Slotless Protocols for Neighbor Discovery in Wireless Networks

点击翻译按钮获取中文摘要

Within the last decade, networks which do not require any fixed infrastructure (mobile ad-hoc networks, MANETS) have become widespread. Before two devices in a MANET can start communicating, they have to discover their mutual existance and synchronize their clocks using a procedure called neighbor discovery. Since all participants of a MANET are powered by batteries, power-efficient neighbor discovery is a crucial requirement. In such protocols, a device repeatedly broadcasts packets on a certain set of channels, whereas other devices repeatedly switch on their receivers for short amounts of time. To save energy, the devices sleep in the meantime. Two devices have discovered each other successfully once both have received a first packet from their opposite. Most common neighbor discovery protocols are referred to as slotted protocols. They subdivide time into multiple, equal-length intervals, called slots. In some slots, the device remains asleep, whereas other ones are active slots used for communication. The discovery procedure is complete once two active slots overlap in time. The pattern of active and sleep slots is defined by a specific schedule. The protocol design problem from the perspective of neighbor discovery is to identify schedules that lead to low energy consumptions, while ensuring low-latencies. In the last years, schedules of active/passive slots, which guarantee deterministic worst-case bounds on the discovery latencies have been an active field of research. In this project, we consider a different paradigm which may be characterized as slotless. In protocols following this paradigm, the transmitting device send packets at periodic intervals, with the transmission duration determined by the size of a packet. The other device switches on its receiver also at periodic intervals and remains on for a specified duration during each period. The lengths of the periodic intervals as well as the on-durations of the receiver are drawn over continuous time. Our preliminary work indicates that for almost all parametrizations of these protocols, deterministic latency bounds can be guaranteed and certain choices of interval lengths lead to significantly shorter discovery-latencies than the best known slotted protocols, while consuming the same amount of energy. The main goal of this project is to develop analytical models for slotless protocols and, based thereupon, to develop novel protocols with higher performances than all existing ones. These protocols are then implemented to verify their behavior in real-world networks.