Network Design and In-network Data Analysis for Energy-Efficient Distributed Sensing

1
Network Design and In-network Data Analysis for
Energy-Efficient Distributed Sensing

• Liang Cheng, Ph.D., Associate Professor
• Laboratory Of Networking Group (LONGLAB)
• Department of Computer Science and Engineering
• In Collaborations with ATLSS Colleagues

2
Outline

• Our research in distributed sensing sponsored by
  NSF
• http//www.cse.lehigh.edu/cheng/LONGLAB_Liang_Che
  ng.pdf
• Wireless sensor networks for bridge monitoring
• Network design for interference mitigation
• Distributed in-network data analysis
• Conclusions

3
Subsurface monitoring techniques
GPR
TDR
air
underground
Sensing Area
Wireless Sensor Node
Wireless Sensor Node
Wireless Signal Networks
Crimp in cable
Global Sensing
Soil Moisture Sensor
S. Yoon, L. Cheng, E. Ghazanfari, S. Pamukcu, and
M. T. Suleiman, A radio propagation model for
wireless underground sensor networks, IEEE
Globecom, Houston, TX, December 2011.
4
Experiments point vs. global sensing
Wireless Vantage Pro2
Water Leakage 2
Water Leakage 1
Soil moisture sensor
MICAz (WiSNS)
5
Point sensing vs. global sensing
S. Yoon, E. Ghazanfari, L. Cheng, S. Pamukcu, M.
T. Suleiman, Subsurface event detection and
classification using wireless signal networks,
Sensors, Vol. 12, No. 11, 2012.
No Change
Water Leakage Event 1
Water Leakage Event 2
6
Outline

• Our research in distributed sensing sponsored by
  NSF
• Wireless sensor networks for bridge monitoring
• Network design for interference mitigation
• Distributed in-network data analysis
• Conclusions

7
Why bridge monitoring?

• Critical to the economy and public safety

• FHWA 2008 25

8
Why wireless sensing?

• Routine visual inspection
• Wired monitoring
• the Stone Cutter Bridge in Hong Kong has more
  than 1200 sensors

9
Wireless sensor network challenges

• Network agility
• June September 2006
• Glen Ellen shaking magnitude 4.4 on 08/02/2006
• 30
• Multi-hop (2008)
• 10 hours for getting 80 seconds of data (1KHz)
  from 56 sensors
• Single-hop (2011)
• 5 minutes for 240KB data from 20 sensors

Liang Cheng and Shamim Pakzad, Agility of
Wireless Sensor Networks for Earthquake
Monitoring of Bridges, the Sixth International
Conference on Networked Sensing Systems
(INSS'09), Carnegie Mellon University,
Pittsburgh, USA, June 17 - 19, 2009.
10
Energy-efficient wireless sensor networks with
resource constraints

• Network design
• Critical radio range determination
• Hidden terminal problem solution
• In-network data analysis
• Distributed system identification

11
Outline

• Our research in distributed sensing sponsored by
  NSF
• Wireless sensor networks for bridge monitoring
• Network design for interference mitigation
• Distributed in-network data analysis
• Conclusions

12
Mitigating exposed interference

• Critical radio range determination
• Reduce wireless collision probability
• Prolong network lifetime

13
Bernoulli graphs

• Infinite radius, unreliable links
• Bela Bollobas, Random Graphs, Cambridge
  University Press, 1985
• A graph consists of N nodes where edges are
  chosen independently and with probability p
• Find the critical p ensuring a connected graph
• PclogNc(N)/N

14
2D wireless networks

• Finite radius, reliable links
• Gupta and Kumar, Critical power for asymptotic
  connectivity in wireless networks, Stochastic
  Analysis, Control, Optimization Applications,
  1998.
• A unit area containing N nodes, each having the
  same communication radius r
• Find the critical r ensuring a connected graph
• RclogNc(N)/N

15
Gap between theory and practice

• RclogNc(N)/N

16
1D wireless networks

• Finite radius, reliable links
• Li and Cheng, Determinate Bounds of Design
  Parameters for Critical Connectivity in Wireless
  Multi-hop Line Networks, IEEE WCNC 2011.
• A unit length containing N nodes, each having the
  same communication radius r
• Find the critical r ensuring a connected graph
• lnN/N lt Rc lt 2lnN/N

17
A bridge sensor network

• Finite radius, unreliable links
• A unit length containing N nodes, each having the
  same communication radius r with link
  connectivity probability p
• Find the critical r ensuring a connected graph
• lnN/N lt Rc lt 2lnN/(pN)

18
Mitigating hidden interference

• Hidden terminal problem
• Collision at will
• Aloha (1971)
• Collision avoidance
• IEEE 802.11 (1997)
• Collision detection
• ?

19
Messages vs. pulses
20
Hidden terminal revisited

• Hidden terminal no longer hidden!
• Collision detection

21
Throughput increased

• J. Peng, L. Cheng, and B. Sikdar, A Wireless MAC
  Protocol with Collision Detection, IEEE
  Transactions on Mobile Computing, Vol. 6, No. 12,
  pp. 1357-1369, 2007.

22
Outline

• Our research in distributed sensing sponsored by
  NSF
• Wireless sensor networks for bridge monitoring
• Network design for interference mitigation
• Critical radio range determination
• Hidden terminal problem solution
• Distributed in-network data analysis
• Distributed system identification
• Conclusions

23
Modal parameters of dynamic systems

• Eigenvalue decomposition of the state matrix (Ad)
  results in the matrices of eigenvalues (?is) and
  eigenvectors (?is)
• The natural frequencies ?i and damping ratios ?i

24
Traditional modal identification

• Expectation-Maximization (EM)
• estimates unknown parameter (?), given the
  measurement data (Y) in the presence of some
  hidden variables (Y ) (Dempster, 1977)

25
Distributed modal identification
26
Evaluation results

• O(1/n) consumed energy comparing to the
  centralized method in n-hop WSNs
• S. Dorvash, S. Pakzad, and L. Cheng, An iterative
  modal identification algorithm for structural
  health monitoring using wireless sensor networks,
  Earthquake Spectra, Vol. 29, No. 2, pp. 339-365,
  May 2013.

27
Outline

• Our research in distributed sensing sponsored by
  NSF
• Wireless sensor networks for bridge monitoring
• Network design for interference mitigation
• Distributed in-network data analysis
• Conclusions

28
Conclusions

• Energy-efficient wireless sensor networks with
  resource constraints
• Network design
• Critical radio range determination (1985, 1998,
  2011)
• Hidden terminal problem solution (1971, 1997,
  2007)
• In-network data analysis
• Distributed system identification
  (Expectation-maximization 1977, frequency
  responses 2004, distributed modal identification
  2011)

29
Acknowledgement

• National Science Foundation (NSF)
• Commonwealth of Pennsylvania
• Department of Community and Economic Development
  via PITA
• Christian R. Mary F. Lindback Foundation
• Siavash Dorvash, Xu Li, Dr. Shamim Pakzad, Dr.
  Jun Peng

30
Q A

• cheng_at_cse.lehigh.edu
• 610-758-5941
• Liang Cheng
• Computer Science Engineering
• 19 Memorial Drive West, Bethlehem, PA 18015

31
Evaluation Scenarios
32
Resource constraints of sensor nodes

• Imote2
• Transceiver CC2420
• Battery
• Rechargeable 300 mWh/cm3
• Zinc-air 1050-1560 mWh/cm3
• CPU 13416 MHz
• Memory 256kB SRAM, 32MB FLASH, 32MB SDRAM
• Demo
• A freshman lab project of my Eng5 students