clustering algorithms for mobile ad hoc networking (slides for my opening defense)

65
TL6)<B=>@? 980257 The Research of Clustering Algorithms for Highly Mobile Ad Hoc Networks 指导教师:钟章队 报告人:倪旻明 2009 年 12 月 轨道交通控制与安全国家重点实验室 博士研究生开题报告

Upload: niminming

Post on 05-Jul-2015

2.165 views

Category:

Technology


0 download

TRANSCRIPT

Page 1: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

TL6)<B=>@?980257The Research of Clustering Algorithms for Highly Mobile Ad Hoc Networks

指导教师:钟章队报告人:倪旻明2009 年 12 月

轨道交通控制与安全国家重点实验室 博士研究生开题报告

Page 2: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

/E

2

选题背景与意义

研究现状与不足

拟研究的问题与方案

已完成工作和安排

Page 3: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

1选题背景与意义

3

Page 4: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

4

IBM: Smarter Planet

“The world isn’t just getting smaller and !atter, it is actually becoming smarter. Today, almost anything—any physical object, process or system—can be instrumented, inter- connected and infused with intelligence.”

- IBM, ”Let's build a smarter planet”

Page 5: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

Banking

Cloud Computing

Food

Infrastructure

Products

Retail

Tra!c

Buildings

Education

Government

Intelligence

Public Safety

Stimulus

Water

Cities

Energy

Healthcare

Oil

Rail

Telecom

Work

Smarter ...

5

Page 6: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

6

无处不在的智能对象,被无处不达的网络与人连接在一起,再被无所不能的超级计算机调度和控制。

Page 7: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

7

Wireless Sensor Network

Vehicle Comm. Network

Wireless Mesh Network

Special-Purpose Network

Ubiquitous Networks Internet Telephone

Network

Electricity Network

Television Network

Self-organized !

Self-organized !

Self-organized !

Self-organized !

Page 8: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

8

自组织网络是现有网络体系实现泛在计算及相关业务的重要补充手段。

而移动无线自组织网络 (Mobile Ad Hoc Network, MANET) 是自组织网络技术的集中体现。

Page 9: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

移动性灵活性

抗毁性强无线通信

无基础设施

9

Page 10: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

10

平面网络结构

层级化网络结构

网络组织简单对移动性、安全性、网络资源复用性支持不好涟漪效应只适合小规模组网

以分簇算法为基础进行组网工作网络可扩展性好降低网络路由复杂度,减少维护开销降低簇内节点功能与资源方面的需求,降低维护成本对移动性、安全性、网络资源复用性提供了良好支持

Page 11: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

2研究现状与不足

11

Page 12: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

2.1 ���

12

19701975

19801985

19901995

PRNet,DARPA, 1973

SURAN,DARPA, 1983

GloMo,DARPA, 1994

TI,DARPA, 1997

IEEE 802.11 WorkGroup “Ad Hoc”, 1991

MANET WorkGroupIETF, 1997.6

Page 13: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

2.1 ���

主要研究机构:

• UCLA, M. Gerla,“无线自适应移动性实验室”(WAM)

• Cornell, Z. Hass,“无线网络实验室" (CWNL)

• UIUC, N. Vaidaya,“Ad Hoc网络研究小组”(WNG)

• Maryland, S. Tripathi,“移动计算与多媒体实验室”(MCML)

• UCSB, E. Belding-Royer,“移动性管理与组网实验室”(MOMENT)

• UCSC, J. Garcia,“计算机通信研究小组”(CCRG)

13

Page 14: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

2.1 ���D.Baker and A. Ephremides, “The architectural organization of a mobile radio network via a

distributed algorithm,” IEEE Trans. on Communications, vol. 29, no. 11, pp. 1694-1701, 1981.

14

通过借鉴传统计算机网络中“子网划分”的思想,将自组织网络中地理上相近的多个节点按照一定的规则构成不同的虚拟组(称为簇), 再在这些簇的基础上组网构成网络。

Page 15: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

M. Gerla and J.T.C. Tsai, “Multicluster, mobile, multimedia radio network,” ACM Journal on Wireless Networks, vol. 1, no. 3, pp. 255-265, 1995.

