Cross-layer optimization of wireless newtworks

Existing wireless network is designed based on the layered network structure, such as physical layer, MAC layer and network layer. The strict layered network structure limits the flexibility of protocol design, and leads to suboptimal network designs. In this dissertation, we investigate the methodology of cross-layer optimization and utilize this methodology to optimize various wireless networks. Specifically speaking, we make three original contributions in this dissertation.;First, we design and implement the first cooperative QoS routing protocol in cooperative multi-hop wireless network. Cooperative communication is a new physical technology, which have the potential to improve QoS support for multi-hop wireless network. The advantages of cooperative communication in the physical layer have been explored. However, its impact to the design of upper-layer protocol, like QoS routing, has not been fully understood. Existing works about cooperative QoS routing did not consider the interference effect among links, and only evaluate their algorithms through simulations. This paper targets at designing and implementing an interference-aware Cooperative QoS routing protocol (CQ-routing) in the real testbed. We formulate the problem of finding cooperative routing path with maximum available bandwidth as an optimization problem, called Coop-routing problem. We prove that the Coop-routing problem is strong NP-hard. We propose both centralized and distributed approximation algorithms to solve the Coop-routing problem. We design and implement a CQ-routing protocol in wireless mesh network testbed and evaluate its performance through both experiments and simulations. The results show that CQ-routing protocol can significantly improve the network performance in terms of available bandwidth and number of admitter flows.;Second, we propose the first multi-link spectrum handoff scheduling algorithms for multi-hop cognitive network. Existing work only considered the problem of minimizing spectrum handoff delay of a single link in single-hop cognitive networks, referred to as the SH-SLSH problem. The objective of minimizing single link's spectrum handoff delay is not applicable for a more challenging problem (referred to as the SH-MLMH problem) in which multiple links perform spectrum handoff in multi-radio multi-hop cognitive networks. The SH-MLMH problem first targets at maintaining the network connectivity so as to keep the communication uninterrupted during spectrum handoff. This results in eliminating the spectrum handoff delay. While maintaining the network connectivity, SH-MLMH problem further aims at minimizing the Total Handoff Completion Time (THCT), which is defined as the time for all the links to finish spectrum handoff. We prove that SH-MLMH problem is NP-hard. We propose both centralized and distributed algorithms to solve it. The simulation results show that our proposed solution can significantly improve the network throughput and reduce the THCT.;Third, we study the non-cooperative channel and bandwidth allocation problem in multi-radio multi-channel wireless network. Existing works ignored two important issues, impact of traffic load to channel's transmission quality, and difference of bandwidth demands for different node pairs. To address these two issues, we extend the problem of non-cooperative multi-radio channel allocation to Non-cooperative Joint Channel and Bandwidth Allocation problem (NJCBA), in which node pairs need to consider not only allocating radios to channels, but also allocating bandwidth to selected channels to maximize its own benefit. We prove that there exist pure Nash Equilibriums (NEs) for the NJCBA game. We also analyze the efficiency of the NEs for NJCBA game, and prove that these NEs can achieve a constant Price Of Anarchy (POA). We design two distributed algorithms, to enable node pairs to converge to a pure NE. The simulation results show that these two algorithms can improve the system throughput by 2 or 3 times compared with a greedy allocation algorithm.