优化WSNs聚合质量：数据驱动的实证分析

165 浏览量更新于2024-06-19 收藏 13.41MB PDF 举报

本篇论文标题为《使用数据提高聚合质量在WSNs中的分析》，作者Alia GHADDAR在Lille1大学和黎巴嫩大学的联合指导下完成了一项重要的研究工作。Lille1大学的博士项目EDSPI（工程师科学博士）和EDST（科学与技术博士）提供了学术支持，论文的指导团队包括Pr. Isabelle SIMPLOT-RYL、Dr. Tahiry RAZAFINDRALAMBO、以及Dr. Abbass HIJAZI等，其中后者同时是黎巴嫩大学的教授。评审团成员涵盖了国际知名学者，如意大利卡拉布里亚大学的Francesca GUERRIERO教授和加拿大的Amiya NAYAK教授。研究焦点集中在如何通过有效利用数据来提升无线传感器网络（WSNs）的聚合质量。WSNs是一种分布式通信系统，它们在物联网（IoT）等领域有广泛应用，对数据的处理和传输效率至关重要。论文探讨了数据的有效整合、优化算法和可能存在的挑战，旨在改进网络的整体性能，比如减少数据包丢失、提高数据准确性，从而增强网络的可靠性。在整个研究过程中，作者Alia Ghaddar与POPS小组成员以及黎巴嫩大学科学与技术博士学院的同事们紧密合作，历时三年完成了这项开创性的工作。论文不仅体现了作者对计算机科学的深入理解，也展示了跨学科合作的力量。此外，论文还包含了致谢部分，表达了对所有参与和支持人员的感激之情，强调了集体智慧和团队精神在科研项目中的不可或缺性。这篇2011年2月12日完成的博士论文，不仅是Alia Ghaddar学术生涯的重要里程碑，也为无线传感器网络领域提供了新的理论和实践洞察，对于推动该领域的发展具有重要意义。同时，它也是对数据驱动技术在物联网环境下的实际应用进行深入探讨的一份有价值的文献。

In what follows, we consider that the sensing consumption is negligible, We tackle the

data acquisition approaches related to communication rate reduction. More speciﬁcally,

we a r e concerned with the data collection and aggregation techniques that reduce, as

possible, the number of communication as well as t he communicated overhead. In the

next subsection, we observe the impact of the communication on the energy consumption

of sensor nodes.

1.3.3 Communication

Typically, the operational states in a sensor node are: Transmit, Receive, Idle and

Sleep [174, 211]. These states for m a cycle called the duty-cycle. It is a periodic cy-

cle between two states: wake-up state (transmit, receive, idle) and sleep state. Ideally, a

node turns its radio ON only when it is going to transmit or receive data. At all other

times, it should keep its radio oﬀ (i. e sleep state). When a radio is on but idle (wa it-

ing f or p ossible traﬃc), it consumes the same amount of energy as when it is receiving

data [1 83, 214]. The communication cost of a node is composed of the cost of transmis-

sion, reception and listening [155]. The cost of receiving or transmitting a packet is an

order of magnitude greater than the cost of computation. Between the two (tr ansmission

and reception), transmitting a bit consumes twice as much energy as receiving a bit [116].

Tilak et al. show in [230] that the execution of 3000 instructions consumes the same

amount of energy as sending 1 bit over 100 meters by ra dio. Therefore, the main concern

for most WSNs algorithms [28, 146, 168, 256] is to keep the communication cost to a

minimum. In this thesis, we focus on the transmission cost. We aim at reducing the

number of transmissions between nodes (i.e the transmission rate) as well as the number

of transmitted bits (i.e the transmission overhead).

In WSNs, data are transferred (i.e routed) from source sensors to the sink through

multi-hop communication par adigm [7]. For an eﬃcient routing, diﬀerent schemes have

been adopted, such as chain [137], tree [14] or clusters [264]. We aim at reducing the

transmission cost on one-hop communication between nodes towards the base station.

Description about these schemes is presented in Chapter 2. In this work, we suppose

having a suitable routing protocols that do not interfere with the algorithms proposed in

the following chapters.

1.3.4 Data processing

Nodes in WSNs gather, process and rely information to one or more other nodes. They

collaborate to produce better outcomes during event monitoring. Ideally, data is expected

to be accurate, arrive on time, processed and transmitted with lower energy consumption.

The amount of data exchanged between nodes is expected to be as minimum as possible.

However, data should be signiﬁcant to respond to applications needs. Fo r these purposes,

collected data need to be processed and a nalyzed in a way to remove any erroneous,

suspicious, redundant or other readings that may cause unnecessary computation and

communication cost. Challenges do not stop here. They continue to grow along with the

trends of producing mobile, smart or a uto no mous objects [235, 272]. Such objects require

intelligent data processing algorithms to avoid bad decisions that aﬀect their energy con-

sumption. For example, an unreliable mobile node produces unpredictable motion and

Thèse d'Alia Ghaddar, Lille 1, 2011

http://doc.univ-lille1.fr

behaviors. It may decide to move in the wrong direction or take the longest route toward

a destination. Bad decisions may harm its source of energy. Malicious nodes may divert

traﬃc from the base station, drain energy of other nodes by producing fake information.

