Coexistence of unlicensed band wireless networks in a non-cooperative environment

There has been an exponential increase in the deployment of wireless networks that operate in the unlicensed band, such as the IEEE 802.11 Wireless Local Area Networks (WLANs) and the Wireless Mesh Networks (WMNs). These networks do not require any additional regulatory approval before deployment and co-located networks often belong to different managing entities. Due to their use of the unlicensed band, no single network can claim exclusive use of a channel. Interference may subsequently arise, leading to suboptimal performance. Besides Medium Access Control (MAC) protocols, radio resource control schemes like channel allocation, power control and link adaptation have been proposed to reduce this interference. In this thesis, we are interested in the coexistence issues of such independent unlicensed band networks. Due to the autonomous nature of these networks, they may not cooperate or even use the same MAC protocol. We investigate the use of radio resource control schemes to improve the performance of such co-located networks. Our proposed schemes make use of utility-based techniques that are derived from game theory and optimization theory.We first model the interactions as a non-cooperative game and study the characteristics of the resultant game. We develop channel selection schemes for respectively, independent multihop WMNs and single-hop WLANs that are located together. We show that our proposed schemes improve the performance of non-cooperative WMNs by as much as 36%. In WLANs, our schemes show as high as 30% increase in aggregate throughput when evaluated against two existing channel selection schemes. Subsequently, we investigate how non-cooperation affects the solution of a cross-layer resource allocation algorithm designed for multi-radio, multi-channel, multihop wireless networks. We show that in the presence of non-cooperative networks, there exists efficiency loss due to the incomplete information of the contention environment. As a result, we propose an adaptation to the algorithm that is shown to improve performance by up to 3.2 times for a general physical/link layer model and 21% for a more realistic CSMA model