W. Chen*, A.K.M.M. Islam**, M. Malkani*, A. Shirkhodaie*, K. Wada**, M. Zein-Sabatto*

* Tennessee State University, USA** Nagoya Institute of Technology, Japan

Novel Broadcast/Multicast Protocols for Dynamic Sensor Networks

Novel Broadcast/Multicast Protocols for Dynamic Sensor Networks

Lecture 3-2: Networking Architecture, Routing Protocols and Algorithms

International Parallel and Distributed Processing Symposium, 2007

Basics in Wireless Sensor Networks (WSNs) Review of Broadcast/Multicast Protocols Proposed Broadcast/Multicast Protocols on

Dynamic WSNs Network Reconfiguration Latency/Energy Analysis and Simulation Summary and Future Work

CONTNETSCONTNETS

• A WSN consists a large number of sensor nodes which are deployed in a large area.

• Inexpensive sensor nodes have limited capabilities in power supply, communication and processing, and they experience failures frequently.

• Communication is in ad-hoc manner.

• Network is dense: overlap and collision in control and communication

• Communication patterns are specific: one-to-many (broadcast/multicast), many-to-one (data gathering)

• Network topology is dynamic: topology self-reconfiguration is necessary

• Tasks need to be self-organized without global centralized control for scalability.

Autonomous tracking in Sensor Network

Basics in A Wireless Sensor NetworkBasics in A Wireless Sensor Network

Problem Statement Problem Statement

Develop time and energy efficient protocols and algorithms for broadcast, Multicast, and Data-gathering on a dynamic sensor network.

Objectives: Maximize the network throughput in Broadcast/multicast

and data gathering Minimize the energy consumption for extending network

life time Enable network self-reconfiguration for dynamic topology

changes

Model of A Flat (unstructured) Sensor NetworkModel of A Flat (unstructured) Sensor Network

All nodes use a single radio channel

Each node has a distinct ID number and it has no any network knowledge

Each node repeats transmission or reception and performs its local computation in a fixed interval, called round. In each round, a node acts as either a transmitter or a receiver.

Nodes have no collision detection

Review of Broadcast/Multicast ProtocolsReview of Broadcast/Multicast Protocols

On Unstructured Networks Without network knowledge: With whole network knowledge:

rounds (n)

network theofdiameter theis

rounds, )log(),(log 52

L

nLOn

A sensor network

On Hierarchical Networks O(n) to for clustering

and construction of cluster-based network

rounds when there are p clusters using depth-first order on backbone

(J. Uchida and others, 2006)

Extra cost for structure maintenance

24 p

rounds )( 2nO

r

memberheadgateway

Cluster-based network

r

member

gateway

There are Trade-offs between communication and network reconfiguration

Results of This paperResults of This paper

Time and energy efficient broadcast and multicast protocols based on collision-free flooding on a cluster-based sensor network

Self-reconfiguration algorithm for a cluster-based network

Theoretical analysis and simulation

Generalization of the proposed protocols to data gathering.

Broadcast Using Flooding: Good or Bad? Broadcast Using Flooding: Good or Bad?

memberhead

gateway

Cluster-based network

member

gateway

sourceBroadcast Based on Depth-First Order

(J. Uchida and others, 2006)

• Broadcast completes in 4p-2 rounds

• Each node is awake until its children finish broadcast.

• Broadcast stops if node/link failure happens.

A sensor network

Broadcast using flooding causes broadcast storm problem !

Broadcast Based on Collision-Free Flooding

• Using TDM to avoid collision

• Each node needs to be awake in a short time

• Broadcast is more robust: even node/link failure happens, broadcast will still continue.

2 1 2

v

x ｙ ｚ2 1 2

v

x ｚx ｚ

TDM: Every node v can receive the broadcast message from at least one node.

Broadcast Using Flooding: Good or Bad? - Contine

Broadcast Using Flooding: Good or Bad? - Contine

Broadcast Algorithm Using Collision-Free FloodingBroadcast Algorithm Using Collision-Free Flooding

CNet(G): Spanning tree of G consisting of clusters and the backbone, where

m: broadcast message

v.time-slot: time-slot assigned to node v

: largest time-slot assigned in CNet(G)

Algorithm CollisionFreeFlooding(CNet(G), ,m) Root r : transmits package (m, r.time-slot, ,0) at its time-slot r.time-slot, Other node v: Assume that v’s depth is i+1. If v is a backbone node received package (m, t, , i), v wait rounds, and then transmits (m, v.time-slot, ,i+1) at its time-slot v.time-slot.If v is a leaf, it receives m and do nothing.

t

Time and Energy Consumption: By using CNet(G) and TDM time-slots, (1) A broadcast can be completed in rounds, h is the height of CNet(G),

and(2) In the broadcast, each node needs to be awake in rounds.

)1( h

2

1

1 2

211 2

21

1 23

3

Broadcast Algorithm Using Collision-Free Flooding -- Improvement

Broadcast Algorithm Using Collision-Free Flooding -- Improvement

Observation: Backbone is relative sparse –the TDM time-slots needed for the broadcast on backbone will be much smaller. Therefore, algorithm can be changed to

Step 1: Broadcast the source message on the backbone using b-time-slot .

Step 2: Heads transmit the source message to its members using l-time-slot .

Time and Energy Consumption: By using CNet(G) and TDM time-slots, (1) A broadcast can be completed in rounds, where h is the height of

CNet(G), and(2) In the broadcast, each node needs to be awake in rounds.

