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

Data Proces s ing Techniques Seminar Sensor Nodes Operation Modes, Networks and Applications SS 2012 Martin Waltl 24.07.2012

M 2 a r ti n Agenda W a lt • l What are data processing techniques? - D • PermaSense project a t • Multi-hop communication a P • Model-based approach for temporal reconstruction r o • Conclusion c e s s i n g T e

3 M a r ti n Why do we need data proces s ing techniques (DPT)? W a lt l - D a t WSN Science Data sink a P • Problems: Duplicates, Packet loss, Unordered arrival, clock drifts r o Reas ons • Solutions: 1. Improve WSN c • Data can be used directly e • High complexity and costs s 2. Post Data Processing s • Requires formal model i n g T e

4 M a r ti n Different approaches for DPT W a • lt Establish a global time base within the WSN  global l synchronization (Flooding Time Synchronization Protocol - FTSP) - [6] D a • Post data processing based on domain specific knowledge [10] t – Microseismics measurements [8] a – Sun light measurements [9] P r o c e • Model-based approach to reconstruct the temporal order of s packets using packet header information [1] s i n g T e

M 5 a r ti n Steps in data proces s ing W a lt l - D a t WSN Science DPT Data sink a Recons truct temporal packet order P r Data Processing T echnique o c e s s i n [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 g 2011 T e

6 M a r ti n PermaSens e project W a lt WSN in high-mountain area • Prototype for deployment in harsh environments • l Goal : Observe permafrost changes in the Swiss Alps • - Challenges : snow cover, lightning strokes, rock falls, • temperature variations [-40°C,60°C], high altitude D Requirements : a • 1. Precision Sensing 2. Reliability in harsh environment t 3. Durability and energy constraints a P r o c e s s i n g T e

7 M a r ti n WSN deployment at the Matterhorn (3450m) W a • lt 25 nodes l • Spacing 10- - 150m D a t a P r o c e s s i http://www.permasense.ch/uploads/pics/2011jf001981-op01.jpg n g T e

8 M a r ti n Sens or nodes W a lt • Components – Shockfish TinyNode l – Sensor Interface Board (SIB) - – 6 sensors – connected to SIB D – 1GB SD card – local storage a – Li-SOCl2 battery t a Sensor node1 • Power consumption P – Ultra low-power Dozer protocol r – Average power consumption o 148µA c – ~ 3 years of operation e s s i (1)http://www.permasense.ch/uploads/RTEmagicC_permabox_exploded.jpg.jpg Deployment at site2 n (2)http://www.permasense.ch/uploads/RTEmagicC_rod_sketch.jpg.jpg g T e

9 M a r ti n Multi-hop communication (Dozer) W a • Dozer multi-hop communication protocol designed for WSN lt • Optimized for low-power consumption of sensor nodes l - • Tree management (dynamic topology) D • Periodic duty cycle: Sensing | Transmission | Sleep a t a P • Communication artefacts r – Unordered arrival o – Duplicate generation c – Data loss e s s i T ree architecture n g T e

M 10 a r ti n Dozer – Trans mis s ion bas ed on TDMA mechanis m W a lt • Transmission is based on a periodic TDMA frame l • Beacon (B) is used for local s ynchronization - • Contention Window - acceptance of children D • Children are assigned to one time slot a • t Communication failure a P r o c e s s i Periodic TDMA frame format in Dozer n g T e

M 11 a r ti n Dozer – Communication example W a • lt T extmasterformat bearbeiten l – Zweite Ebene • - Dritte Ebene – Vierte Ebene D » Fünfte Ebene a t a P r o c e s s i [7] Figure 3 n g T e

12 M a r ti n Wrap up W a Filtered lt Data l - D a t WSN Science Data sink DPT a P • Our goal: Reconstruct temporal order of packets r o • What we know: c – WSN architecture e – Transmission via multi-hop communication s • Next steps: s 1. Specify a system model to describe the data acquisition and transmission i process n 2. Develop a formal model to reconstruct temporal packet order g T e

