Limited Feedback Design for Multiuser Networks

OF THE DISSERTATION Limited Feedback Design for Multiuser Networks By Xiaoyi Liu Doctor of Philosophy in Electrical and Computer Engineering University of California, Irvine, 2017 Professor Hamid Jafarkhani, Chair In this dissertation, the potential of limited feedback in multiuser/multinode networks is explored, and our goal is to design efficient and practical quantizers to mitigate the performance loss brought by limited feedback. For the multiple amplify-and-forward relay network, we propose variable-length quantizers (VLQs) with random infinite-cardinality codebooks in contrast to the fixed-length quantizers (FLQs) with finite-cardinality codebooks that cannot attain the full-channel-state-information (full-CSI) performance. We validate through both theoretical proofs and numerical simulations that the proposed VLQs can achieve the fullCSI outage probabilities with finite average feedback rates. We also apply the idea of VLQ to the multicast network, and show that the global VLQ can achieve the minimum full-CSI outage probability with a low average feedback rate. For the two-user interference network where interferences are treated as noise, we introduce the idea of cooperative quantization to allow multiple rounds of feedback communication in the form of conferencing between receivers. For both time-sharing and concurrent transmission strategies, the proposed cooperative quantizers are able to achieve the full-CSI network outage probability of sum-rate and the full-CSI network outage probability of minimum rate, respectively, with only finite average feedback rates. For non-orthogonal multiple access (NOMA) which is recognized as a key technique for 5G, we propose efficient quantizers using variable-length encoding, and prove that in the typical application with two receivers, the losses in the minimum rate xii and outage probability decay at least exponentially with the minimum feedback rate. In addition, a sufficient condition for the quantizers to achieve the maximum diversity order is provided. For NOMA with K receivers where K > 2, the minimum rate maximization problem is solved within an accuracy of in time complexity of O ( K log 1 ) .