Recent years have witnessed the phenomenal penetration rate of wireless networks in our daily lives, ranging from 802.11-based wireless LANs that provide ubiquitous Internet access, to 802.15.4-based wireless sensor networks that carry out various mission-critical tasks such as security surveillance and patient monitoring. However, despite the advances in the field of wireless networking, how to design high-performance wireless networks remains an open problem because of the fundamental challenges of wireless interference and coexistence.Interference is the fundamental factor that limits the link concurrency of wireless networks. Due to the broadcast nature of wireless channel, concurrent transmissions on the same frequency interfere with each other over the air, resulting in lower throughput and higher delivery delay. Handling interference in wireless networks is difficult because of the hidden terminal problem and the exposed terminal problem. Although the former is well studied in existing literature, the later is not, especially in networks where multiple bit rates are available.Another challenge is that interference significantly hinders the coexistence of different wireless technologies. In particular, recent studies show that wireless coexistence is becoming a growing issue due to the unprecedented proliferation of wireless devices in the unlicensed 2.4GHz band. When devices of heterogeneous physical layer operating on the same frequency, interference is more difficult to resolve as devices cannot decode the signals of each other. Moreover, co-existing devices commonly transmit at different powers, which leads to unfair channel usage. The issue is particularly critical to lower power wireless devices. This thesis tackles the fundamental challenges of wireless interference and coexistence to the link layer design of wireless networks. In particular, we identify two problems in the design of existing link layer protocols, and advance the state-of-the-art by offering practical solutions: (1) the rate-adaptive exposed terminal problem where link concurrency cannot be fully exploited by conventional link layers because of they are oblivious to the bit rate diversity; and (2) the blind terminal problem where existing link layer protocols fail to work in co-existing environments due to the heterogeneous physical layer and power asymmetry of co-existing devices. We motivate this research by showing that existing link layer protocols are surprisingly ineffective in handling these problems. Our experiments pinpoint the fundamental reasons of such ineffectiveness and reveal their implications on the design of link layer protocols. We then develop practical solutions to tackle the identified problems. Extensive testbed-based experiments validate the design of proposed solutions, and demonstrate their significant performance gains over existing link layer protocols.
Read
