Dynamic rate-control and scheduling algorithms for quality-of-service in wireless networks

Abstract

Rapid growth of the Internet and multimedia applications, combined with an increasingly ubiquitous deployment of wireless systems, has created a huge demand for providing enhanced data services over wireless networks. Invariably, meeting the quality-of-service requirements for such services translates into stricter packet-delay and throughput constraints on communication. In addition, wireless systems have stringent limitations on resources which necessitates that these must be utilized in the most efficient manner. In this thesis, we develop dynamic rate-control and scheduling algorithms to meet quality-of-service requirements on data while making efficient utilization of resources. Ideas from Network Calculus theory, Continuous-time Stochastic Optimal Control and Convex Optimization are utilized to obtain a theoretical understanding of the problems considered, and to develop various insights from the analysis. We, first, address energy-efficient transmission of deadline-constrained data over wire-less fading channels. In this setup, a transmitter with controllable transmission rate is considered, and the objective is to obtain a rate-control policy for transmitting deadline- constrained data with minimum total energy expenditure. Towards this end, a deterministic model is first considered and the optimal policy is obtained graphically using a novel cumulative curves methodology. We, then, consider stochastic channel fading and introduce the canonical problem of transmitting B units of data by deadline T over a Markov fading channel. This problem is referred to as the "BT-problem" and its optimal solution is obtained using techniques from stochastic control theory.

(cont.) Among various extensions, specific setups involving variable deadlines on the data packets, known arrivals and a Poisson arrival process are considered. Using a graphical approach, transmission policies for these cases are obtained through a natural extension of the results obtained earlier. In the latter part of the thesis, a multi-user downlink model is considered which consists of a single transmitter serving multiple mobile users. Here, the quality-of-service requirement is to provide guaranteed average throughput to a certain class of users, and the objective is to obtain a multi-user scheduling policy that achieves this using the minimum number of time-slots. Based on a geometric approach we obtain the optimal policy for a general fading scenario, and, further specialize it to the case of symmetric Rayleigh fading to obtain closed-form relationships among the various performance metrics.