Quality of service assurances in multihop wireless networks

• Authors: Kuang-Ching Wang;Parameswaran Ramanathan
• Venue: Quality of service assurances in multihop wireless networks
• Year: 2003

Wireless networks are expected to be an integral part of the next generation communication infrastructure with mobile users demanding better coverage, higher rates, and robust connectivity. When users agree to forward packets for each other, multiple hop routes facilitate communication beyond users' radio range. This dissertation addresses the problem of providing diverse end-to-end communication quality of service (QoS) assurances in such multihop wireless networks. Since bandwidth at each node and routes of end-to-end flows change over time, existing QoS mechanisms that require explicit resource allocation and absolute QoS assurances are difficult to realize. Instead, the dissertation proposes a new Neighborhood Proportional Delay Differentiation (NPDD) service model providing multiple classes with consistent proportional rather than absolute QoS. Based on NPDD, an application can choose a satisfactory class to service its packets and adapt its class to combat network dynamics. Algorithms are proposed for dynamic class selection (DCS), scheduling, and medium access priority adaptation to provide NPDD-based end-to-end QoS assurances in IEEE 802.11e like networks. The effectiveness of the proposed algorithms in achieving concurrent end-to-end delay and throughput assurances is evaluated with simulations. Multiple DCS applications in the same network essentially compete with each other by choosing classes high enough relative to others to meet their desirable QoS. Such competitive interactions are modeled as a multi-agent noncooperative game. For various network settings, the conditions for the existence of an equilibrium and the guaranteed convergence to a feasible equilibrium are derived. These findings demonstrate that DCS over NPDD is reliable in delivering diverse QoS assurances among multiple applications and admission control is achievable in such networks. The dissertation concludes by identifying open issues that need further investigation.