Distributed synchronization algorithms for wireless sensor networks
Department of Electrical and Computer Engineering
Doctor of Philosophy
The ability to distribute time and frequency among a large population of interacting agents is of interest for diverse disciplines, inasmuch as it enables to carry out complex cooperative tasks. In a wireless sensor network (WSN), time/frequency synchronization allows the implementation of distributed signal processing and coding techniques, and the realization of coordinated access to the shared wireless medium. Large multi-hop WSN's constitute a new regime for network synchronization, as they call for the development of scalable, fully distributed synchronization algorithms. While most of previous research focused on synchronization at the application layer, this thesis considers synchronization at the lowest layers of the communication protocol stack of a WSN, namely the physical and the medium access control (MAC) layer. At the physical layer, the focus is on the compensation of carrier frequency offsets (CFO), while time synchronization is studied for application at the MAC layer. In both cases, the problem of realizing network-wide synchronization is approached by employing distributed clock control algorithms based on the classical concept of coupled phase and frequency locked loops (PLL and FLL). The analysis takes into account communication, signaling and energy consumption constraints arising in the novel context of multi-hop WSN's. In particular, the robustness of the algorithms is checked against packet collision events, infrequent sync updates, and errors introduced by different noise sources, such as transmission delays and clock frequency instabilities. By observing that WSN's allow for greater flexibility in the design of the synchronization network architecture, this work examines also the relative merits of both peer-to-peer (mutually coupled - MC) and hierarchical (master-slave - MS) architectures. With both MC and MS architectures, synchronization accuracy degrades smoothly with the network size, provided that loop parameters are conveniently chosen. In particular, MS topologies guarantee faster synchronization, but they are hindered by higher noise accumulation, while MC topologies allow for an almost uniform error distribution at the price of much slower convergence. For all the considered cases, synchronization algorithms based on adaptive PLL and FLL designs are shown to provide robust and scalable network-wide time and frequency distribution in a WSN.
njit-etd2011-020 (196 pages ~ 1,970 KB pdf)
Please complete this Feedback Form to inform us about your experience using this website. It will assist us in better serving your information needs in the future. Thank You!
Created June 28, 2011