System-level performance analysis of multiple-antenna wireless networks with limited channel state information

Cellular communication systems are interference limited because of frequency reuse. To manage the interference, information about the channel state can be used. Specifically, under the premise of perfect and global channel state information at transmitters (CSIT), prior work showed significant spectral efficiency improvement in many cases of wireless networks by using advanced interference management techniques. Unfortunately, obtaining perfect and global CSIT is infeasible in practice due to finite capacity feedback links and associated overheads. For this reason, characterizing the performance of wireless networks with limited CSIT is important to understand the performance achievable in practical wireless networks. Motivated by this, in this dissertation, I analyze the spectral efficiency of wireless networks with various scenarios of limited CSIT. Leveraging the analytical results, I propose strategies to obtain spectral efficiency gains under limited CSIT.

First, I determine appropriate feedback rate used in a multi-antenna cellular network. Specifically, I analyze the net spectral efficiency, which is defined as the downlink spectral efficiency normalized by the uplink overheads caused by using limited feedback. Subsequently, I obtain the optimal feedback rate to maximize the net spectral efficiency. In addition, I extend this result to a multiple-antenna device-to-device network, where I derive the optimal feedback rate to maximize the net spectral efficiency. Next, I consider a cooperative cellular network and propose a semi-static base station (BS) clustering strategy by exploiting the graph theory. By using the proposed strategy, I show that the same spectral efficiency gain with dynamic BS clustering is achieved while avoiding the associated complexity. I also study the spectral efficiency of K-tier heterogeneous networks with limited feedback. Considering non-cooperative and cooperative heterogeneous networks, I formulate and solve adaptive feedback partition problems to maximize the ergodic spectral efficiency in each case. Last, I assume a spectrum-shared millimeter wave (mmWave) downlink cellular network. I characterize the rate coverage performance assuming that inter-operator BS coordination is used. By using the analytical results, I show that inter-operator BS coordination is valuable when sharing the spectrum with a dense operator in a mmWave cellular network.