19
Infocom'04 Ossama Younis, Purdue University 1 Distributed Clustering in Ad-hoc Sensor Networks: A Hybrid, Energy-Efficient Approach Ossama Younis and Sonia Fahmy Department of Computer Sciences, Purdue University

Distributed Clustering in Ad-hoc Sensor Networks: A Hybrid, Energy-Efficient Approach

  • Upload
    helene

  • View
    42

  • Download
    0

Embed Size (px)

DESCRIPTION

Distributed Clustering in Ad-hoc Sensor Networks: A Hybrid, Energy-Efficient Approach Ossama Younis and Sonia Fahmy Department of Computer Sciences, Purdue University. Contributions. A new distributed clustering protocol for sensor networks that has the following properties: Energy-efficient - PowerPoint PPT Presentation

Citation preview

Page 1: Distributed Clustering in Ad-hoc Sensor Networks: A Hybrid, Energy-Efficient Approach

Infocom'04 Ossama Younis, Purdue University 1

Distributed Clustering in Ad-hoc Sensor Networks: A Hybrid, Energy-Efficient

Approach

Ossama Younis and Sonia FahmyDepartment of Computer Sciences,

Purdue University

Page 2: Distributed Clustering in Ad-hoc Sensor Networks: A Hybrid, Energy-Efficient Approach

Infocom'04 Ossama Younis, Purdue University 2

Contributions

A new distributed clustering protocol for sensor networks that has the following properties: Energy-efficient Terminates rapidly Considers cluster quality, e.g., load-balanced

clusters or dense clusters Has low message/processing overhead

Page 3: Distributed Clustering in Ad-hoc Sensor Networks: A Hybrid, Energy-Efficient Approach

Infocom'04 Ossama Younis, Purdue University 3

Sensor Networks Application-specific, e.g.,

Monitoring seismic activities Surveying military fields Reporting radiation levels at

nuclear plants Nodes are usually:

Densely deployed Limited in processing, memory,

and communication capabilities Constrained in battery lifetime Left unattended

Page 4: Distributed Clustering in Ad-hoc Sensor Networks: A Hybrid, Energy-Efficient Approach

Infocom'04 Ossama Younis, Purdue University 4

Goals Scalability, data and state aggregation,

robustness, and prolonged network lifetime

Time until the first node dies

Time until the last node dies

How to prolong the network lifetime? Deploy redundant nodes Apply energy-efficient protocols, e.g.,

MAC layer protocols can reduce energy waste Topology management can distribute energy

consumption

What is network lifetime?

Page 5: Distributed Clustering in Ad-hoc Sensor Networks: A Hybrid, Energy-Efficient Approach

Infocom'04 Ossama Younis, Purdue University 5

Topology management

Cell-based approach Cluster-based approachobserver

Page 6: Distributed Clustering in Ad-hoc Sensor Networks: A Hybrid, Energy-Efficient Approach

Infocom'04 Ossama Younis, Purdue University 6

Outline

System model and requirements The Hybrid, Energy-Efficient, Distributed

clustering protocol (HEED) HEED properties Evaluation Related Work Conclusion

Page 7: Distributed Clustering in Ad-hoc Sensor Networks: A Hybrid, Energy-Efficient Approach

Infocom'04 Ossama Younis, Purdue University 7

System Model

A set of n sensor nodes are dispersed uniformly and independently in a field

Sensor nodes are Quasi-stationary Unattended Equally significant Location un-aware Homogeneous (similar capabilities) Serving multiple observers Possess a fixed number of transmission power levels

Page 8: Distributed Clustering in Ad-hoc Sensor Networks: A Hybrid, Energy-Efficient Approach

Infocom'04 Ossama Younis, Purdue University 8

Requirements

Our goal is to design a new clustering approach that has the following properties: Completely distributed Terminates in O(1) iterations Has low message/processing overhead Generates high energy, well-distributed cluster

heads Can provide other characteristics, such as

balanced or dense clusters

Page 9: Distributed Clustering in Ad-hoc Sensor Networks: A Hybrid, Energy-Efficient Approach

Infocom'04 Ossama Younis, Purdue University 9

Approach (HEED)

We propose the Hybrid, Energy-Efficient, Distributed clustering approach (HEED)

Heed is hybrid: Clustering is based on two parameters

HEED is distributed: Every node only uses information from its 1-hop

neighbors (within cluster range) HEED is energy-efficient:

Elects cluster heads that are rich in residual energy Re-clustering results in distributing energy

consumption

Page 10: Distributed Clustering in Ad-hoc Sensor Networks: A Hybrid, Energy-Efficient Approach

Infocom'04 Ossama Younis, Purdue University 10

HEED - Parameters Parameters for electing cluster heads

Primary parameter: residual energy (Er) Secondary parameter: communication cost (used

to break ties)i.e., maximize energy and minimize cost

Page 11: Distributed Clustering in Ad-hoc Sensor Networks: A Hybrid, Energy-Efficient Approach

Infocom'04 Ossama Younis, Purdue University 11

HEED – Algorithm at node v Initialization

Main processing

Finalization

Discover neighbors within cluster range Compute the initial cluster head probability

CHprob = f(Er/Emax)

If v received some cluster head messages, choose one head with min cost

If v does not have a cluster head, elect to become a cluster head with CHprob .