M 13 a r ti n Sys tem model W a • lt Formal model of a sensor network l – Tree architecture – - Multi-hop communication – No global synchronization D • Sink has an absolute clock and does not suffer from restarts a • t Nodes suffer from local clock drifts and restarts a – Warm restarts  reset time – Cold restarts  resets time, sequence counter and message queue P (data los s !!) r o • c Packet = (o, s, p, ts~, tb) e – o – sensor node I – s – sequence number  (s + 1) mod smax s – p – payload s – ts~ – estimated sojourn time i – tb – absolute arrival time a the sink node n g T e

M 14 a r ti n Model-bas ed approach W a lt l - D a t a P r o • Goal : Reconstruct temporal packet order and enhance c data quality e • Only performed on header information (no sensor data) s – Sequence number s – Arrival time stamp (tb) at sink i – System model n g T e

15 M a r ti n Step 1: Es timation of packet generation time W a lt l - D a t a P r o c e s s i n g T e

16 M a r ti n Step 2: Duplicate filtering W a lt l - D a t a P r o c e s s i G = (V,E) n g T e

17 M a r ti n Step 3: Epoch as s ignment – What is an epoch? W a lt l - D a t a P r o c e s s i n [1] Figure 3 g T e

18 M a r ti n Step 3: Epoch as s ignment - Algorithm W a lt l - D a t a P r o c e s s i n g T e

19 M a r ti n Step 4: Forward / Backward reas oning W a lt l - D a Forward reasoning Backward reasoning t a P r o c e s s i n g T e

M 20 a r ti n Evaluation and performance W a • lt Case study using data from the PermaSense project l – 3 deployment phases - – Ground truth data from SD card of sensor nodes (reference) D a • Metrics: t – Packet acceptance rate a – Correctness of the derived packet sequence P – Improvement of generation time intervals r o c • Results e • Comparison to simple heuristics s • High packet acceptance rate for model-based approach s • Delivered correct temporal packet order i • Generation time intervals reduced by 90%  from 100s – 2.6s n g T e

21 M a r ti n Conclus ion W a • lt Reconstruction of temporal order requires a formal system model l – What is the architecture? - – How is the data transmitted within the network? D + Reduce complexity of WSN  cheaper sensor nodes a + Enhance system life time t − Model formulation is complex and never correct  simplification a P r Data Processing T echnique o c e s s i n g T e

Thanks for your attentation! Ques tions ?

M 23 a r ti n PermaDAQ architecture W a • T extmasterformat bearbeiten lt – Zweite Ebene l • Dritte Ebene - – Vierte Ebene D » Fünfte Ebene a t a P r o c e s http://www.permasense.ch/typo3temp/pics/2cab5fb s 804.png • Live-Data Viewer: http://data.permasense.ch/ i • Sensor nodes position: n http://www.permasense.ch/de/data/permasense-data.html g T e

Backup s lides

25 M a r ti n References (1) W a lt [1] M. Keller, L. Thiele, and J. Beutel. Reconstruction of the correct temporal order of sensor network data. In Information Processing in Sensor Networks l (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, D C. T schudin, M. Woehrle,and M. Yuecel. PermaDAQ: A scientific instrument for a precision sensing and data recovery in environmentalextremes. In t Information Processing in Sensor Networks, 2009. IPSN 2009. International a Conference on, pages 265 { 276, April 2009 P [3] G. Werner-Allen, K. Lorincz, J. Johnson, J. Lees, and M. Welsh. Fidelity and r yield in a volcano monitoring sensor network. In Proceedings of the 7th o symposium on Operating systems design and implementation, OSDI '06, c pages 381{396, Berkeley, CA, USA, 2006. USENIX Association e [4] G. Barrenetxea, F . Ingelrest, G. Schaefer, M. Vetterli, O. Couach, and M. Parlange. 2008. SensorScope: Out-of-the-Box Environmental Monitoring. In s s Proceedings of the 7th international conference on Information processing in sensor networks (IPSN '08). IEEE Computer Society, Washington, DC, USA, 332-343. i n g T e