Impact of wireless channels on queue behavior: Modeling, analysis and optimization

Abstract

The quality of service (QoS) experienced by traffic streams in a wireless link is closely related to the dynamics of the wireless channel. Hence, many recent studies have considered cross-layer interaction between the physical layer and the scheduling layer in the wireless system. These studies typically use simplistic channel models like first-order Markov models to represent the wireless channel. Since the higher-layer building blocks are variants of Markov processes, these channel models match the higher-layer setup and simplify cross-layer analysis and optimization. However, first-order Markov models cannot adequately model even simple wireless channel scenarios and thus can lead to erroneous QoS predictions.

This thesis begins by introducing hidden Markov models (HMM) to model the wireless channel behavior in QoS analysis contexts. HMM random processes are general enough to model wireless channel processes adequately. The structure of HMM processes permits the development of exact and approximate methods to analyze QoS performance. Then, a low-complexity approximate QoS analysis technique based on “burstiness” properties of the channel random process is developed. The complexity reduction is achieved by functional mapping of the random process using peakedness functional. A closed form analytic expression is derived for computing the peakedness functional from the HMM parameters of the process. Then, HMM-based channel models are used to formulate a general framework for exact QoS analysis and QoS-constrained cross-layer optimization. An efficient solution technique based on linear programming (LP) techniques is developed to solve this formulation. Further, a lower bound and upper bound for the optimal scheduling problem are developed as LP formulations. To conclude the thesis, the matrix-geometric method is used to provide an alternate QoS analysis technique for a special case of wireless channels.