PhD Thesis Proposal Defence "Bandwidth Allocation in Wireless Ad Hoc Networks" By Mr. Zuyuan Fang Abstract: Nowadays the emergence of nomadic applications has led to intensive research interest on providing Quality of Service (QoS) in Mobile Ad hoc Networks (MANETs). In the networking protocol stack, a QoS capable MAC protocol plays an important role in a QoS scheme. However, currently the widely used MAC protocol IEEE802.11 Distributed Coordination Function (DCF) does not provide any QoS guarantee and suffers from a problem of fairness. This thesis proposal formulates the design of MAC protocols as a bandwidth allocation problem. A flow contention graph is used to model the contention relations between link flows. Based on the flow contention graph formalism, different algorithms are proposed to solve the problem in such a way that the obtained MAC protocols for MANETs satisfy QoS objectives such as fairness or the minimal bandwidth guarantee. Firstly, a max-min fair allocation algorithm combined with a measurement-based algorithm are presented to guarantee max-min fair distribution of bandwidth across the network at the link layer. In the max-min fair share allocation algorithm, each flow can calculate its max-min fair share based on flow information in two hop diameter in the contention graph. Once each flow knows its fair share and those of its neighbors, the measurement based backoff algorithm is invoked to achieve weighted fairness. Secondly, in order to reduce the communication overhead and meet different fairness objectives, the fair bandwidth allocation problem is also modelled as a general utility based constrained maximization problem, which is shown to admit a unique solution that can only be obtained when global coordination between all nodes is possible. Invoking Lagrange relaxation and duality theory, both a non-cooperative and a cooperative game formulation of the problem are presented. The algorithm derived from the non-cooperative game does not depend on any local flow information. The algorithm derived from the cooperative game depends only on one hop flow information. The fairness objective of these algorithms is controlled by the adopted utility function. Both previous solutions' implementation relies on some variations of the random access mechanism known as carrier sense multiple access. As an alternative, a series of collision-free scheduling algorithms are proposed. In these algorithms, with the help of one hop local flow information, one flow can determine a local collision-free transmission schedule. Comparisons with other schemes proposed in the literature show that the max-min fairness scheme and the cooperative game framework based algorithm not only can provide better fairness but also have the capability of quantifying the bandwidth, which is a very important factor to do admission control. The collision-free algorithms have the advantage of guaranteeing a minimal throughput. These algorithms show that collecting local contention information helps to design effective MAC schemes. However, they are not suitable for highly mobile networks. Since the algorithms derived from the non-cooperative game do not depend on any flow information, they are more suitable for such networks. Date: Friday, 30 January 2004 Time: 11:00a.m.-1:00p.m. Venue: Room 2406 lifts 17-18 Committee Members: Prof. Brahim Bensaou (Supervisor) Prof. Jogesh Muppala (Chairman) Prof. Gary Chan Prof. Danny Tsang (ELEC) **** ALL are Welcome ****