Improving performance of wireless networks using anycast

Joohwan Kim, Purdue University

Abstract

The wireless medium is characterized by a highly variable and unreliable environment that could lead to performance degradation in wireless systems, such as large delays, high energy consumption, and low throughput. In this thesis, we are interested in exploiting the broadcast nature of the wireless medium to mitigate the performance degradation caused by the wireless channel. The class of control solutions that exploit the resulting multi-receiver diversity gain are called “anycast.” Essentially, with anycast, if the originally intended receiver is unavailable to relay the packet, the sending node can let another candidate node that successfully overhears the packet forward the packet. This opportunistic behavior can potentially lead to lower delay, higher throughput, and lower energy consumption. These performance gains, however, can be achieved only when the anycast policy is carefully designed. Otherwise, the dynamic routing choices made by the anycast policy could route packets to undesirable paths, resulting in deterioration of network performance. Furthermore, in a decentralized wireless network, the anycast policy must be designed such that every node determines its forwarding policy independently based only on limited local information. In this dissertation, we develop distributed anycast policies for various wireless applications that meet the above requirements. We first study how to design anycast policies for delay-sensitive event-driven wireless sensor networks, where events occur rarely. However, once an event occurs, event information needs to be reported to the command center as soon as possible. We jointly optimize the anycast policy with the sleep-wake scheduling policy to reduce the event reporting delays and to extend the network lifetime. The second application is for a heavy-traffic wireless sensor network, where in-network data aggregation is employed to reduce the multi-hop traffic volume. The anycast policy that we studied can opportunistically forward a packet to a neighboring node that can best compress the packet and hence can achieve the minimum energy consumption. The last application is a multi-hop wireless network with lossy wireless channels, where packets are often lost during transmission. We develop an anycast policy that can increase the reception probability of the packets and achieve the optimal throughput.

Degree

Ph.D.

Advisors

Shroff, Purdue University.

Subject Area

Electrical engineering

COinS