)1(h

2

Time and Energy Consumption with k-channels: By using CNet(G) and TDM time-slots, if k channels are available,

(1) A broadcast can be completed in G in rounds, where h is the height of CNet(G), and

(2) In the broadcast, each node needs to be awake in rounds.

kh /))1((

k/)2(

1

1 2

211 2

21

1 23

3

3

2

From Broadcast Protocol to Multicast ProtocolFrom Broadcast Protocol to Multicast Protocol

Structure of MCNet(G):

At each backbone node, there are two lists: (group-list) and [relay-list]

(1) (1) (1) (1) (1) (1) (1) (1)

(1,2)[1,2]

(1,2) (1) (2) (1) (1) (2) (2) (2) (1)

(1,2) (1) (1,2) (2)

(2)[1,2]

(1,2)[1] (1)

[2]

(1,2)[1,2]

(2)(1,2)[1]

(1,2)[1,2]

(1)[1]

(1)[1]

vu

(1) (1) (1) (1) (1) (1) (1) (1)

(1,2)[1,2]

(1,2) (1) (2) (1) (1) (2) (2) (2) (1)

(1,2) (1) (1,2) (2)

(2)[1,2]

(1,2)[1] (1)

[2]

(1,2)[1,2]

(2)(1,2)[1]

(1,2)[1,2]

(1)[1]

(1)[1]

vuMulticast on group p

• Broadcast source massage using Collision-Free Flooding.

• A node will stop relay if the group ID is not included in its group-list.

Network Initialization and ReconfigurationNetwork Initialization and Reconfiguration

r

memberheadgateway

Cluster-based network

r

member

gateway

Each node’s knowledge (1) 1-hop information:

• Node’s status: head, member, gateway

• If the node is a member, it knows its head.

• If the node is a head, it has a member-list.

• If the node is the node on the backbone, it has a backbone-neighbor-list.

• TDM time-slots: b-time-slot and l-time-slot.

(2) Root knows the largest b-time-slot

Operations used for network initialization and reconfiguration• Node-joining

• Node-leaving

r

new

)(GCNet

memberheadhead

Change to gateway

Node-Joining Operation

Update 1-hop information in expected O(log d) rounds:• update status• update member-list• update backbone-neighbor-list

headGateway

new

w

new

w

u

Case 1: new’s parent w is an internal node of CNet(G). w will have a new leaf of new, w has to update its l-time-slot.

Case 2: new’s parent w is a leaf of CNet(G). w will have a new leaf new, w will change to an internal node, and w’s parent u will have a new internal node w. w has to update its l-time-slot, and u has to update its b-time-slot and l-time-slots.

• Update TDM time-slotsNode-Joining Operation – Continue

Calculate-b-time-slot (l-time-slot) (CNet(G), y)(1) Node y sends a request to ask the nodes of C(y) sending their b-time-slots

back in turn. (2) When node v in C(y) receives the message, it checks the b-time-slots at the

nodes of P(v). If v can find two distinct b-time-slots which are different from those of the others in P(v), v sends back nothing.

Otherwise, v packs all different b-time-slots at the nodes of P(v) and send them back to y at v’s turn.

(3) When y receives the b-time-slots from the nodes of C(y), y selects the minimum positive integer which is different from all received b-time-slots and set it to be y’s b-time-slot.

(4) y sends its b-time-slot back to the root.

y

v

P(v)

C(y): cnnected with y from next level

• Update TDM time-slotsNode-Joining Operation – Continue

• Calculate y’s b-time-slot in rounds d

Node-Joining Operation – Continue

Worst Case Analysis • In node-joining operation, TDM time-slots: d-time-slot and l-time-slot can be updated in rounds.

• , where

d is the largest degree of backbone nodes, and D is largest degree of member nodes.

is the largest b-time-slots, and is the largest l-time-slots

hd 2

)1(2 and )1(2 DDdd

Simulation shows that the worst cast seldomly happen!

Time and Energy Consumption in Broadcast: By using CNet(G) and TDM time-slots,

(1) A broadcast can be completed in rounds, where h is the height of CNet(G), and

(2) In the broadcast, each node needs to be awake in rounds.

)1(h

2

Simulation Simulation

0

5

10

15

20

25

30

35

1 2 3 4 5

number of nodes

Deg

rees

and

tim

e sl

ots

D dΔ δ

100 200 300 400 500

Simulation Model• a cluster-based WSN with the scales of 8×8 units, 10×10 units and 12×12 units, where each unit is 100 meters. • Communication range of a node is 50 meters.• Number of nodes used for testing varies from 64 to 720.

(1) Comparison of the largest degree and largest TDM time-slot

Simulation - ContinueSimulation - Continue

100 200 300 400 5000

100

200

300

400

500

600

1 2 3 4 5

number of nodes

num

ber

of r

ound

s

rounds by CFF Broadcastrounds by DFO Broadcast

100 200 300 400 500

(2) Latency of the broadcast based on depth-first-order (DFO) and collision-free-flooding (CFF)

0

100

200

300

400

500

600

1 2 3 4 5

number of nodes

roun

ds to

be

awak

e

Rounds to be awake for CFFRounds to be awake for DFO

100 200 300 400 500

(3) Energy consumption of the broadcast based on depth-first-order and collision-free-flooding

Summary and Future WorkSummary and Future Work

Summary Time and energy efficient broadcast and multicast

protocols using collision-free flooding Algorithms for reconfiguration of a cluster-based

network Theoretical analysis and simulation

Future Work Revise the broadcast/multicast protocols for Data

gathering (Assign TDM time-slot in opposite direction)

Relax synchronization Add more functionalities for network

reconfiguration:

-- Network split and network merge

2 1 22 1 2

Broadcast

1 2 3 1 3 2 4

Data gathering

Homework & Assignment

•Revise the broadcast/multicast protocols for Data gathering (Assign TDM time-slot in opposite direction). How to solve the problem that more and more data need to be relayed back?

•Relax synchronization.

•Add more functionalities for network reconfiguration:

-- Network split and network merge