Adaptive Sampling System fWireless Sensor Network Distributed Intelligent Systems EPF Lausanne, December 2011
Corina Stoenescu –Graduate Student EneSergiu Gaman- Graduate Student Comput
for Light Monitoring using a of MICAz Motes
rgy Management and Sustainabilityter Science
|Motivation
importance of wireless sensor neenvironmental, Home Energy Mangyunderlying issue- energy autono
|Obj i|Objective
D i d ti l th t
Corina Stoenescu –Graduate Student Ene
Design an adaptive algorythm to re- decreasing the number of meas- while keeping a satisfactory rep
Sergiu Gaman- Graduate Student Computp g y p
Evaluate the deployment of the alg
etworks in a multitude of applications: nagement, industrial, structural, military… g yomy limitation
d th ti i WSN b
rgy Management and Sustainability
educe the power consumption in WSN by:surements & transmissions,presentativity of the studied phaenomena.ter Science p y pgorithm on a real WSN
|Experimental Setup
7 MICAz motes - aligned7 MICAz motes - aligned- measuring light- transmitting the- traversed by an
! Ambient light reduction
Corina Stoenescu –Graduate Student Ene
! Ambient light reduction
! Minimum distance between motes
Sergiu Gaman- Graduate Student ComputAdditional Simulator – offline te- distinctio
t intensitye values to a sink e-puck carrying a shadowing blade
rgy Management and Sustainabilityter Science esting of different algorithmson between code and hardware limitations
|Methoda combination of :...a combination of :
Signal adaptive transmission rCollaborative adaptation of mp
|Performance Metricrelative to the results of the uniforrelative to the results of the unifor
Corina Stoenescu –Graduate Student EneSergiu Gaman- Graduate Student Comput
rates measurement rates
rm sampling method (Mc=1/3)rm sampling method (Mc=1/3)
rgy Management and Sustainabilityter Science
|1. Signal adaptive transmissi
Key points:
! Relate to a forecasted value ! Relate to a forecasted value rather than the last transmission
! Adaptive threshold
threshold=t%* mean(s1,s2)( )
* t- slightly above the observed noise
ion rates
|1. Signal adaptive transmiss
simulated results for one mote
Light intenssity
Time
ion rates
|2. Collaborative adaptation o
Task allocation:
Divide the motes into guardians (G)Divide the motes into guardians (G)sleepers (S)
+ highly beneficial in decreasing the highly beneficial in decreasing themeasurement and transmission rates
+ less risk of skipping important measurempp g p
! dependent on the given environment
! single fault vulnerable unless dynamic tas
of measurement rates
)
G GS S S S S S
)
ments
k allocation
|2. Collaborative adaptation oGuardian
*counter’s role:
of measurement ratesSleeper
: going back to sleep after five samples
|Combined method on-boa3D representation3D representation
Norm
alized intensityM
Corina Stoenescu –Graduate Student Ene
Mote
Sergiu Gaman- Graduate Student Comput
=
ard results
rgy Management and Sustainabilityter Science
Time
=>
|Combined method on-boa2D representation2D representation
Mc~ TMc~ T~ Ns/Ng~ Ns/Nggg
Corina Stoenescu –Graduate Student EneSergiu Gaman- Graduate Student Comput
ard results
Mot
e
rgy Management and Sustainabilityter Science
TiTime
|Combined method on-boaSink interpolationSink interpolation
Corina Stoenescu –Graduate Student Ene
Norm
alized i(interpolated
Sergiu Gaman- Graduate Student Comput
intensity d)
Tim
ard results
rgy Management and Sustainabilityter Science
me
|Error sources|deployment on real hardwarepower consumption (especially sinl d d bl limited and unstable transmission r
|Optimization possibilitiesDynamic allocation of the guardian
Corina Stoenescu –Graduate Student Ene
y gThreshold check before propagatinEnvironment learning stage (befor
Sergiu Gaman- Graduate Student Comput
nk& guardians) ranges
n role (increase in robustness)
rgy Management and Sustainability
ng an alert messagere deciding upon t% and role allocation)
ter Science
|Conclusion
Adaptive sapling algorithms have to the increase in the autonomy of wto the increase in the autonomy of w
The best adaptation techniques reenvironment.
! Th h d d l t d b
Corina Stoenescu –Graduate Student Ene
! The hardware deployment and robcarefully addressed.
Sergiu Gaman- Graduate Student Comput
an important potential of contributing wireless sensor network nodeswireless sensor network nodes.
ely on an observation of the given
b t f th l ith h t b
rgy Management and Sustainability
bustness of the algorithms have to be
ter Science
Thank you!