comprises compute and forward (C&F) and integer forcing (IF), is employed to avoid the potentially enormous backhaul load. Firstly, novel algorithms for coefficient selection in C&F are proposed. For the first time, we propose a low polynomial complexity algorithm to find the optimal solution for the complex valued case. Then we propose a sub-optimal simple linear search algorithm which is conceptually sub-optimal, however numerical results show that the performance degradation is negligible compared to the exhaustive method. The complexity of both algorithms are investigated both theoretically and numerically. The results show that our proposed algorithms achieve better performance-complexity trade-offs compared to the existing algorithms. Both algorithms are suitable for lattices over a wide range of algebraic integer domains. Secondly, the performance of LNC in a realistic distributed massive MIMO model (including fading, pathloss and correlated shadowing) is investigated in this thesis. By utilising the characteristic of pathloss, a low complexity coefficient selection algorithm for LNC is proposed. A greedy algorithm for selecting the global coefficient matrix is proposed. Comprehensive comparisons between LNC and some other promising linear strategies for massive MIMO, such as small cells (SC), maximum ratio combining (MRC), and minimum mean square error (MMSE) are also provided. Numerical results reveal that LNC not only reduces the backhaul load, but also provides uniformly good service to all users in a wide range of applications. Thirdly, the inevitable loss of information due to the quantisation and modulo operation under different backhaul constraints are investigated. An extended C\&F with flexible cardinalities is proposed to adapt to the different backhaul constraints. Numerical results show that by slightly increasing the cardinality, the gap between C\&F to the infinite backhaul case can be significantly reduced.