| Wireless technologies represent a rapidly emerging area of growth and importance for providing anytime, anywhere ubiquitous communications. Compared with their wireline counterparts, wireless networks are constrained by limited bandwidth and power resources. Furthermore, they have to cope with the detrimental effects of time-varying channel fading and various forms of interference. For these reasons, the conventional layered network architecture designed for wireline networks should be modified in order to address the challenges emerging with the broadcast wireless air-interface.; This thesis focuses on cross-layer resource allocation at both link-level (medium access control (MAC) layer) and flow-level (transport layer) for wireless networks. The objective is to provide a unified framework for theoretical analysis as well as to design efficient and distributed algorithms for existing and envisioned applications. This thesis consists of two interrelated thrusts, one focusing on joint medium access and congestion control, and the other on random access.; In the first thrust, joint medium access and congestion control is considered for hybrid networks as well as wireless ad hoc networks. For both cases, novel approaches are developed using tools from the theory of convex/nonlinear optimization and are motivated by economic (utility-based) models to provide optimal benchmarks. The well-established utility maximization framework enables vertical decomposition across the layered network stack, resulting in nicely decoupled and matched schemes at both transport and MAC layers. The overall framework offers a systematic and holistic cross-layer approach for distributed fair resource allocation in wireless networks.; In the second thrust, throughput of random access protocols is improved by capitalizing on physical-layer (PHY) interference cancellation and opportunism. The goal here is to migrate PHY benefits to the MAC layer. A novel random access scheme called SICTA exploits the conventional structure of contention tree algorithms and employs successive interference cancellation (SIC) to resolve collisions. It is shown that the binary SICTA protocol achieves a high throughput 0.693, outperforming the best-known 0.487 tree algorithm. To take advantage of opportunistic transmissions in random access, channel-aware slotted Aloha schemes are designed. Binary scheduling turns out to be throughput optimal and exploits distributed multiuser diversity while guaranteeing user fairness. |
