[PPT] - Data Proces s ing Techniques Seminar Sensor Nodes Operation Modes, PowerPoint Presentation

SLIDE 1

Data Proces s ing Techniques

Seminar Sensor Nodes Operation Modes, Networks and Applications SS 2012 Martin Waltl 24.07.2012

SLIDE 2

Agenda

What are data processing techniques?
PermaSense project
Multi-hop communication
Model-based approach for temporal reconstruction
Conclusion

2 M a r ti n W a lt l

D

a t a P r

c

e s s i n g T e

SLIDE 3

Why do we need data proces s ing techniques (DPT)?

Problems:

Duplicates, Packet loss, Unordered arrival, clock drifts

Solutions:
1. Improve WSN
Data can be used directly
High complexity and costs
2. Post Data Processing
Requires formal model

3 WSN Data sink Science M a r ti n W a lt l

D

a t a P r

c

e s s i n g T e Reas ons

SLIDE 4

Different approaches for DPT

Establish a global time base within the WSN  global

synchronization (Flooding Time Synchronization Protocol - FTSP) [6]

Post data processing based on domain specific knowledge [10]

– Microseismics measurements [8] – Sun light measurements [9]

Model-based approach to reconstruct the temporal order of

packets using packet header information [1] 4 M a r ti n W a lt l

D

a t a P r

c

e s s i n g T e

SLIDE 5

Steps in data proces s ing

5 WSN Data sink Science DPT Data Processing T echnique

[1] M. Keller, L. Thiele, and J. Beutel. Reconstruction of the correct temporal order of sensor network data. In Information Processing in Sensor Networks (IPSN), 2011 10th International Conference on, pages 282 - 293, April 2011

Recons truct temporal packet order M a r ti n W a lt l

D

a t a P r

c

e s s i n g T e

SLIDE 6

PermaSens e project

WSN in high-mountain area
Prototype for deployment in harsh environments
Goal:

Observe permafrost changes in the Swiss Alps

Challenges : snow cover, lightning strokes, rock falls,

temperature variations [-40°C,60°C], high altitude

Requirements :
1. Precision Sensing
2. Reliability in harsh environment
3. Durability and energy constraints

6 M a r ti n W a lt l

D

a t a P r

c

e s s i n g T e

SLIDE 7

WSN deployment at the Matterhorn (3450m)

25 nodes
Spacing 10-

150m 7 M a r ti n W a lt l

D

a t a P r

c

e s s i n g T e

http://www.permasense.ch/uploads/pics/2011jf001981-op01.jpg

SLIDE 8

Sens or nodes

Components

– Shockfish TinyNode – Sensor Interface Board (SIB) – 6 sensors – connected to SIB – 1GB SD card – local storage – Li-SOCl2 battery

Power consumption

– Ultra low-power Dozer protocol – Average power consumption 148µA – ~ 3 years of operation

8

Deployment at site2 Sensor node1

(1)http://www.permasense.ch/uploads/RTEmagicC_permabox_exploded.jpg.jpg (2)http://www.permasense.ch/uploads/RTEmagicC_rod_sketch.jpg.jpg

M a r ti n W a lt l

D

a t a P r

c

e s s i n g T e

SLIDE 9

Multi-hop communication (Dozer)

Dozer multi-hop communication protocol designed for WSN
Optimized for low-power consumption of sensor nodes
Tree management (dynamic topology)
Periodic duty cycle:

Sensing | Transmission | Sleep 9

Communication artefacts

– Unordered arrival – Duplicate generation – Data loss T ree architecture

M a r ti n W a lt l

D

a t a P r

c

e s s i n g T e

SLIDE 10

Dozer – Trans mis s ion bas ed on TDMA mechanis m

10

Transmission is based on a periodic TDMA frame
Beacon (B) is used for local s ynchronization
Contention Window - acceptance of children
Children are assigned to one time slot
Communication failure

Periodic TDMA frame format in Dozer

M a r ti n W a lt l

D

a t a P r

c

e s s i n g T e

SLIDE 11

Dozer – Communication example

11

T

extmasterformat bearbeiten

– Zweite Ebene

Dritte Ebene

– Vierte Ebene » Fünfte Ebene

[7] Figure 3

M a r ti n W a lt l

D

a t a P r

c

e s s i n g T e

SLIDE 12

Wrap up

Our goal:

