Date of this Version
Multimedia communications, especially real-time video communications, is expected to be the major application of the next-generation wireless networks. However, bringing delay-sensitive and loss-tolerant multimedia services based on the current wireless Internet is a very challenging task. In this dissertation, we address cross-layer optimized wireless multimedia networking from both theoretical and practical perspectives. In the first part of the dissertation, we propose cross-layer optimization frameworks for real-time video communications over wireless networks, where the expected received video quality is adopted as the objective function. With the user-centric objective function, we first study content-aware video communications in single-hop wireless networks by exploring the transmission of video summaries. Then, we investigate the routing issue for real-time video streaming over multi-hop wireless networks. Lastly, we study the performance of video summary transmission over cooperative wireless networks to exploit the spatial diversity of cooperative communications. Extensive theoretical and experimental results demonstrate that significant performance gains are obtained by our solutions. In the second part of the dissertation, we theoretically study the methodology of cross-layer design and optimization. Despite rich literature in cross-layer design and optimization schemes, most current research on cross-layer design has been carried out in various piecemeal approaches and lacks a methodological foundation to gain in-depth understanding of complex cross-layer behaviors. We focus on the quantitative analysis of the interactions among design variables towards to the design objective. The interaction measure is calculated based on the non-additive measure theory with network observation data. We conduct a case study on cross-layer optimized wireless multimedia communications to illustrate the major cross-layer design tradeoffs and validate the proposed theoretical framework. The proposed framework can significantly enhance our capability for cross-layer behavior characterization and provide insights for future design.