Abstract
In this thesis, we consider coordinated beamforming design for a multiple-input single-output (MISO) bursty interference channel. In realistic scenarios, distributed medium access control mechanisms or the decentralized networking protocols across different users contributes to the bursty nature of network traffic. To exploit the potential gain provided by such burstiness, the interference state information needs to be fed back to the transmitters. In practice, errors may occur during such feedback, leading to uncertainties of user traffic in the beamforming design problem. Here, we model such traffic uncertainties via a hidden Markov model (HMM). Assuming that perfect channel state information (CSI) or imperfect CSI is available at transmitters, our goal is to maximize the average system utility subjected to average power constraints under traffic uncertainties. The resulting problem is highly non-convex. We hence apply a series of convex approximation techniques to handle the non-convex problem. We further improve the approximation accuracy by our proposed successive convex approximation (SCA) algorithms (\textbf{HMM-SCA-1} and \textbf{HMM-SCA-2}). In our simulation results, we compared the system utilities achieved by the exhaustive search (\textbf{ES}) method and our proposed \textbf{HMM-SCA-1} when perfect CSI is available at transmitters. We demonstrated that our proposed algorithm provide a low-complexity near-optimal solution for the considered optimization problem. Specifically, \textbf{HMM-SCA-1} is 728 times faster than the \textbf{ES} algorithm on average. Even under traffic uncertainties, our proposed \textbf{HMM-SCA-1} still performs better ($\bf 9.5\%$ improvement) compared with the case where the bursty nature of the interference is not exploited at all (\textbf{Non-bursty-SCA}). For the imperfect CSI case, the simulation showed that our proposed \textbf{HMM-SCA-2} provided a $41.4\%$ improvement over \textbf{Non-bursty-SCA}. Compared with the perfect CSI case, we can observed that the performance improvement is much more significant for the imperfect CSI case. Exploiting bursty traffic is hence critical for a practical wireless communication system when channel estimation errors are considered.