Ph.D Thesis

Ph.D StudentTan Hwee Pink
SubjectQuality of Service Provisioning in Wireless Networks
DepartmentDepartment of Electrical and Computer Engineering
Supervisors PROFESSOR EMERITUS Moshe Sidi


Wireless scheduling is important for the QoS provisioning in advanced wireless networks. Compared to its wired counterpart, the design of wireless schedulers is a much harder and challenging problem due to the unique characteristics of the wireless channel. While recent work focused on the design of wireless schedulers to meet a given performance objective, our research aims to characterize the QoS performance of a generic wireless scheduler in a downlink centralized scheduling scenario.

We characterize each input-flow in terms of its demand for resources and channel characteristics. We define a channel-independent scheduler that aggregates all input-flows prior to scheduling, and a generic channel-state dependent wireless scheduler that can be abstracted into functional components. We propose novel schemes for these components to achieve trade-offs amongst statistical, higher-order performance metrics. For a Markovian wireless channel, we derive ergodic Markov models for each scheduler, and develop an analytical framework from which performance metrics can be computed efficiently.

Numerical results highlight the performance again achieved with our proposed schemes for the channel-state dependent scheduler. In addition, while flow aggregation achieves better QoS in an uncorrelated channel, the channel-state dependent scheduler is superior in a persistent channel. Hence, we propose a hybrid scheduler that partitions flows according their channel characteristics before applying a suitable scheduling mechanism to each partition. Compared to its component schedulers, the hybrid scheduler achieves superior performance in a channel-heterogeneous scenario.