CHprob = min(CHprob * 2, 1) Repeat until CHprob reaches 1

If cluster head is found, join its cluster Otherwise, elect to be cluster head

Page 12: Distributed Clustering in Ad-hoc Sensor Networks: A Hybrid, Energy-Efficient Approach

Infocom'04 Ossama Younis, Purdue University 12

HEED - Example

Compute CHprob

and costElect to become

cluster head

Resolve ties

Select your cluster head

(0.2,2)

(0.4,3)

(0.2,3)

(0.1,2)

(0.1,4)(0.6,2)

(0.2,5) (0.5,3)

(0.8,4) (0.2,3)

(0.6,4)

(0.5,4)

(0.1,4)(0.9,4)

(0.3,2)

(0.7,5)(0.3,2)

(0.2,3)

a1

c4a3

a2

a5a6

c3a12

a11a13

a9

a7 a8

a4

a10c2

c1a14

Discover neighbors

Page 13: Distributed Clustering in Ad-hoc Sensor Networks: A Hybrid, Energy-Efficient Approach

Infocom'04 Ossama Younis, Purdue University 13

HEED - AnalysisHEED has the following properties:

Completely distributed Clustering terminates in O(1) iterations:

Message overhead: O(1) per node Processing overhead: O(n) per node Cluster heads are well distributed. Pr{two CHs

are within the same cluster range}: (p = initial CHprob )

1p1logNmin

2iter

1)1(log

0

221p

i

inbr pp

Page 14: Distributed Clustering in Ad-hoc Sensor Networks: A Hybrid, Energy-Efficient Approach

Infocom'04 Ossama Younis, Purdue University 14

HEED – Inter-cluster communication

Lemma 1 (Blough and Santi’02): Assume n nodes are dispersed uniformly and

independently in an area R=[0,L]2. If Rc2n=aL2lnL, for

some a>0, Rc << L, and n>>1, then limn,N→∞E(number of empty cells) = 0, where a cell is an

area22cc RR

Lemma 2: There exists at least one cluster

head a.a.s. in any area of size cc RR )

212()

212(

Page 15: Distributed Clustering in Ad-hoc Sensor Networks: A Hybrid, Energy-Efficient Approach

Infocom'04 Ossama Younis, Purdue University 15

HEED – Inter-cluster communication

Theorem 1: Two cluster heads in two

neighboring areas can communicate if

Theorem 2: HEED produces a connected

multi-hop cluster head graph (structure) asymptotically almost surely

ct RR 6

2.7Rc

2.7Rc

Rt

CH1

CH2

Page 16: Distributed Clustering in Ad-hoc Sensor Networks: A Hybrid, Energy-Efficient Approach

Infocom'04 Ossama Younis, Purdue University 16

Performance evaluation

2000x2000 network field with 1000 nodes

Demonstrating HEED properties: fast termination, rich-energy cluster heads, and cluster quality

Page 17: Distributed Clustering in Ad-hoc Sensor Networks: A Hybrid, Energy-Efficient Approach

Infocom'04 Ossama Younis, Purdue University 17

Performance evaluation (cont’d)

Apply HEED to an application where nodes directly contact a far observer

Compare to multi-hop LEACH clustering 100x100 network Initial Er = 2 Joule 1 round = 5 TDMA

frames

Page 18: Distributed Clustering in Ad-hoc Sensor Networks: A Hybrid, Energy-Efficient Approach

Infocom'04 Ossama Younis, Purdue University 18

Related Work Topology management protocols suffered from at

least one of the following problems: Dependence on location awareness (e.g., GAF) Slow convergence (i.e., dependent on the network

diameter) (e.g., DCA) Energy efficiency was not the main goal of many

protocols, e.g., Max-Min D-clustering No focus on clustering quality, such as having

cluster heads well-distributed in the network (e.g. LEACH)

Page 19: Distributed Clustering in Ad-hoc Sensor Networks: A Hybrid, Energy-Efficient Approach

Infocom'04 Ossama Younis, Purdue University 19

Conclusion We have proposed HEED clustering HEED is fast and has low overhead HEED can provide other features, such as load-

balancing HEED is independent of:

Homogeneity of node dispersion in the field Network density or diameter Distribution of energy consumption among nodes Proximity of querying observers

HEED can be extended to provide multi-level hierarchy