Advisor : Prof. Yu-Chee Tseng Student : Yi-Chen Lu 12009/06/26.

1

An Energy Efficient, Load Balanced Multicast Protocol with Probabilistic Anycast for ZigBee Wireless Sensor Networks

Advisor : Prof. Yu-Chee TsengStudent : Yi-Chen Lu

2009/06/26

2

Outline

Introduction Related Work Motivation Goal Protocol Design Simulation Conclusion

3

Introduction

A WSN is composed of numerous inexpensive wireless sensor nodes, each of which is normally powered by batteries and has limited computing ability

Wireless sensor nodes are capable of not only collecting, storing, processing environmental information, but also communicating with neighboring nodes

Many research works have been dedicated to WSNs, such as routing, self-organization, deployment, and localization

4

Introduction

Multicast is a fundamental routing service of network communication

In WSN, a single message can be delivered to multiple destinations efficiently via multicast communication

In WSN, members may dynamically join and leave the groups

Fruits Area

Drinks Area

Join banana group

Join coke group

5

Introduction

ZigBee is a cost-effective wireless networking solution that supports low data-rates, low-power consumption, security, and reliability

Most WSN industries have adopted ZigBee as their communication protocol and developed numerous products

6

ZigBee Multicast

In ZigBee, multicast members are physically separated by a hop distance of no more than MaxNonMemberRadius

ZigBee multicast exploits regional flooding to deliver the multicast message

Region bounded by MaxNonMemberRadius

Member Another Member

Drawbacks of ZigBee multicast Heavy traffic

overhead High energy cost Unreliable

7

Outline

IntroductionRelated Work

Overlay Multicast Geographic Multicast Relay-Selection Multicast

MotivationGoalProtocol DesignSimulationConclusion

8

Overlay Multicast

PAST-DM (Wireless Networks 2007) Applying unicast leads to excessive

energy consumption and redundant transmissions

AOM (ICPPW 2007) Applying broadcast eliminates

redundant transmissions Packet header overhead

Overlay multicast needs extra cost to support dynamic member actions

Fixed delivery paths lead to single-node failure problem

redundant

6 transmissions

4 transmissions

Destination List & Forwarder List

9

Geographic Multicast

GMREE (COMCOM 2007) Cost over progress ratio

Drawbacks Packet header overhead Location information must be available Suffer from the face routing cost Do not support dynamic member

joining/leaving Single-node failure problem

S SS

S

S

S

S

S

S

S

S

S

S

S

SS

S

10

Relay-Selection Multicast

Steiner-tree based multicast BIP and MIP (MONET 2002)▪ Based on Prim’s algorithm to find a minimum-cost spanning tree

NJT and TJT (COMCOM 2007 ) ▪ Minimum cost set cover heuristics

Single-node failure problem Computing complexity is high Centralized algorithm must keep global

information Do not support dynamic member joining/leaving Source tree construction overhead

11

Outline

Introduction Related Work Motivation Goal Protocol Design Simulation Conclusion

12

Motivation

Due to the limited power resource, energy efficient multicast is a critical issue in WSN

ZigBee multicast is not only energy inefficient but also unreliable

Many approaches have been proposed to study on the energy efficient multicast issues in WSN

13

Motivation

However, these proposed approaches either have significant drawbacks or are not compatible with ZigBee Single-node failure problem (all) Do not support dynamic member joining/leaving

(all) Packet header overhead (overlay & geographic) Location information (geographic) High computing complexity (geographic & relay) Must keep global information (relay-selection)

14

Outline

Introduction Related Work Motivation Goal Protocol Design Simulation Conclusion

15

Goal

Propose a multicast routing protocol which has the following features ZigBee Compatible Energy efficient▪ Less energy consumption

Reliable▪ Higher delivery ratio

Load balanced▪ Avoid single-node failure problem▪ Prolong the network lifetime

Support dynamic member joining/leaving

16

Outline

Introduction Related Work Motivation Goal Protocol Design Simulation Conclusion

Protocol Overview

17

S

S

S

S

S

S

SS

S

Probabilistic

Anycast

Random Backoff

Packet Forwardin

g

Coverage Over Cost Ratio

Residual Energy

Forwarding Strategy

Ack Mechanism

Multicast Informatio

n Table (MIT)

S

S

S

S

S

S

18

Protocol Flow

MIT Maitenance

Radom Backoff

Discard

Forward

Rebroadcast

Ack Mechanis

m

Forwarding Strategy

Initiate A Multicast

Receive A Multicast Packet

Coverage Over Cost Ratio

Residual Energy

Wait for twait

Backoff for tb

Multicasting

19

Outline

Introduction Related Work Motivation Goal Protocol Design

MIT Maintenance Multicasting

Simulation Conclusion

20

MIT Maintenance

Multicast Information Table (MIT) Reachable members within

MaxNonMemberRadius hops Hop distances to the reachable members

MIT

Member Hop Count

m1 h1

m2 h2

… …

mn hn

21

MIT Maintenance Example

1

5

8

15

21

MIT 21

15 2

MIT 1

5 2

8 2

15 3

MIT 5

1 2

8 3

MIT 8

1 2

5 3

15 3

MIT 15

1 3

8 3

21 2

MaxNonMemberRadius = 2

MIT keeps the information of only the members located within the region bounded by MaxNonMemberRadius hops