Reconstruct temporal order of packets

What we know:

– WSN architecture – Transmission via multi-hop communication

Next steps:

1. Specify a system model to describe the data acquisition and transmission process 2. Develop a formal model to reconstruct temporal packet order

12 WSN Data sink Science DPT Filtered Data M a r ti n W a lt l

D

a t a P r

c

e s s i n g T e

SLIDE 13

Sys tem model

Formal model of a sensor network

– Tree architecture – Multi-hop communication – No global synchronization

Sink has an absolute clock and does not suffer from restarts
Nodes suffer from local clock drifts and restarts

– Warm restarts  reset time – Cold restarts  resets time, sequence counter and message queue (data los s !!)

Packet = (o, s, p, ts~, tb)

–

– sensor node I

– s – sequence number  (s + 1) mod smax – p – payload – ts~ – estimated sojourn time – tb – absolute arrival time a the sink node

13 M a r ti n W a lt l

D

a t a P r

c

e s s i n g T e

SLIDE 14

Model-bas ed approach

14

Goal:

Reconstruct temporal packet order and enhance data quality

Only performed on header information (no sensor data)

– Sequence number – Arrival time stamp (tb) at sink – System model

M a r ti n W a lt l

D

a t a P r

c

e s s i n g T e

SLIDE 15

Step 1: Es timation of packet generation time

15 M a r ti n W a lt l

D

a t a P r

c

e s s i n g T e

SLIDE 16

Step 2: Duplicate filtering

16 G = (V,E) M a r ti n W a lt l

D

a t a P r

c

e s s i n g T e

SLIDE 17

Step 3: Epoch as s ignment – What is an epoch?

17 M a r ti n W a lt l

D

a t a P r

c

e s s i n g T e

[1] Figure 3

SLIDE 18

Step 3: Epoch as s ignment - Algorithm

18 M a r ti n W a lt l

D

a t a P r

c

e s s i n g T e

SLIDE 19

Step 4: Forward / Backward reas oning

19 Backward reasoning Forward reasoning M a r ti n W a lt l

D

a t a P r

c

e s s i n g T e

SLIDE 20

Evaluation and performance

Case study using data from the PermaSense project

– 3 deployment phases – Ground truth data from SD card of sensor nodes (reference)

Metrics:

– Packet acceptance rate – Correctness of the derived packet sequence – Improvement of generation time intervals

Results
Comparison to simple heuristics
High packet acceptance rate for model-based approach
Delivered correct temporal packet order
Generation time intervals reduced by 90%  from 100s – 2.6s

20 M a r ti n W a lt l

D

a t a P r

c

e s s i n g T e

SLIDE 21

Conclus ion

Reconstruction of temporal order requires a formal system model

– What is the architecture? – How is the data transmitted within the network?

+ Reduce complexity of WSN  cheaper sensor nodes + Enhance system life time − Model formulation is complex and never correct  simplification 21 Data Processing T echnique M a r ti n W a lt l

D

a t a P r

c

e s s i n g T e

SLIDE 22

Thanks for your attentation! Ques tions ?

SLIDE 23

PermaDAQ architecture

23

T

extmasterformat bearbeiten

– Zweite Ebene

Dritte Ebene

– Vierte Ebene » Fünfte Ebene

http://www.permasense.ch/typo3temp/pics/2cab5fb 804.png

Live-Data Viewer: http://data.permasense.ch/
Sensor nodes position:

http://www.permasense.ch/de/data/permasense-data.html

M a r ti n W a lt l

D

a t a P r

c

e s s i n g T e

SLIDE 24

Backup s lides

SLIDE 25

References (1)

[1] M. Keller, L. Thiele, and J. Beutel. Reconstruction of the correct temporal

rder of sensor network data. In Information Processing in Sensor Networks

(IPSN), 2011 10th International Conference on, pages 282 - 293, April 2011 [2] J. Beutel, S. Gruber, A. Hasler, R. Lim, A. Meier, C. Plessl, I. T alzi, L. Thiele,

C. T

schudin, M. Woehrle,and M. Yuecel. PermaDAQ: A scientific instrument for precision sensing and data recovery in environmentalextremes. In Information Processing in Sensor Networks, 2009. IPSN 2009. International Conference on, pages 265 { 276, April 2009 [3] G. Werner-Allen, K. Lorincz, J. Johnson, J. Lees, and M. Welsh. Fidelity and yield in a volcano monitoring sensor network. In Proceedings of the 7th symposium on Operating systems design and implementation, OSDI '06, pages 381{396, Berkeley, CA, USA, 2006. USENIX Association [4] G. Barrenetxea, F . Ingelrest, G. Schaefer, M. Vetterli, O. Couach, and M.

