This thesis deals with interference mitigation in multiuser multiple antenna (MIMO) networks. We focus on the broadcast channel (BC) and the interference channel (IC), where the first models point-to-multipoint transmission and the latter independent point-to-point transmission. The considered network scenarios exclude the exchange of user payload data (in the IC case) and joint signal processing among the receivers. Hence, we aim to limit the necessary backhaul communication and prohibit joint MIMO transmission. Data transmission to different users over wireless channels generates interference. We study linear beamforming which allows to separate spatially the transmitted signals of different users. Interference suppression at the receiver side is implemented through linear filtering, avoiding computationally expensive multiuser decoding.
In the first part, we design an improved channel state information (CSI) feedback metric in a MIMO-BC. The data transmission to different users is organized through random beams. The channel observation used to compute the CSI feedback is noisy. Our proposed estimation method is based on the perturbation of the measured channel. We show through simulation that the resulting signal-to-interference plus noise ratio (SINR) feedback can noticeably reduce the outage probability. The increased transmission reliability is achieved at the cost of a minor reduction of the achievable goodput compared to state-of-the-art link quality estimation methods.
The second part is inspired by the seminal work of Cadambe and Jafar on the multiuser interference channel (IC). They recently showed that interference alignment (IA), based only on linear beamforming and receive-site filtering, achieves a sum-rate that scales proportionally to the number of users at high signal-to-noise ratio (SNR), with the number of degrees of freedom (DoF) as the slope.
We attempt to explore the application of IA in practical systems operating at non-asymptotic SNR. To this end, we extend the idealized model thoroughly studied in the IA literature where all desired and interfering signals are of comparable strength and consider spatial networks, with geographical node distributions and distance dependent pathloss. We find closed-form IA solutions for the multiuser MIMO-IC, avoiding long symbol extensions that are necessary to achieve the optimal multiplexing gains in single-antenna systems. Furthermore, in large networks with a numerousness of transmitters, we investigate IA among a finite number nodes, which we group into clusters. This modular cooperation strategy is chosen in order to keep the number of antennas that need to be deployed per node and the IA training overhead finite.
We characterize the resulting uncoordinated interference. We use IA with diversity techniques to circumvent link outages and show that the overall network area spectral efficiency can be increased, albeit fewer users per cluster must be accommodated in the IA procedure.