Analysis of Structured and Un-Structured Network Protocols for Data Aggregation Over Distributed Wireless Sensor Networks.

摘要

The focus of this thesis is on design and evaluation of one-shot data aggregation protocols for static and mobile wireless sensor networks (WSNs). The goal in one-shot data aggregation is to compute a statistical summary of sensor data such as max, average, sum, count and min, when initiated by a special node such as the base station. WSNs have wide range of applications in both static and mobile/dynamic systems. Static sensor networks are especially useful when monitoring is required in harsh, inaccessible environments and when the region to be monitored is really large. Examples of static sensor network applications include environmental monitoring systems, monitoring of industrial control systems, monitoring of degradation in slagging gasifiers, distributed object detection and tracking. Example of mobile applications include vehicular ad-hoc networks and networks of personal radios used in emergency dispatch and battlefields.;For data aggregation in static networks with stable links, structured approaches such as spanning trees are generally preferred. This is because, once a data aggregation structure has been established, link topologies remain fixed and there is minimal need to actively maintain and change the routing structures. In this thesis, one such tree based data aggregation protocol has been designed and evaluated using simulations in networks ranging from 100-1000 nodes. The protocol has also been implemented at a smaller scale in the context of a smart refractory environment, where slag penetration in gasifiers is remotely monitored using smart bricks that are embedded with sensors. In mobile networks and networks with frequent link changes, topology driven structures are likely to be unstable and to incur a high communication overhead. Therefore, self-repelling random walks have been recently proposed as an attractive alternative for data aggregation in mobile systems. In this thesis, a brief overview of random walk based data aggregation has been presented and systematic evaluation of tree based and random walk based data aggregation protocols in networks ranging from 100-1000 nodes under varying degrees of node mobility has been done. The conditions under which unstructured protocols become more attractive in terms of convergence time and messaging efficiency as compared to tree based structured approaches have been quantified.