Efficient Cross Layer Designs for IEEE 802.11 Wireless Networks

Abstract

Various properties of wireless networks, such as mobility, frequent disconnections and varying channel conditions, have made it a challenging task to design networking protocols for wireless communications. In this dissertation, we address several problems related to both the routing layer and medium access control (MAC) layer in wireless networks aiming to enhance the network performance. First, we study the effect of the channel noise on the network performance. We present mechanisms to compute energy-efficient paths in noisy environments for ad hoc networks by exploiting the IEEE 802.11 fragmentation mechanism. These mechanisms enhance the network performance up to orders of magnitude in terms of energy and throughput. We also enhance the IEEE 802.11 infrastructure networks with a capability to differentiate between different types of unsuccessful transmissions to enhance the network performance. Second, we study the effects of the physical layer capture phenomena on network performance. We modify the IEEE 802.11 protocol in a way to increase the concurrent transmissions by exploiting the capture phenomena. We analytically study the potential performance enhancement of our mechanism over the original IEEE 802.11. The analysis shows that up to 35% of the IEEE 802.11 blocking decisions are unnecessary. The results are verified by simulation in which we show that our enhanced mechanism can achieve up to 22% more throughput. Finally, we exploit the spatial reuse of the directional antenna in the IEEE 802.11 standards by developing two novel opportunistic enhancement mechanisms. The first mechanism augments the IEEE 802.11 protocol with additional information that gives a node the flexibility to transmit data while other transmissions are in its vicinity. The second mechanism changes the access routines of the IEEE 802.11 data queue. We show analytically how the IEEE 802.11 protocol using directional antenna is conservative in terms of assessing channel availability, with as much as 60% of unnecessary blocking assessments and up to 90% when we alter the accessing mechanism of the data queue. By simulation, we show an improvement in network throughput of 40% in the case of applying the first mechanism, and up to 60% in the case of applying the second mechanism.