Parlange. 2008. SensorScope: Out-of-the-Box Environmental Monitoring. In

Proceedings of the 7th international conference on Information processing in sensor networks (IPSN '08). IEEE Computer Society, Washington, DC, USA, 332-343.

25 M a r ti n W a lt l

D

a t a P r

c

e s s i n g T e

SLIDE 26

References (2)

[5] G. T

lle, J. Polastre, R. Szewczyk, D. Culler, N. T

urner, K. T u, S. Burgess, T. Dawson, P . Buonadonna, D. Gay, and W. Hong. 2005. A macroscope in the

redwoods. In Proceedings of the 3rd international conference on Embedded networked

sensor systems (SenSys '05). ACM, New York, NY, USA, 51-63 [6] M. Maróti, B. Kusy, G. Simon, and Á. Lédeczi. 2004. The flooding time synchronization protocol. In Proceedings of the 2nd international conference on Embedded networked sensor systems (SenSys '04). ACM, New York, NY, USA, 39- 49 [7] N. Burri, P . von Rickenbach, and R. Wattenhofer. Dozer: Ultra-Low Power Data Gathering in Sensor Networks. In Information Processing in Sensor Networks, 2007. IPSN 2007. 6th International Symposium on, pages 450 - 459, April 2007 [8] M. Lukac, P . Davis, R. Clayton, and D. Estrin. 2009. Recovering temporal integrity with Data Driven Time Synchronization. In Proceedings of the 2009 International Conference on Information Processing in S ensor Networks (IPSN '09). IEEE Computer Society, Washington, DC, USA, 61-72

26 M a r ti n W a lt l

D

a t a P r

c

e s s i n g T e

SLIDE 27

References (3)

[9] J. Gupchup, R. Musâloiu-E., A. Szalay, and A. T

erzis. Sundial: Using sunlight

to reconstruct global timestamps. In U. Roedig and C. Sreenan, editors, Wireless Sensor Networks, volume 5432 of Lecture Notes in Computer Science, pages 183 - 198. Springer Berlin / Heidelberg, 2009 [10] R. Szewczyk, J. Polastre, A. Mainwaring, and D. Culler. Lessons from a sensor network expedition. In H. Karl, A. Wolisz, and A. Willig, editors, Wireless Sensor Networks, volume 2920 of Lecture Notes in Computer Science, pages 307 - 322. Springer Berlin / Heidelberg, 2004

27 M a r ti n W a lt l

D

a t a P r

c

e s s i n g T e

SLIDE 28

Environmental monitoring

Gathering of environmental data

– Understanding of processes – Improve existing models – Oberservaton of long term evolutions

High data accuracy  Science
Long system life time with unattended operation
Examples:

– PermaSense – observation of permafrost changes in the Swiss Alps [2] – Volcano monitoring sensor network – seismic and acoustic data [3] – SensorScope – Monitoring of climate changes [4] – Monitoring of redwood trees – temperature, humidity, photosynthetically active solar radiation [5]

28 M a r ti n W a lt l

D

a t a P r

c

e s s i n g T e

SLIDE 29

PermaDozer Integration

29 M a r ti n W a lt l

D

a t a P r

c

e s s i n g T e

SLIDE 30

Dozer protocol in detail (1)

30 M a r ti n W a lt l

D

a t a P r

c

e s s i n g T e

SLIDE 31

Dozer protocol in detail (2)

31 M a r ti n W a lt l

D

a t a P r

c

e s s i n g T e

SLIDE 32

Epoch as s ignment – Algorithm s teps

32 M a r ti n W a lt l

D

a t a P r

c