Multiuser diversity and scheduling in wireless communications

The number of wireless subscribers has been growing rapidly and this growth has coincided with and encouraged the development of more bandwidth intensive applications. As a result, modern systems are being designed with more base-stations and smaller cell sizes. The base-station transmit power remains the same as in previous systems and thus inter-cell interference becomes significant. This thesis considers how techniques from information theory and signal processing such as user scheduling, diversity and inter-cell coordination can be used to enhance modern systems by increasing data rates and mitigating interference.This thesis explores the use of opportunistic scheduling to a single user per cell in downlink multicellular wireless systems. Scaling laws are derived for system throughput in the interference limited regime with J interferers, K users per cell, M T transmit and MR receive antennas. The scaling starts at 1J log K for Multiple Input Single Output (MISO) systems and 1J-MR+1 log K for single input multiple output (SIMO) systems and transitions to log log K as the number of users becomes large. Simulation for practical system values indicates that the scaling laws obtained are in fact realistic.The performance of two interfering MISO downlinks is investigated under various assumptions of channel state information at the transmitter (CSIT) and coordination between the base-stations. With only signal channel CSIT, multiuser diversity provides log log K scaling but is unable to benefit from a large signal to noise ratio (SNR). With signal and interference CSIT but no coordination in user selection, a high SNR scaling of 1MT log K is achieved to within a constant in contrast to the log K scaling to within a constant for global CSIT and full coordination. For any SNR, the asymptotic scaling of both of these schemes is log log K + log SNR so that for large enough K, the gap between the schemes disappears and global CSIT and coordination are not necessary.The impact of different relative received interference powers on the system is discussed and the Max Percentile scheduler is proposed to ensure fairness among users. For general values of J, the Max Percentile scheduler yields a signal to interference ratio (SIR) scaling that is asymptotically correct as K &rarr infinity. Next, the Max Percentile scheduler is made to incorporate delay and to be practical. Simulations show that this scheduler and its practical version compare well to the popular Proportional Fair scheduler. An attractive feature of the Delay Constrained Max Percentile schedulers is their amenability to analytical analysis and the fact that each user's performance depends solely on that user's SINR distribution and the total number of users being served by the base-station.Finally, opportunistic scheduling is considered for orthogonal frequency division multiplexing (OFDM) systems when there is channel uncertainty. In particular, the optimal balance between frequency and multiuser diversity is explored. Jointly optimizing the number of coherence bandwidths coded over for every transmission and the choice of user for each transmission results in scheduling gains of more than 10% in the outage rate. It is shown that the per tone capacity is well-approximated by a Gaussian. The Gaussian approximation enables the derivation of analytic scaling laws that are close to the true ergodic capacity for realistic values of K.