UTILITY-BASED RESOURCE ALLOCATION STRATEGIES AND PROTOCOL DESIGN FOR SPECTRUM-ADAPTIVE WIRELESS NETWORKS

Abstract

Resource allocation strategies, including power control, rate adaptation, and dynamic spectrum access, have been the keys to improving the performance of dynamic (mobile) wireless networks. In this dissertation, we propose several resource optimization schemes for various wireless network architectures, with the goal of maximizing the system throughput and/or minimizing the total energy consumption. These schemes are integrated into the design of distributed medium-access control (MAC) protocols. We propose a game theoretic power control scheme for single-channel ad-hoc networks, and design an efficient MAC protocol, called GMAC, that implements such a scheme in a distributed fashion. GMAC allows for multiple potential transmitters to contend for the channel through an admission phase that these transmitters to determine their appropriate transmission powers. Successful contenders proceed concurrently following the admission phase. We then study the operation of spectrum-agile (cognitive) radios in multi-channel, multi-hop wireless network setting. Two principal cases are considered: exclusive-occupancy and interference-based channel models. For the case of exclusive-occupancy channel models, we design a MAC protocol that exploits the "dual receive" capabilities of the radios to maximize the network throughput. We then propose a cross-layer framework for joint adaptive load/medium access controls. Under this framework, the traffic loads of individual node are adapted based on local MAC parameters. For the case of interference-based channel models, when system throughput is the primary performance metric, we apply "price-based" iterative water-filling (PIWF) algorithms for resource allocation. When energy consumption is the primary metric, we propose a selfish update algorithm and an incentive-based update algorithm for minimizing the power consumption while satisfying the rate and power mask requirements. These algorithms are implemented by having nodes repeatedly negotiate their best power/spectrum to reach a good Nash Equilibrium. An efficient multi-channel MAC protocol is proposed to facilitate the radio negotiation and convergence phase. Simulation results indicate that our proposed protocols achieve significant throughput/energy improvements over existing protocols.