Alternatives for Geosensors Network Data Analysis

1
Alternatives for Geosensors Network Data Analysis
GEOPRO Geoinformation Group


Ilka Afonso Reis Instituto Nacional de
Pesquisas Espaciais Universidade Federal de Minas
Gerais
V Brazilian Symposium on GeoInformatics. Campos
do Jordão, November 20-23, 2005
2
but
3
The goal of this presentation
link to
link to
Geosensors network as a means to collect
Space-time models as a means to analyse
spatio-temporal data
4
What are geosensors networks ?

• Sensor networks consist of (small) nodes that can
  measure characteristics of their local
  environment, perform computations, and
  communicate with each other over a wireless
  network.
• (Paskin et al., 2005)

When these devices are deployed over a geographic
area and collect data whose geospatial
information is important, they form a geosensors
network. (Nittel and Stefanidis, 2005)
5
How they work ?
6
Geosensors networks applications

• Monitoring applications
• habitat and wildlife
• environments
• Forest fires detection, air
  pollution, oil and gas escaping, glacier
  movements
• security
• Intruders detection
• disasters alerts
• Floods, earthquakes
• structural
• Building, bridges, towers
• military

• Retrospective Studies data are stored to
  posterior analysis
• Floods simulation (FloodNet project)
• habitat and wildlife
• Great Duck Island
• PODS
• RedWood Trees
• ZebraNet

7
Geosensors networks features

• The most important are
• Nodes have power limitations
• Nodes must know their geographic location
• low-power GPS receivers or
• techniques that use reference points
• Nodes are prone to failures
• Nodes must be unattended
• Nodes are densely deployed

Some networks are designed to collect and send
data continuously
8
Data Routing
Sending data from nodes to the base-station

• Data routing involves communication between the
  nodes
• Communication spends a lot of energy ?

• Routing all data spends a lot of energy
• Data aggregation reduces the messages size and
  provides energy saving. ?

Data aggregation is one of the proposals to save
energy in data routing
9
Data Aggregation and Routing

• Hierarchical Routing Protocols
• Before each data transmission
• Nodes form clusters around a node chosen as the
  cluster-head
• Nodes send their data to the cluster-head
• Cluster-head aggregates these data and sends the
  result to the base-station. ?

Hierarchical routing protocols are the most
efficient alternative when data has to be sent
continuously (Heinzelman, 2000)
10
Geosensors network data analysis
11
(No Transcript)
12
Spatially continuous data analysis
If geosensors can be seen as a point like data
source, geosensors data can be treated as
spatially continuous data ?
13
Spatially continuous data analysis
Geostatistical Model
C(s1, s2 t1, t2) C( d ? ) (under
stationarity and isotropy)
Spatial distance
Temporal distance
14
Spatially continuous data analysis
Separable process
C( d ? )
Non-separable process
Gneiting, 2002
15
Spatially continuous data analysis
Another approach to deal with space-time models
Kriged Kalman Filter (Mardia et al., 1998)
Space-time Kalman Filter (Cressie and Wikle,
2002)
Zt Z(s1 ,t), Z(s2,t), , Z(snt,t) is the
column vector that contains the data for a time
period t ?
16
Spatially continuous data analysis
The prediction for a non-observed location S at
time t Zt(S) is Gaussian Ft(S) mt Ft(S)
Ct Ft(S)
The one-step ahead forecast Zt1 is Gaussian
Ft1 Gt1mt Qt1
17
Geosensors network data analysis
Spatially continuous data is the result of a
network that sends all nodes data. But... what
if the routing protocol aggregates the data ?
18
When data are aggregated during the routing
Cluster-head
data aggregation
19
Area data analysis
If the areas are the same in each time period t
Zit data in area i at time t Zit has mean
µit e variance ?2it ?
20
Area data analysis

• Bayesian framework
• spatial structure is obtained through
  ?i and di prior distributions.