22

Outline

Introduction Related Work Motivation Goal Protocol Design

MIT Maintenance Multicasting

Simulation Conclusion

23

Multicasting

MIT Maitenance

Radom Backoff

Discard

Forward

Rebroadcast

Ack Mechanis

m

Forwarding Strategy

Initiate A Multicast

Receive A Multicast Packet

Coverage Over Cost Ratio

Residual Energy

Wait for twait

Backoff for tb

Multicasting

24

Multicasting

Our protocol adopts a probabilistic anycast mechanism based on the coverage over cost ratio and each node’s residual energy

Our protocol is similar to the relay-selection approaches

However, the selection of relay nodes is determined by the receivers, rather than by the senders

25

Probabilistic Anycast

Random Back-off

Packet Forwarding

Probabilistic Anycast

•Coverage Over Cost Ratio• Residual Energy

•Forwarding Strategy•Ack Mechanism

26

Initiating A Multicast

Packet

M H

m1 2

m2 2

m3 3

Eavg

S

•Multicast to {m1, m2, m3}•Hop distance to them is {2, 2, 3 }•The average residual energy of my

neighbors is Eavg

Destination Set M = {m1, m2, m3}Distance Set H = {2, 2, 3 }

Average residual energy of the neighbors = Eavg

MIT

m1 2

m2 2

m3 3

27

Upon Receipt of A Multicast Packet

Packet

M H

m1 2

m2 2

m3 3

Eavg(S)

S

MIT

m1 1

m2 3

m3 2

S XPacke

t

M H

m1 1

m2 3

m3 2

Eavg(X)

Remove member originator/previous hop to avoid

loop

Remove the members which are further from me

than from the previous hop to avoid detours

MIT

m1 2

m2 2

m3 3

Generate a random Backoff period

28

Random BackoffRandom Backoff

Packet Forwarding

Probabilistic Anycast

•Coverage Over Cost Ratio• Residual Energy

29

Coverage Over Cost Ratio Coverage Over Cost Ratio

The coverage over cost ratio is targeted at reaching as many member nodes as possible while consuming as little energy as possible

ontransmissibroadcast single one ofcost energy the:

MIT theofentry each in counts hop of sum the: TL

MIT in the members reachable ofnumber the: D

(1) 1DTL

D

tx

tx

E

Ef

X

A

B

Y C

B

A

Number of covered

members

Estimated energy cost Superior

in forwarding

30

Radom Backoff

The backoff timer interval tb is generated randomly within the range [0, T]

With greater f value, T should be smaller The single-node failure problem is still

unsolved

erRadiusMaxNonMembEf

E

Nf

ff

ff

TT

tx

tx

max

1

)1(

min

max

minmax

min

max

Normalize f to a parameter α to show the influence of coverage over cost ratio on the backoff interval

31

Random Backoff

We further introduce the idea of load balance to our protocol

Therefore, a node which has more energy and covers more destination members with less energy cost has a better chance to generate a shorter backoff interval

The data delivery paths are dynamically adjusted during each propagation according to the instant network condition

(2) )1( maxr

avg

E

ETT S

S

S

S

Superior in forwarding

32

Packet Forwarding

Random Back-off

Packet Forwarding

Probabilistic Anycast

•Forwarding Strategy•Ack Mechanism

33

Forwarding Strategy

During the backoff period, any member covered by other nodes is removed from M

When the backoff period expires, and M is not Φ

Rebroadcast the packet with up-to-date M, H and Eavg

34

ACK Mechanism

After sending out the multicast packet, the sender waits for a period

of time twait to confirm the forwarding

status of the destination members If not all the members in set M of the

sender are forwarded when twait

expires, the sender retransmits the multicast packet

35

Outline

Introduction Related Work Motivation Goal Protocol Design Simulation Conclusion

36

Simulation

Simulation Environment

Simulation Duration 105 sec

Area 35m * 35m

Number of Nodes 100~500 (Random deployment)

Number of Members 10 (Randomly generated)

MaxNonMemberRadius 5

Transmission Range 6m

MAC IEEE 802.15.4 MAC with unslotted CSMA-CA

Max Backoff Interval 5ms

Transmission Rate 250Kbps

37

Simulation

100 200 300 400 5000

5

10

15

20

25

30

ZigBee Proposed protocol

Network size

Late

ncy (

ms)

100 200 300 400 5000

100

200

300

400

500

600

700

800

900

1000

ZigBee Proposed protocol

Network Size

Nu

mb

er

of

packets

38

Outline

Introduction Related Work Motivation Goal Protocol Design Simulation Conclusion

39

Conclusion

Energy efficient multicast is a critical issue in WSN Many approaches have been proposed, but they fail

to achieve energy efficiency and load balance at the same time

We propose a ZigBee compatible multicast protocol Energy efficient Load balanced Reliable Support dynamic member joining/leaving

Simulation result shows that our protocol outperforms ZigBee in energy consumption and latency

40

Q&A

Thanks