Date of Award
Doctor of Philosophy (PhD)
Hallstrom, Jason O
Srimani , Pradip K
Malloy , Brian A
Jacobs , David P
Wireless sensor networks are composed of large numbers of resource-lean sensors that collect low-level inputs from the physical world. The applications present challenges for programmers. On the one hand, lightweight algorithms are required given the limited capacity of the constituent devices. On the other, the algorithms must be scalable to accommodate large networks. In this thesis, we focus on the design and implementation of fast and lean (yet scalable) algorithms for classification, simulation, and target tracking in the context of wireless sensor networks. We briefly consider each of these challenges in turn.
The first challenge is to achieve high precision classification of high-level events in-network using limited computational and energy resources. We present in-network implementations of a Bayesian classifier and a condensed kd-tree classifier for identifying events of interest on resource-lean embedded sensors. The first approach uses preprocessed sensor readings to derive a multi-dimensional Bayesian classifier used to classify sensor data in real-time. The second introduces an innovative condensed kd-tree to represent preprocessed sensor data and uses a fast nearest-neighbor search to determine the likelihood of class membership for incoming samples. Both classifiers consume limited resources and provide high precision classification. To evaluate each approach, two case studies are considered, in the contexts of human movement and vehicle navigation, respectively. The classification accuracy is above 85% for both classifiers across the two case studies.
The second challenge is to achieve high performance parallel simulation of sensor network hardware. This is achieved by reducing the synchronization overhead among distributed simulation processes. Traditional parallel simulation strategies introduce significant synchronization overhead, reducing the simulation speed. We present an optimistic simulation algorithm with support for backtracking and re-execution. The algorithm reduces the number of synchronization cycles to the number of transmissions in the network under test. Concretely, we implement SnapSim, an extension to the popular Avrora simulator, based on this algorithm. The experimental results show that our prototype system improves the performance of Avrora by 2 to 10 times for typical network-centric sensor network applications, and up to three orders of magnitude for applications that use the radio infrequently.
The third challenge is to efficiently track a moving target in a network. The difficulty again lies in the conflict between the limited resource capacity of typical sensors and the significant processing requirements of typical tracking algorithms. We introduce an in-network object tracking framework for tracking mobile objects using resource-lean sensors. The framework is based on a distributed, dynamically scoped tracking algorithm which adaptively scopes the event detection region based on object speed. A leader node records the samples across an event region (without the aid of time synchronization) and estimates the object's location in situ. To minimize the number of radio transmissions, the location snapshotting rate is also adjusted based on the object speed.
In this dissertation, focusing on the above challenges, we present the design, implementation, and evaluation of classification, simulation, and tracking contributions.
Jiang, Hao, "Fast and Efficient Classification, Tracking, and Simulation in Wireless Sensor Networks" (2012). All Dissertations. 993.