Quality Fairness-oriented Cross-layer Resource Allocation for Scalable Video Delivery over OFDMA Wireless Networks

Abstract

This thesis addresses the problem of scalable video delivery to multiple users over OFDMA wireless networks, where quality fairness and system e ciency are jointly considered. Fairness deals with the provision of a similar video quality to all users, while system e ciency concerns the maximization of the overall received video quality. This problem was recently tackled by Cical o and Tralli who proposed a cross-layer optimization framework with the aim of maximizing the sum of the ergodic rates while minimizing the distortion di erence among multiple videos. The optimization problem was \vertically" decomposed into two subproblems: a source adaptation problem at the application (APP) layer and a resource allocation problem at the medium access control (MAC) layer. An iterative local approximation (ILA) algorithm was proposed to solve the two subproblems iteratively until the optimal solution is obtained. One drawback of the above work is that the APP layer algorithm to solve the source adaptation problem is unnecessarily complex. Moreover, the optimal solution may not be accurate since the adopted semi-analytical rate-distortion (R-D) model used for source adaptation is only an approximation of the empirical R-D data. Our rst main contribution is to overcome the aforementioned two drawbacks. To this end, we propose a quality fairness-oriented cross-layer optimization framework that solves a joint resource allocation and source adaptation (JRASA) problem where the objective is to maximize the sum of the PSNRs while minimizing the PSNR di erence among the received videos. The JRASA problem is equivalent to the aforementioned sum-rate maximization problem and capable of being solved by the ILA approach. On the other hand, it has a di erent formulation which naturally leads to the development of a considerably faster APP layer algorithm based on the bisection search method. Furthermore, we show that the above optimization framework can be extended to solve e ciently the JRASA problem based on accurate empirical R-D models, as well. The solution to the JRASA problem using the empirical R-D models can be used as a benchmark to assess the performance of solutions based on approximate R-D models, such as the semi-analytical R-D model. Our second main contribution is an adjustable quality-fair cross-layer optimization framework, which is able to achieve trade-o s between quality fairness and system e ciency, aspect which was not considered by Cical o and Tralli. Our procedure consists of two steps. First the aforementioned JRASA problem is solved. The second step seeks to maximize the sum of the PSNRs while limiting the absolute value of the relative di erence between the PSNR of each video and the common PSNR value obtained in the rst step. We show that the second problem is a general utility-based resource allocation problem, for which e cient algorithms are available to obtain an almost surely optimal solution. Numerical results show that the proposed quality-fair optimization framework provides signi cantly better performance in terms of quality fairness and the provision of better quality to high-complexity videos with respect to an equal-rate adaptation scheme. Moreover, various trade-o s between fairness and system e ciency can be achieved using the adjustable quality-fair cross-layer optimization framework.