2.1 ���

IEEE Xplore: “Ad Hoc” & “Clustering”(in Journal & Top Conferences)

0

8

15

23

30

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

15

Page 16: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

2.1 ���分簇算法的研究目标

减少控制信息的开销减少重新成簇的开销降低通信复杂度降低能量开销

提供移动性方面的支持提供安全性方面的支持

16

Page 17: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

2.2 �*80�;

J. Yu and P. Chong, “A Survey of Clustering Schemes for Mobile Ad Hoc Networks,” IEEE Communications Surveys & Tutorials, vol. 7, no. 1, pp. 32-48, 2004.

a. 基于 支配集 (Dominating Set)

b. 基于 移动性感知 (Mobility-aware)

c. 基于 能量高效 (Energy E!cient)

d. 基于 负载均衡 (Load Balancing)

e. 基于 组合权值 (Combined-metrics-based)

17

Page 18: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

a. ��%MR2980 “在一个无加权图 (Unweighted Graph) G 中,如果存在一个节点集合 S,使得 G 中的任意节点都属于集合 S 或是与 S 中至少一个节点相邻,则称 S 是一个支配集 (Dominating Set)。如果这个 DS 内所有节点都可以相互联通,则称这个集合为联通支配集 (Connected Dominating Set)。”

18

Page 19: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

b. ��6� 4 2980

19

Page 20: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

b. ��6� 4 2980

19

Page 21: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

b. ��6� 4 2980

20

Page 22: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

b. ��6� 4 2980

20

Page 23: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

c. ��AOT&2980

F

E

Th 获取剩余能量信息

低于Th?N

等待 T

21

Page 24: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

c. ��AOT&2980

F

E

Th

获取剩余能量信息

低于Th?N

等待 T

进行簇头重选

Y

22

Page 25: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

d. ��HJ�C2980

No. of Cluster Members∈ [Opt. Lower Bound, Opt. Upper Bound]

23

!

Page 26: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

e. ��=�,�2980

!Parameter i x Weighting Factor i

i W1 W2 W3 W4

场景1场景2场景3场景4

24

Page 27: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

25

Dominating Set - Based

Mobility - Aware

Energy Efficient

Load Balancing

Combined-metrics-based

Newly Proposed Algorithms

Page 28: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

2.3 1*80��2�I

26

分簇信令开销对网络性能的影响

忽视 时变无线信道对分簇过程的影响

参量的灵活性对分簇算法的影响

Page 29: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

2.3 1*80��2�I

27

造成算法对场景改变的适应性较差

Scenario 1

Scenario 2

✓!

Threshold 1

A

Weighting Factor 1

BOpt. Upper Bound

COpt. Lower Bound D

Weighting Factor 2 E

Page 30: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

600 1800 3000

3

6

9

18

��������

死亡节点数目百分比 (%

)

1200 2400 3600

12

15

28

S. Bouk, and I. Sasase, “Energy E!cient and Stable Weight Based Clustering for mobile ad hoc networks,” in Proc. Signal Processing and Communication Systems 2008, pp. 1-10.

最大速度 5m/s 场景下,论文中的 WCA 性能曲线

最大速度 5m/s 场景下,论文中的 EECA 性能曲线

最大速度 15m/s 场景下,采用论文中的权重系数组合,仿真得到的 EECA 性能曲线

最大速度 15m/s 场景下,采用修正的权重系数组合,仿真得到的 EECA 性能曲线

Page 31: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

成功建立的通信链路无法建立的通信链路簇头的理想覆盖范围

可能造成异常不稳定的网络拓扑结构

29

2.3 1*80��2�I

Page 32: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

成功建立的通信链路无法建立的通信链路簇头的实际覆盖范围

可能造成异常不稳定的网络拓扑结构

30

2.3 1*80��2�I

Page 33: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

M. Chatterjee, S. K. Das, and D. Turgut, “An On-Demand Weighted Clustering Algorithm (WCA) for Ad hoc Networks,” in Proc. IEEE Globecom 2000, pp. 1697–701.

论文中的 WCA 性能曲线

论文中的 DWCA 性能曲线

加入中尺度衰落后的 WCA

加入中尺度衰落后的 DWCA

32

Page 34: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

可能造成大量的网络带宽消耗

33

2.3 1*80��2�I

平面结构 分簇结构( 理论上 )

需要传输的控制信息

for Mobility

for Re-Transmission

for Speci"c Functions

for Others

分簇结构( 实际中 )

for Basic Operations

Page 35: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

34

“A lack of realism regarding of the scenario in which MANET will be applied coupled with a lack of realism during the design of MANET are the main causes of MANET running a high risk of failure.”

M. Conti and S. Giordano, “Multihop Ad Hoc Networking: The Theory”,

Communications Magazine, IEEE (2007) vol. 45 (4) pp. 78 - 86

Page 36: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

3拟研究的问题与方案

35

Page 37: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

3.1 #572PS

36

对变化场景更具适应性的分簇算法

考虑无线信道影响的新型分簇算法

消耗更少网络带宽的新型分簇算法

Page 38: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

3.2 #N�257(.

