Interference management techniques in large-scale wireless networks
In this thesis, advanced interference management techniques are designed and evaluated for large-scale wireless networks with realistic assumptions, such as signal propagation loss, random node distribution and non-instantaneous channel state information at the transmitter (CSIT). In the first part of the thesis, the Maddah-Ali and Tse (MAT) scheme for the 2-user and 2-antenna base station (BS) broadcast channel (BC) is generalised and optimised using the probabilistic-constrained optimisation approach. With consideration of the unknown channel entries, the proposed optimisation approach guarantees a high probability that the interference leakage power is below a certain threshold in the presence of minimum interference leakage receivers. The desired signal detectability is maximised at the same time and the closed-form solution for the receiving matrices is provided. Afterwards, the proposed optimisation approach is extended to the 3-user BC with 2-antenna BS. Simulation results show substantial sum rate gain over the MAT scheme, especially with a large spatial correlation at the receiver side. In the second part, the MAT scheme is extended to the time-correlated channels in three scenarios, in which degrees of freedom (DoF) regions as well as achievability schemes are studied: 1) 2-user interference channel (IC) using imperfect current and imperfect delayed CSIT; 2) K-user BC with K-antenna BS using imperfect current and perfect delayed CSIT; 3) 3-user BC with 2-antenna BS using imperfect current and perfect delayed CSIT. Notably, the consistency of the proposed DoF regions with the MAT scheme and the ZF beamforming schemes using perfect current CSIT consents to the optimality of the proposed achievability schemes. In the third part, the performance of the ZF receiver is evaluated in Poisson distributed wireless networks. Simple static networks as well as dynamic networks are studied. For the static network, transmission capacity is derived whereby the receiver can eliminate interference from nearby transmitters. It is shown that more spatial receive degrees of freedom (SRDoF) should be allocated to decode the desired symbol in the presence of low transmitter intensity. For the dynamic network, in which the data traffic is modelled by queueing theory, interference alignment (IA) beamforming is considered and implemented sequentially. Interestingly, transmitting one data stream achieves the highest area spectrum efficiency. Finally, a distance-dependent IA beamforming scheme is designed for a generic 2-tier heterogeneous wireless network. Second-tier transmitters partially align their interferences to the dominant cross-tier interference overheard by the receivers in the same cluster. Essentially, the proposed IA scheme compromises between enhancing the signal-to-interference ratio and increasing the multiplexing gain. It is shown that acquiring accurate distance knowledge brings insignificant throughput gain compared to statistical distance knowledge. Simulation results validate the derived expressions of success probabilities as well as throughput, and show that the distance-dependent IA scheme significantly outperforms the traditional IA scheme in the presence of path-loss effect.