UVM Theses and Dissertations
Format:
Print
Author:
Chen, Lei
Dept./Program:
Electrical Engineering
Year:
2012
Degree:
Ph. D.
Abstract:
Wireless sensor networks have been used to solve many real world problems. In many applications, we are interested in the statistics of the sensed data instead of their individual values. Therefore, we need an effective way to aggregate data from the network. Generally, there are two approaches to accomplish this: centralized processing and distributed consensus. The centralized way relies on a base station to collect and aggregate data from the nodes in the network. In distributed consensus, the aggregation is performed through local communication, where each individual node updates its consensus state based on information exchanged with others. Under certain conditions, the network will converge to a status with all nodes having the same state value, which is often referred to as consensus status. Compared to centralized approaches, distributed schemes are often more reliable and scalable.
While individual sensors could have unreliable information or be vulnerable to environmental or network dynamics, all sensors iIi the network, via effective cooperation, can produce consensus decisions that are higher in reliability. Prior work on distributed consensus focuses on the theoretical analysis of the asymptotic behavior of consensus algorithms. However, practical issues including energy efficiency and robustness to link failures, which are crucial to the efficacy of a sensor network, have not been sufficiently explored. In practice, communication between sensors constitutes a significant part of energy consumption in wireless sensor network. Therefore, algorithms should be designed to limit communication cost while being resilient to link failures.
In this work, we present several approaches to solve such practical problems. Specifically, the contributions of this work are as follows.
First, a self-aware criterion is developed to detect consensus based on the local information of each node. This approach does not rely on the states of other nodes and leads to substantial saving of communication resources. A systematic analysis and derivation of the optimal parameters that minimize the upper bound of the number of iterations needed to reach consensus is presented.
In addition, a scheme for distributed consensus that can achieve a good balance between convergence rate and energy efficiency is presented. This scheme actively selects a subset of links with significant contribution to the formation of consensus at each iteration, thus adapting the network topology dynamically to the changes of the sensor states. An optimization problem for optimal link selection is formulated and an efficient algorithm to solve it by quadratic programming relaxation and random sampling is derived. Finally, a series of empirical simulations under different configurations and a hardware implementation to locate centroid of WSN deployment are presented that demonstrate the improved energy efficiency and robustness to link failures.
While individual sensors could have unreliable information or be vulnerable to environmental or network dynamics, all sensors iIi the network, via effective cooperation, can produce consensus decisions that are higher in reliability. Prior work on distributed consensus focuses on the theoretical analysis of the asymptotic behavior of consensus algorithms. However, practical issues including energy efficiency and robustness to link failures, which are crucial to the efficacy of a sensor network, have not been sufficiently explored. In practice, communication between sensors constitutes a significant part of energy consumption in wireless sensor network. Therefore, algorithms should be designed to limit communication cost while being resilient to link failures.
In this work, we present several approaches to solve such practical problems. Specifically, the contributions of this work are as follows.
First, a self-aware criterion is developed to detect consensus based on the local information of each node. This approach does not rely on the states of other nodes and leads to substantial saving of communication resources. A systematic analysis and derivation of the optimal parameters that minimize the upper bound of the number of iterations needed to reach consensus is presented.
In addition, a scheme for distributed consensus that can achieve a good balance between convergence rate and energy efficiency is presented. This scheme actively selects a subset of links with significant contribution to the formation of consensus at each iteration, thus adapting the network topology dynamically to the changes of the sensor states. An optimization problem for optimal link selection is formulated and an efficient algorithm to solve it by quadratic programming relaxation and random sampling is derived. Finally, a series of empirical simulations under different configurations and a hardware implementation to locate centroid of WSN deployment are presented that demonstrate the improved energy efficiency and robustness to link failures.