37

WiW!j

.! j

对变化场景更具适应性的分簇算法

以基于复合权重的分簇算法模型为基础进行改进

对相应参量的获取方法进行改进

Page 39: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

38

WiW U!1. !3.!2. !4.U U U...

Page 40: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

39

WiW U!1. !3.!2. !4.U U U...v

无需额外的控制信息开销计算复杂度低对高速、低速环境适应性好

目标:设计一种独立的速度信息获取算法

源自:基于移动性感知的分簇算法

速度相关信息

v

现状:基于外部系统 (如, GPS) 的速度信息获取方法

Page 41: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

40

Doppler Shift - based Relative Speed Estimation Algorithm

A!"

D

E

!1

C

B

!2

v

fA

fdB

fdC

m

t!

PB

PC

Approaching Scenario

A!"

!1

!2

vB

D

C

E

fA

fdB

fdCt!

PB

PC

m

Receding Scenario

v =fdB · c

fA·

2 · 4

�P∆

PBK+

�P∆

PCK−

�P∆

PBK

Page 42: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

41

N � 3, 5, 7平均窗口

!1 = 15˚, 30˚, 45˚, 60˚, 75˚初始角度

n � 2.5, 3.0, 3.5衰落系数

v � 5, 10, 15, 20, 25 m/s移动速度

3 " 5 " 3 " 5 W 225! 100

22500

Page 43: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

42

Analyzation of Estimation Error in !1 = 45˚ , N = 5 Analyzation of Estimation Error in !1 = 45˚ , N = 5 Analyzation of Estimation Error in !1 = 45˚ , N = 5 Analyzation of Estimation Error in !1 = 45˚ , N = 5 Analyzation of Estimation Error in !1 = 45˚ , N = 5 Analyzation of Estimation Error in !1 = 45˚ , N = 5 Analyzation of Estimation Error in !1 = 45˚ , N = 5 Analyzation of Estimation Error in !1 = 45˚ , N = 5

v"

v (m/s) 5 10 15 20 25

v"

#en = 2.5 0.435 0.926 1.931 2.702 3.645

v"

#e n = 3.0 1.456 2.039 3.325 4.155 4.747

v"

#en = 3.5 1.333 2.298 4.747 5.890 6.429v"

$en = 2.5 2.347 4.521 5.904 7.218 8.120

v"

$e n = 3.0 2.842 5.863 8.472 9.433 10.687

v"

$en = 3.5 5.506 5.159 6.736 11.433 12.723

Page 44: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

43

WiW U!1. !3.!2. !4.U U U...v

目的:尽可能选择剩余能量多的节点担任簇头

源自:基于能量高效的分簇算法

能量相关信息

E

簇头节点 70%30%成员节点

全网能量消耗

Page 45: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

44

Fixed Data Generation Model

Dynamic Data Generation Model

Page 46: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

45

WiW U!1. !3.!2. !4.U U U...v

目标:设计一种簇成员数目最优区间自适应计算方法

源自:基于负载均衡的分簇算法

拓扑相关信息

E #

拟采用的理论方法:

拓扑控制网络容量图论 [a, b]

Page 47: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

46

WiW U!1. !3.!2. !4.U U U...v

目的:选择拥有更多“近邻”的节点担任簇头

位置相关信息

E #

!

" "" """ "# #

!"

""""""#

#

d

Page 48: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

47

WiW U!1. !3.!2. !4.U U U...v

目的:更加灵活的调整各参量的权重系数,以适应不同场景

权重系数序列

E # d

目标:设计基于灰色理论的权重系数自适应计算方法

Scenario 1

Scenario 3

Scenario 2

"1"2"3

"4

"1 "3"2 "4

"1"2"3"4

Gray Theory based Algorithm

Page 49: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

3.2 #N�257(.

48

考虑无线信道影响的新型分簇算法

无线信道会给分簇工作带来恶性的不确定效果

在考虑节点的移动性之后这种影响会更加恶劣

节点间的相对位置(距离)是重要的决定因素

结合以随机几何为代表的随机数学工具进行分析

Page 50: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

49

3.2 #N�257(.考虑无线信道影响的新型分簇算法

网络覆盖网络干扰 网络吞吐量

网络联通度

一些物理层接收技术的工作模式

一种无线信道感知的高速移动自组织网络分簇算法

Stochastic Geometry

Point Process Theory

Random Graph

The Probabilistic Method

Page 51: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

3.2 #N�257(.

