energy-efficient wake-up scheduling for data collection and aggregation yanwei wu, member, ieee,...

Energy-Efficient Wake-Up Scheduling for Data Collection and Aggregation

Yanwei Wu, Member, IEEE, Xiang-Yang Li, Senior Member, IEEE,

YunHao Liu, Senior Member, IEEE, and Wei Lou

IEEE TPDS, vol. 21, no. 2, 2010, pp. 275-287.

Outline

2

Introduction System model and assumption Homogeneous wireless sensor networks Heterogeneous wireless sensor networks Formation of data gathering tree Performance evaluation Conclusion

Introduction Wireless sensors are often powered by batteries and have

limited computing and memory resources. Schedule the nodes’ activities to reduce energy consumption.

Previous studies did not consider all possible energy consumption by wireless sensors, Wasted listening, and the state transitions.

3

Introduction

4

Traditionally, the scheduling algorithms often schedule the individual activities for each sensor one by one. Find the best time slots for sending and receiving data.

Communication and interference range. Homogeneous wireless sensor networks Heterogeneous wireless sensor networks

Introduction

5

A scheduling should reduce the state transitions to increase the lifetime of a sensor.

To minimize the sensor’s wake-up times in a scheduling period. Any sensor node needs only to wake up at most twice in our

protocol. Once for continuously receiving all packets from its children nodes

and once for sending its own data to its parent node.

System model and assumption

6

Network System Models Problem Description

Network System Models

7

Tree-based All data will be collected and sent to the sink.

Each wireless node will use a fixed power to communicate with its neighboring sensors. The fixed power transmission by a node vi will define an

interference range. RI(vi) such that the transmission of node vi will interfere the reception of

any node vk when

The physical link <vi, vj >is reliable if vi can communicate with vj.

)( iIik vRvv

)(*)1()( iTiII vRvRR )( iI vR

)( iT vR

vk

vi

vj

Problem Description

8

Energy-Efficient Scheduling Data Collection Tree Construction

Problem Description-Energy-Efficient Scheduling

9

The amount of slots assigned to a node vi for transmitting should be enough.

A node vi with children nodes u1, u2, . . . , uj should be active for receiving at the time slots when these children nodes send data to vi.

T

t

itSi p

wX

1,,

1,1 ,,,, tRitSj XX tSjtRi XX ,,,,

The observed link reliability

Packets received from its children nodes

{0,1}


10

Any node can only be in one of the states.

All transmissions should be interference-free.

1,,,,,,,, tLitPitRitSi XXXX

1,,,, tSjtSi XX

{0,1}


11

Notice that the energy cost by a node vi in all states is

The energy cost for state transitions is

T

tsslptPilsttLircvtRitxtSi tPXPXPXPX

1,,,,,,,, *)****(

T

tLPtLitPiRPtRitPiSPtSitPi EXXEXXEXX

1,1,,,,,1,,,,,1,,,, )******(

The objective of a schedule S is to minimize the summation of these two energy costs.

Slot sizeEnergy consumption of #

{0,1}

12

Problem Description-Data Collection Tree Construction Tree T is given for the data collection or aggregation.

The total energy cost of the optimum activities scheduling based on this tree is the lowest.

The objective is to find a data collection tree T that should satisfy the data requirements of all nodes. NP-hard problem.

Goals

13

This paper use a TDMA for scheduling node activities to reduce the energy consumption.

Focused on the energy cost by the radio. Transmitting, receiving, listening, and sleeping.

Outline Introduction System model and assumption Homogeneous wireless sensor networks Heterogeneous wireless sensor networks Formation of data gathering tree Performance evaluation Conclusion

14

Homogeneous wireless sensor networks-Centralized Activity Scheduling

15

)( iTj vCvii wW

The total number of time slots that node vi should wake up toreceive the data from its children is:

6iW

6iW

4iW

Parent node

Child node

Homogeneous wireless sensor networks- Centralized Activity Scheduling

16

Conflicting Cluster


17

4

1-hop

4-hop

3-hop

2-hop

1/2

The node z is within the distance at most 3RI from node p. The sensors from conflicting clusters Cj,l can only be distributed inside the circle with the radius 3RI

z

q

v

p

RI(p)

RT(p)

)(*)1()( pRpRR TII

u


18

Schedule the clusters in the decreasing order of their weight Wi. Then each child vj will be assigned a consecutive wj time slots

from this chunk.

Time slot

Wj,i :The clusters which conflict with cluster Ci and are scheduled before cluster Ci.

gj,i : Gaps ,non-conflicting clusters, which could be assigned to cluster Ci.

Wi

{w1,w2,w3…}

Homogeneous wireless sensor networks- Centralized Activity Scheduling

19

SW

SYN

SW

SYN

SW

SYN

SW

SYN

Time slotgj,1 gj,2gj,3 gj,4wj,1 wj,2 wj,3 wj,4

7,4 iWC

6,5 iWC

4,6 iWC

3,1 iWC 3,3 iWC

4,2 iWC

4,7 iWC

C4 C5C2

....,,,, 76254 CCCCC

Homogeneous wireless sensor networks- Centralized Activity Scheduling (More discussions)

20

Besides reducing the energy consumption and increasing network throughput Another important issue in WSNs is to reduce the delay.

Instead of scheduling using the available earliest time slots, this paper use the latest available time slots.

SW

SYN

SW

SYN

SW

SYN

SW

SYN


di ui

di,wi=3

ui,wi=4

diui

Homogeneous wireless sensor networks- Distributed Activity Scheduling (TTL)

21

34 CC

7,4 iWC

6,5 iWC4,6 iWC

3,1 iWC 3,3 iWC

3,2 iWC

4,7 iWC

SW

SYN

SW

SYN

SW

SYN

SW

SYN


C4 C7

IamScheduled

27 CC

IamScheduled

Outline Introduction System model and assumption Homogeneous wireless sensor networks Heterogeneous wireless sensor networks Formation of data gathering tree Performance evaluation Conclusion

22

Heterogeneous wireless sensor networks -Centralized Activity Scheduling

23

7,4 iWC

6,5 iWC

4,6 iWC

3,1 iWC 3,3 iWC

3,2 iWC

4,7 iWC

SW

SYN

SW

SYN

SW

SYN

SW

SYN


C2C3

},,{ 1763 CCCB

First divide the sensors into buckets according to their interference radii.

},{ 321 CCB

},{ 452 CCB

Interference Range : B1<B2 <B3

C4

Heterogeneous wireless sensor networks -Distributed Activity Scheduling (TTL)

24

7,4 iWC

6,5 iWC

4,6 iWC

3,1 iWC 3,3 iWC

3,2 iWC

4,7 iWC

SW

SYN

SW

SYN

SW

SYN

SW

SYN


C6C7

},,{ 1763 CCCB

First divide the sensors into buckets according to their interference radii.

C1

Formation of data gathering tree -Connected dominating set

25

Energy efficiency is a critical issue in WSNs since the sensor nodes are with limited energy.

Performance evaluation

26

Randomly placing 32 sensors in a square 5*5 square meters.

Transmission radius as 1m. Interference radius as 2m.


27

Impact of Data Rate


28

Impact of Number of Nodes


29

Impact of Heterogeneous Nodes

Conclusion

30

In this paper proposed an efficient centralized and distributed scheduling algorithms. Remove the unnecessary listening cost Reduce the energy cost for state switching and clock

synchronization.

Every node needs only to wake up at most twice in one scheduling period One for receiving data from its children and one for sending

data to its parent.