UC San Diego

This dissertation addresses key aspects of opportunistic multiuser scheduling, one of the main components of next generation wireless networks responsible for achieving maximum spectrum efficiency and superior user experience. The main topics include the exploitation of multiuser diversity and spatial frequency reuse while maintaining user fairness, the design and analysis of fair scheduling policies that can provide optimal performance, the utilization of historical channel data in scheduling, and the Quality-of-Service (QoS) guarantee.

First, we consider the scheduling problem in cellular networks which support Device-to-Device Communications (D2D). In a D2D network, additional communication modes including orthogonal and non-orthogonal D2D communications make it very challenging to perform fair resource allocation while maximizing the system performance. Based on the Cumulative Distribution Function (CDF) scheduling approach, we propose two orthogonal scheduling policies that guarantee access fairness among all users while taking advantage of multi-user diversity and local D2D communications. For even higher spectral efficiency, spatial frequency reuse is proposed via our novel Group Fairness Scheduling (GFS) policy, that exploits both spatial frequency reuse and multiuser diversity in order to deliver marked improvements to system performance with perfect fairness among the users.

Next, we examine the optimality of opportunistic scheduling under temporal fairness. In general, obtaining the optimal scheduling gain under user fairness or other QoS criteria for a heterogeneous environment is very difficult. Two main methods are proposed to obtain the optimal scheduling solutions under temporal fairness. In the first method, we derive linear program-based scheduling (LPS) algorithms using a window of past Channel State Information (CSI) to compute the scheduling decisions that can approach the optimal policy as the window size gets large. In the second method, we cast the scheduling problem as one of statistical classification and propose a novel supervised classification-based scheduling (SCS) framework for obtaining optimal scheduling decision boundaries.

Finally, we address one of the key QoS aspects of opportunistic multiuser scheduling, the user service delay time. Keeping the service delay time under control while maximizing the scheduling gain subject to some fairness criterion is a very hard problem in general. In order to reduce the service delay, we introduce the concept of delay profile shaping. Based on this concept and large system approximations, we develop a novel delay reduction scheme, the Opportunistic Bernoulli Mixing (OBM) policy, which can support different user channel conditions and resource allocation requirements with an excellent performance/delay tradeoff.