Dynamic multicast and time synchronization in wide-area sensor networks
Liang Lun Cheng, Qing Ye · 2008
The event surveillance applications need to collect spatiotemporal event observations from different interested sensing areas. This objective can be achieved by using technologies of Wide-Area Sensor Networks (WASNs), which are emerging wireless networks that have a two-tier infrastructure: (i) the lower tier consists of multiple Wireless Sensor Networks (WSNs) which are deployed to monitor the interested events; and (ii) the upper tier consists of mobile data collectors (such as unmanned automatic vehicles) which move around the sensing areas to forward event queries and reports. The mobile data collectors form the so-called Intermittently Connected Networks (ICNs). The connectivity of this infrastructure is unreliable, because (i) different sensor networks may be deployed in geographically disconnected areas; and (ii) the end-to-end paths may not exist among mobile data collectors due to node mobility. The key challenge is how to handle the dynamic topology variations in disseminating queries and collecting event reports. Also, globally synchronizing clocks in WASNs is required such that the observed events have accurate timestamps. This dissertation is the first one to address these issues in the overall data communication environment of WASNs. The first contribution of our work includes a set of dynamic multicast protocols. On-demand Situation-aware Multicast (OS-Multicast) protocol is proposed for ICNs to distribute event queries to the mobile data collectors which can further forward the queries to the distributed sensor networks. OS-Multicast dynamically builds up the multicast structure according to the current network situations. It forwards data whenever there is a discovered opportunity to reach the destinations. Therefore, OSMulticast improves the successful message delivery ratio by introducing redundant traffic. OS-Multicast doubles the message delivery ratio than DTBR, the first dynamic multicast protocol for ICNs, when the total unavailability of contacts is more than 70% of the overall system operating time. Priced Track and Transmit (PTNT) protocol is proposed to help mobile data collectors gather event reports from the sensors. By distributedly counting how long the data collectors stay within the neighborhood of each sensor, PTNT is capable of tracking the movement of mobile data collectors without requiring GPS information. PTNT also gives system designers the capability of making a tradeoff between the message delivery efficiency and the average message forwarding delays, by controlling how often the queries are flooded. The second contribution of our work involves a set of time synchronization protocols. Double-pairwise Time Protocol (DTP) is proposed for ICNs to keep clocks of mobile data collectors ticking at the same speed. By doubling the synchronization messages and applying a new message filter to estimate the relative clock drift, DTP cuts the average time-synchronization error by about half comparing to NTP-Core, which models the synchronization exchange method of NTP which is the Internet time standard for decades. The idea of DTP can also be directly implemented in the current synchronization architecture of NTP. The LEvel Synchronization using Sender, Adjuster and Receiver (LESSAR) is then proposed to synchronize clocks in WSNs. LESSAR is a lightweight approach to addressing the concern of energy consumption in WSN time synchronization by only requiring the adjusters to conduct two-way time message exchange with the time servers. Thus, LESSAR always uses less synchronization traffic than TPSN to achieve similar time accuracy, which is the default time protocol used in TinyOS, i.e. the operating system of off-the-shelf wireless sensors. In summary, we describe solutions of providing dynamic multicast and time synchronization services in WASNs. Analytical models for performance analyses of some existing solutions and the proposed approaches are also presented in this dissertation.