Cross-Layer Resource Allocation for Scalable Video Over OFDMA Wireless Networks: Tradeoff Between Quality Fairness and Efficiency

This work addresses the tradeoff between quality fairness and system efficiency for scalable video delivery to multiple users over OFDMA wireless networks. We consider a cross-layer optimization framework seeking to maximize the sum-PSNR corresponding to average user rates, subject to relaxed PSNR-fair constraints. More specifically, a pure quality-fairness (PF) problem is solved first to determine the maximum PSNR value obtained by imposing the same PSNR level to all users. Next the constraints in the PF problem are relaxed by allowing the relative difference between the PSNR of each video and the PF PSNR value to be within some range $[0,\sigma ]$. Thus, the parameter $\sigma$ controls the tradeoff between quality fairness and system efficiency. The PF problem is equivalent to the quality fairness problem proposed by Cical and Tralli, which was solved using a vertical decomposition approach. Further, we convert the optimization problem with the relaxed fairness constraints into a convex problem and solve it using established techniques. Our simulation results show that by varying the value of $\sigma$, a wide range, densely populated, of trade-off points between quality fairness and efficiency can be achieved. Additionally, a subjective quality assessment reveals that while the maximum efficiency scheme (ME), i.e., when $\sigma =\infty$ , may compromise the quality of the high demanding videos, the PF scheme may sacrifice the quality of the low demanding videos. On the other hand, by providing a trade-off between PF and ME, the proposed scheme has the potential of finding a middle ground where all users are satisfied.