Wij 1, if areas i and j are neighbors Wij 0,
otherwise
CAR
21
Area data analysis
Usually, hierarchical routing protocols change
the cluster-heads and their clusters
periodically. This must to be done because the
tasks of a cluster head spend much energy. So the
clusters areas change in each time period !! ?
22
Final Remarks

• Geosensors networks can be seen as an instrument
  to sample space-time processes and generate lots
  of data
• This main goal of this presentation was to link

• Geosensors networks pose challenges for several
  disciplines (electronics, geographical
  positioning, communication systems, DBMS and data
  analysis).
• They are a new research subject and promise a
  revolution in the physical world observation,
  offering the possibility of a dense sensing of
  the environment.

23
Thank you !!
Ilka Afonso Reis
24
References

• Akyildiz, I. F., Su, W. , Sankarasubramaniam, Y.
  and Cayirci, E. A Survey on Sensor Networks, IEEE
  Communications Magazine August 2002, pages 103
  to 114
• Cressie, N., Wikle, C. (2002) Spacetime Kalman
  filter. In El-Shaarawi, A. , Piegorsch, W (ed.)
  Encyclopedia of Environmetrics, vol. 4, pp
  20452049, John Wiley Sons Ltd, Chichester.
• Elson. J. Estrin, D. (2004) Sensor networks a
  bridge to the physical world. In Raghavendra, C
  S. Sivalingam, K.M. e Znati, T (ed.) Wireless
  Sensor Networks, Kluwer
• Gneiting, T. (2002) Nonseparable, stationary
  covariance functions for space-time data. Journal
  of the American Statistical Association, 97,
  590-600.
• Heinzelman, W. B. Chandrakasan, A.
  Balakrishnan, H. Energy-Efficient Communication
  Protocol for Wireless Microsensor Networks. In
  Proc. of the 33rd. Hawaii Int. Conference on
  System Science, 2000
• W. Heinzelman, Application specific protocol
  architectures for wireless networks, PhD Thesis,
  MIT, 2000.

25
References

• Intanagonwiwat, C. Govindan,R. Estrin, D.
  Directed Diffusion_A Scalable and Robust
  Communication Paradigm for Sensor Networks. In.
  Proc. ACM/MOBICOM, 2002, pp. 5667
• Mardia, K.V., Goodall, C.R., Redfern, E.
  Alonso, F.J. (1998). The kriged Kalman filter
  (with discussion), Test 7, 217285.
• Meng, T. Low-power GPS Receiver Design. In
  Proceedings of the 1998 IEEE Workshop on Signal
  Processing Systems (SiPS '98), October 1998.
• Musunuri, R., Cobb, J.A. Hierarchical-Battery
  Aware Routing in Wireless Sensor Networks., 2005.
  In http//www.utdallas.edu/musunuri/Academics/vt
  c05.pdf
• Nittel, S., Stefanidis, A. (2005) GeoSensor
  Networks and Virtual GeoReality. In Nittel, S.,
  Stefanidis, A. (ed.) GeoSensors Networks, CRC
  Press, 296 p.
• Paskin, Mark A. , Guestrin, Carlos E. and
  McFadden, Jim (2005). A Robust Architecture for
  Inference in Sensor Networks. In Proceedings of
  the Fourth International Symposium on Information
  Processing in Sensor Networks 2005 (IPSN-05).

26
Back up Slides
27
Physical world sensing
Atmospheric pressure Temperature Humidity Wind
Speed ...
Data Collection Platform
Platforms localization
Brazilian Environmental Data Collection System
28
A brief introdução to bayesian inference
29
A brief introdução to bayesian inference
Predictive distributions
30
Area data analysis
Spatio-temporal Model f(µit) a fi ?i dt
?t, i 1, , n e t 1, ,m. ?t
Gaussian (?t-1 ?2?) dt Gaussian (0
?2d)
31
Sensor WEB
Fonte http//sensorweb.geoict.net/whatIsSensorWe
b.htm