Cross-layer Optimization for Video Delivery over Wireless Networks

As video streaming is becoming the most popular application of Internet mo- bile, the design and the optimization of video communications over wireless networks is attracting increasingly attention from both academia and indus- try. The main challenges are to enhance the quality of service support, and to dynamically adapt the transmitted video streams to the network condition. The cross-layer methods, i.e., the exchange of information among different layers of the system, is one of the key concepts to be exploited to achieve this goals. In this thesis we propose novel cross-layer optimization frameworks for scalable video coding (SVC) delivery and for HTTP adaptive streaming (HAS) application over the downlink and the uplink of Long Term Evolution (LTE) wireless networks. They jointly address optimized content-aware rate adaptation and radio resource allocation (RRA) with the aim of maximiz- ing the sum of the achievable rates while minimizing the quality difference among multiple videos. For multi-user SVC delivery over downlink wireless systems, where IP/TV is the most representative application, we decompose the optimization problem and we propose the novel iterative local approxi- mation algorithm to derive the optimal solution, by also presenting optimal algorithms to solve the resulting two sub-problems. For multiple SVC de- livery over uplink wireless systems, where healt-care services are the most attractive and challenging application, we propose joint video adaptation and aggregation directly performed at the application layer of the transmit- ting equipment, which exploits the guaranteed bit-rate (GBR) provided by the low-complexity sub-optimal RRA solutions proposed. Finally, we pro- pose a quality-fair adaptive streaming solution to deliver fair video quality to HAS clients in a LTE cell by adaptively selecting the prescribed (GBR) of each user according to the video content in addition to the channel condi- tion. Extensive numerical evaluations show the significant enhancements of the proposed strategies with respect to other state-of-the-art frameworks.