Computer Science Department Past Features

Featured Defense - November 2010

Hady AbdelSalam
PhD, Old Dominion University

 

Bio: Hady received his Bachelor and Master degrees in Computer Science from Alexandria University in 1999 and 2004, respectively. His passion towards the theory of computer science has encouraged him to have his master thesis on the verification of minimum-redundancy prefix codes and Huffman trees. He joined the PhD program in the Computer Science Department at Old Dominion University in Spring 2006. Initially, Hady joined the Policy research group to work in several research projects with IBM and OCCS. Hady's work during this period was reported in several publications in different autonomic management related conferences and journals. In Fall 2007, Hady decided to pursue his dissertation under the supervision of Prof. Stephan Olariu in the area of Wireless Sensor Networks. Hady's main research was focused on backbone construction protocols, and on developing energy efficient protocols for workforce selection and data aggregation in sensor networks. He also worked on terrain modeling using advanced 3D localization techniques and on developing adaptive sleep scheduling protocols for sensors. Hady's Ph.D. work flourished into more than 10 publications in different IEEE and ACM conferences and workshops and 3 IEEE transactions manuscripts. In Fall 2010, Hady successfully defended his thesis to join the Stream Insight team in Microsoft Corporation. Ph.D. Thesis Thesis Title: A Virtual Infrastructure for Mitigating Typical Challenges in Sensor Networks Abstract: Sensor networks have their own distinguishing characteristics that set them apart from other types of networks. Typically, the sensors are deployed in large numbers and in random fashion and the resulting sensor network is expected to self-organize in support of the mission for which it was deployed. Because of the random deployment of sensors, the resulting network is not easy to manage since the sensors do not know their location, do not know how to aggregate their sensory data and where and how to route the aggregated data. The limited energy budget available to sensors makes things much worse. To save their energy, sensors have to sleep and wake up asynchronously. However, while promoting energy awareness, these actions continually change the underlying network topology and make the basic network protocols more complex. Several techniques have been proposed in different areas of sensor networks. Most of these techniques attempt to solve one problem in isolation from the others, hence protocol designers have to face the same common challenges again and again. This, in turn, has a direct impact on the complexity of the proposed protocols and on energy consumption. Instead of using this approach we propose to construct a lightweight backbone that can help mitigate many of the typical challenges in sensor networks and allow the development of simpler network protocols. Our backbone construction protocol starts by tiling the area around each sink using identical regular hexagons. After that, the closest sensor to the center of each of these hexagons is determined -- we refer to these sensors as "backbone sensors". We define a ternary coordinate system to refer to hexagons. The resulting system provides a complete set of communication paths that can be used by any geographic routing technique to simplify data communication across the network. We show how the constructed backbone can help mitigate many of the typical challenges inherent to sensor networks. In addition to sensor localization, the network backbone provides an implicit clustering mechanism in which each hexagon represents a cluster and the backbone sensor around its center represents the cluster head. As cluster heads, backbone sensors can be used to coordinate task assignment, workforce selection, and data aggregation for different sensing tasks. They also can be used to locally synchronize and adjust the duty cycle of non-backbone sensors in their neighborhood. Finally, we propose "Backbone Switching", a technique that creates alternative backbones and periodically switches between them in order to balance energy consumption among sensors by distributing the additional load of being part of the backbone over larger number of sensors.