Please use this identifier to cite or link to this item:
Title: Delay-constrained and energy-efficient distributed algorithms for computation-intensive applications in WSNS
Authors: Lin, Wanyu
Degree: M.Phil.
Issue Date: 2015
Abstract: In recent years, research on wireless sensor networks(WSNs) has gradually spread from the traditional applications such as environmental monitoring to more domain-specific applications which are computation intensive due to large amount of collected physical data and complexity of the computation, such as structural health monitoring (SHM), volcano tomography, smart grid. However, for these applications, due to severe limitations of energy and bandwidth, it is necessary to utilize the computation capability of sensors and allow them to process raw signals within the network, and only transmit processed information. This implies a new revolution of designing energy-efficient distributed computing architecture for computation-intensive applications in WSNs. The distributed computing systems are scalable in the sense that all the computational and networking capacities scattered across the network could be utilized in a cooperative and distributed (not master-to-slave) manner. Considering the data-intensive and computation-intensive properties for some domain-specific applications, several unique challenging issues arise. Firstly, those algorithms designed by domain experts usually only consider the design aspects from domains such as accuracy, and they are usually sophisticated signal processing algorithms. Most algorithms are associated with complex computations such as large matrix inversion, matrix multiplication in which matrices are constructed by the raw data from different sensors. Therefore, designing a distributed version to perform matrix operations when considering the severe constraint of wireless network resources (bandwidth, energy, computing capability, memory, etc) is difficult.
In this research, we propose a framework focusing on how to implement sophisticated processing of intensive physical information within a network. We focus on the design of distributed estimation algorithms for least squares estimation. Recent years, researchers have proposed a wide range of strategies for distributed least squares estimation. However, each strategy has its own design objectives and applications scenarios. No guided schemes exists for current practical usage, making it difficult to evaluate their relative effectiveness and performance. Thus, we propose a 3 dimensional framework which provides a basis for designing, analyzing and evaluating strategies to address parameters estimation issues using least squares estimation algorithms in wireless sensor networks. In the 3D framework, we propose three design aspects of designing distributed least squares estimation, and then we study the existing works from the design aspects and then discuss their advantages and disadvantages, respectively. Finally, based on our proposed framework, we wish to conserve energy by minimizing communication with our new design, constraints on communication delays will also need to be satisfied. Thus, we propose E3, a new distributed algorithm specifically designed to guarantee the precision of least squares estimation in sensor networks, with the objective of minimizing the energy consumption incurred during communication, while observing constraints on application-specific communication delays. To evaluate the performance of our proposed framework and algorithms, we conduct simulations and structural damage detection experiments in a real environment to do test. Compared to previous works, we show that E3 maintains the same level of estimation precision while incurring much lower energy costs. Finally, we address that our 3D framework is the first work which can facilitate the design, classification and evaluation of the current distributed least squares estimation strategies in sensor networks.
Subjects: Wireless sensor networks.
Estimation theory.
Hong Kong Polytechnic University -- Dissertations
Pages: xiv, 63 pages : illustrations
Appears in Collections:Thesis

Show full item record

Page views

Last Week
Last month
Citations as of Jun 4, 2023

Google ScholarTM


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.