Throughput enhancements in multiuser downlink systems using imperfect feedback
Abstract
With the increasing demand for higher data rates in wireless communication systems, there is concern that traditional approaches to the physical layer design will not be capable of providing such enhancements. For this reason, new non-traditional approaches to the design of wireless communication systems need to be adopted. One recent area of research that is promising in that direction is the use of feedback in channel coding to increase throughput in multiuser downlink systems. Two types of feedback that have not been fully explored in multiuser downlink systems are: Channel Output Feedback (COF) and Network-Coded Hybrid Automatic Repeat Request (NC-HARQ) feedback. In COF, the transmitter has access to all previous channel outputs of each receiver through feedback links from each receiver to the transmitter. In NC-HARQ, the transmitter has ARQ feedback from each receiver to the transmitter, and since transmission is packet-based, network coding principles are applied. Both types of feedback have been shown to provide promising throughput enhancements in multiuser downlink systems under perfect feedback. However, the assumption of perfect feedback is far from practical. In this work, we explore these two types of feedback under the presence of feedback imperfections. For both types of feedback, we present schemes that operate under imperfect feedback and analyze their performance to show that significant throughput enhancements can still be obtained. For COF, we consider the additive white Gaussian noise (AWGN) broadcast channel (BC) with AWGN feedback channels from each receiver to the transmitter. Under this setup, we show an achievable rate region for noiseless feedback of linear processing based coding schemes, which include the majority of the currently proposed schemes in the literature, to be arbitrarily closely achieved with sufficiently small feedback noise powers. This result was obtained by designing a concatenated coding scheme using linear processing schemes as inner codes. Then, we present a practical realization of the inner linear processing scheme for the K-user symmetric case with noisy feedback and optimized it for use in the concatenated coding scheme. For NC-HARQ, we consider the two-user symmetric block fading AWGN-BC with imperfect ARQ feedback. Under this setup, we show through analysis and simulations that significant throughput improvements over basic HARQ can still be obtained for moderate-to-high signal-to-noise ratio (SNR); however, basic HARQ outperforms NC-HARQ for low SNR. Probability of packet loss is shown to be worse in NC-HARQ compared to basic HARQ and more pronounced for low SNR, especially for low feedback noise. We also present and analyze a new NC-HARQ scheme that limits the number of NC packets to only one per data packet. The latter scheme is shown to attain most of the throughput of the original NC-HARQ scheme.
Degree
Ph.D.
Advisors
Love, Purdue University.
Subject Area
Information Technology|Electrical engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.