They may modify the collected data to cause the base station to make mis-informed deci-

sions. Such decisions are related t o errors in data collection or errors in data processing,

or shortly the quality of data collection and processing. The collection of undetected erro-

neous or incomplete data (due to data loss, misbehaviors, etc.) aﬀects the quality of data

analysis and the decisions made by sensors. Another negative impact is caused by the

presence of data redundancy. Nodes that are densely deployed (20 nodes/m

[207]), and

sense the same property are expected to provide the same information. The transmission

of redundant data may waste energy. However, if data is ﬁltered and redundancies are

detected, the communication overhead can be reduced and energy consumption can be

omitted. Fusion techniques were proposed to reduce the amount of data traﬃc, ﬁlter

noisy measurements, and make predictions about a monitored entity [163, 164]. However,

these techniques can lose much of the original structure in the data during compression,

aggregation or prediction. Therefore, they need to be convenient with the required quality

of data to prevent errors. For example, when sensor readings, from a number of sensors,

are aggregated into one value by taking the average over all these readings. This value is

used to represent each individual sensor reading. However, it deviates from the individual

sensor readings and thus introduces an error. Similarly, a loss of data may occur during

compression or predictions. In Chapter 2, we observe diﬀerent data processing techniques

for WSNs, such as, data ag gregation, compression, prediction, anomaly detection. In the

following section, we present the motivation and contribution of our work.

1.4 Motivation and contributi on

WSNs are designed to sense, gather and transmit useful information to interested users

or applications. In some applications, sensor networks are expected to interact with the

physical world. They may resp ond to the sensed events by performing corresponding

actions and assist people in their life, such as in Wireless Sensors and Actuator Net-

works (WSANs). For example, in a ﬁre handling system, the actuators can turn on the

water sprinklers upon receipt of a ﬁre report [5]. Sensors can monitor and manipulate

the temperature and lighting in a smart oﬃce or the speed and direction of a mobile

robot [257]. To produce useful outcomes in a speciﬁc WSNs application, many sensors

collaborate in monitoring and gathering information. In patient monitoring system, sen-

sors can signal a possible health trouble, by gathering information from bloo d pressure,

glucose and other monitoring body sensors. An accurate data analysis and processing is

an important issue to increase the smartness and lifetime of nodes. However, data man-

agement (i.e collection, processing and communication) is not a trivial task, especially

for energy-limited sensor networks. The energy of a node is consumed in three aspects:

sensing, processing and communication. It has been mentioned in [192] that the amount

of energy consumed by the transceiver var ies a bout 15% to about 35% of the total energy.

Nonetheless, the maj ority of energy is consumed in radio communication rather than in

阿利亚∙加达尔的博士论文，里尔第一大学，2011年

Thèse d'Alia Ghaddar, Lille 1, 2011

http://doc.univ-lille1.fr

计算。例如，在Sensoria传感器5和Berkeley6

motes中，通信和计算能量消耗之间的比例范围从103到104

[271]。此外，发送100米的1位与执行3000条指令的能量消耗相同[230]。至于传输和接收成

本，发送1位的能量消耗是接收1位的两倍[116]。因此，我们的重点是减少通信成本，而不是

计算成本，正如许多方法所做的一样[74，218，271]。更具体地说，我们致力于减少传输成本

。由于每个节点都是可以配备多个传感器的数据源，因此可以收集、处理和传输大量数据，例

如湿度、温度、位置值等。数据的传输主要通过多跳网络架构来实现，因为这种架构具有能源

效率[175]。在这种多跳传输过程中，通信开销可能会增加。通过网络内处理可以聚合来自其

他传感器的附加相关信息[118]。聚合的数据可能是大的、错误的或冗余的，这可能会损害发

送方的能源来源。因此，在聚合过程中观察数据的质量可以省略不必要的通信。我们的目标是

减少聚合过程中的能量消耗。我们试图降低节点之间的传输速率和开销。我们关注的是在传输

成本减少和网络中收集到的数据质量之间取得平衡。我们的目标是进行较少但有用的数据传输

。如果数据错误或异常，并且根据应用标准未被检测到，它们肯定会影响传感器的行为。它们

将降低网络性能，因为它会影响所产生信息/决策的质量，以及能量消耗。在这项工作中，我

们不研究节点的计算成本。相反，我们的目标是在能量有限的传感器网络中减少传输成本。在

本论文中，我们提出以下内容：

•第3章描述了一种预测技术。

•第4章描述了一种数据相关性和相似性检测技术。

•第5章描述了一种异常检测技术，以提高感知数据的质量。

1.5论文的组织结构

本论文分为五章。

•

第2章介绍了文献中提出的不同方法，用于在无线传感器网络（WSN）中节省通信能量。在

本章中，我们观察到数据驱动的技术，如网络内处理、数据压缩、数据预测等。从通信的角

度来看，这些技术被认为是能源高效的[101]。我们的算法依赖于数据聚合和估计来降低传

输速率和开销。

5Sensoria公司：http://www.sensoria.com6

http://www-bsac.eecs.berkeley.edu/

阿利亚∙加达尔的博士论文，里尔第一大学，2011年

Thèse d'Alia Ghaddar, Lille 1, 2011

http://doc.univ-lille1.fr

剩余113页未读，继续阅读

cpongm

粉丝: 5
资源: 2万+

优化WSNs聚合质量：数据驱动的实证分析

网络游戏-基于智能代理的无线传感器网络数据聚合路径规划方法.zip

物联网-智慧传输-基于典型相关性分析的无线传感网数据融合.pdf

基于无线传感器网络的实时数据采集系统

路由层无线分析脚本文件

电信设备-基于移动节点的树形无线传感网数据收集方法.zip

无线传感器网络中分布式信息融合研究进展.pdf

无线传感器网络中基于移动代理的合作数据处理算法

无线传感器网络：安全模式码数据融合协议

无线传感器网络向移动云可靠传输更多有用数据的研究进展

无线传感器网络在关键应用中的性能保证研究

最新资源