50

消耗更少网络带宽的新型分簇算法

方案一:高速移动自组织网络的混合分簇算法

路由算法

Hybrid Routing

Active Routing

Passive Routing

分簇算法Active Clustering (Almost All)

Passive Clustering (Proposed)

Hybrid Clustering ( ? )

Page 52: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

51

3.2 #N�257(.消耗更少网络带宽的新型分簇算法

方案二:高速移动自组织网络的 k-hop 分簇算法

1-hop 3-hop

Page 53: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

4已完成工作与安排

52

Page 54: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

4.1 ��!��V�3��

53

GUI

FunctionNodes

Algorithms

Scenarios

MatLab

Page 55: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

4.1 ��!��V�:QG

J.G. Proakis, 《数字通信》A. Goldsmith, 《无线通信》

T.S. Rappaport, 《无线通信原理与应用》S. Haykin, 《现代无线通信》

钟章队 等, 《铁路综合数字移动通信系统》

J.G. Proakis, 《现代通信系统》

B.A. Forouzan, 《数据通信与网络技术》

基础理论方面:

54

Page 56: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

4.1 ��!��V�:QG

Ad Hoc 网络相关:

<Ad-hoc Networks: Fundamental Properties and Network Topologies>R. Hekmat

A. Boukerche <Algorithms and Protocols for Wireless and Mobile Ad Hoc Networks>

M. Barbeau <Principles of Ad Hoc Networking>

S. Basagni <Mobile Ad Hoc Networking>

P. Santi <Topology Control in Wireless Ad Hoc and Sensor Networks>

L. Gavrilovska <Ad Hoc Networking Towards Seamless Communications>

55

Page 57: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

4.1 ��!��VF' D

期刊:1. M. Ni, H. Wu, B. Ai and Z. Zhong, “Composite Recon!gurable Multi-Clustering Ad Hoc Network”, ���K���", Vol. 33, No.2, 2009, p 94-97.

2. M. Ni, Z. Zhong and H. Wu, “A Novel Energy E"cient Clustering Algorithm for Dynamic Wireless Sensor Network”, to appear in Journal of Internet Technology, No.4, 2009.

国际会议:1. M. Ni, Z. Zhong, H. Wu and D. Zhao, “An Energy E"cient Clustering Scheme for Mobile Ad Hoc

Networks”, Submitted to IEEE VTC ‘2010 Spring.2. M. Ni, Z. Zhong, H. Wu and D. Zhao, “A New Stable Clustering Scheme for Highly Mobile Ad Hoc

Networks”, Accepted for IEEE WCNC 2010.3. X. Qiao, M. Ni and Z. Tan, “A Directional Antennas-Based Topology Control Algorithm for Two-

tiered Wireless Sensor Network”, IEEE WiCOM 2009.4. M. Ni, Z. Zhong and R. Xu, “An Energy E"cient Routing Scheme for Wireless Sensor Network in

Heavy Haul Railway Transportation”, International Conference of International Heavy Haul Association (IHHA) 2009.

56

Page 58: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

4.2 +-���$

书籍阅读:J.A. Gubner, <Probability and Random Process for ECE>P.V. Mieghem, <Performance Analysis of Communications Network and System>A. Baddeley, <Spatial Point Process and their Applications>

57

论文阅读:IEEE Wireless Comm., IEEE Comm., IEEE Network, IEEE Trans. on Networking,ACM MobiCom, ACM MobiHoc, IEEE InfoCom, IEEE GlobeCom, IEEE ICC 等

论文发表:��������� 8-10

Page 59: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

58

1 2 3 4 5 6 7 8 9 10 11 12

2010

Stage 1

阶段1在已进行的工作基础上完善对变化场景更具适应性的分簇算法

分解 整合"

Page 60: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

58

1 2 3 4 5 6 7 8 9 10 11 12

2010

Stage 2

阶段2完成对随机几何等相关数学工具的学习

对网络的覆盖、干扰、联通度、吞吐量进行初步分析

Page 61: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

58

1 2 3 4 5 6 7 8 9 10 11 12

2010

Stage 3

阶段3整合阶段2的工作成果

设计、实现考虑无线信道影响的新型分簇算法

进行性能仿真,对比试验

Page 62: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

59

111 12 2 3 4 5 6 7 8 9 10 11 12

2011

Stage 4

阶段4设计、实现消耗更少网络带宽的新型分簇算法

进行性能仿真,对比试验

与阶段1设计的算法进行开销对比

Page 63: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

59

111 12 2 3 4 5 6 7 8 9 10 11 12

2011

Stage 5

阶段5整合前4个阶段获得的各种零散结论

进行相应的完善工作

Page 64: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

59

111 12 2 3 4 5 6 7 8 9 10 11 12

2011

Stage 6

阶段6完成博士论文

Page 65: clustering Algorithms for Mobile Ad Hoc Networking (Slides for my opening defense)

谢谢2009.12