Smart Dust Sensor Mote Characterization, Validation, Fusion and ...

Presentation for Master of Science
Smart Dust Sensor Mote
Characterization, Validation, Fusion
and Actuation
Yao-Jung Wen
11/19/2004

Outline
• Introduction of “Intelligent daylighting control system for
commercial buildings” project
• Introduction of smart dust mote
• Mote sensor characterization
– Illuminance characterization
– Temperature characterization
– Accelerometer evaluation
• BESTnet
• Mote-FVF: Fuzzy validation and fusion for sensor network
• Mote-based actuation
• Conclusion
• Future research

Intelligent Daylighting Control
System for Commercial Buildings
• Goal:
– Balancing conflicting lighting preferences of
occupants sharing common lighting switch
– Energy conservation (~50% annual energy saving
by implementing this system)
• Mote sensor network in the system
– Perception (illuminance and occupancy sensing)
– Actuator (lighting actuation)

Smart Dust Motes Technology
• Smart Dust Mote
– Characteristic
• Capacity of sensing, computation, wireless
communication and data storage
• Miniature size, inexpensive, network ready
– Limitation
• Limited energy
• Limited communication range
• Uncalibrated sensors
• Less accurate sensors
• Small sensing scope
• Low robustness to local disturbances

Mote Sensor Characterization
• Illuminance characterization • Temperature characterization
1200
1000
Data points
Fitting curve
90
80
70
Illuminance (lux)
Illuminance (lux)
800
600
400
200
0
600 650 700 750 800 850 900 950 1000
Digital Reading
1000
900
800
700
600
500
400
300
200
100
Sensorboard01
Sensorboard02
Sensorboard03
Sensorboard04
Sensorboard05
Sensorboard06
Sensorboard07
Sensorboard08
Sensorboard09
Sensorboard10
Sensorboard11
Sensorboard12
0
600 650 700 750 800 850 900 950 1000 1050
Digital Reading
Temperature residuals ( °C )
Temperature residuals ( °C )
60
50
data1
data2
40
data3
data4
30
data5
20
data6
data7
10
data8
data9
0
data10
data11
-10
0 100 200 300 400 500 600 700 800
SensorReadings
100
Data points
Global mapping curve
80
Xbow's curve
60
40
20
0
-20
0 100 200 300 400 500 600 700 800
Sensor readings

Mote Sensor Characterization
850
• Accelerometer evaluation
Sensor Reading
800
750
700
650
600
550
Chair is empty
mote13 x
mote13 y
mote14 x
mote14 y
Start sitting
in the chair
Sit steady in the chair
Swing the chair
while sitting on it
Leave the chair
500
0 200 400 600 800 1000 1200 1400 1600 1800 2000
Time (sec)
Accelometer Readings
545
540
535
Sensor Reading
530
525
520
515
510
600 700 800 900 1000 1100 1200 1300 1400 1500 1600
Time (sec)

BESTnet
• BESTnet v. 1.0
– Centralized single-hop sensor network
– 6 motes in a 3-by-2 matrix under a dimmable lighting fixture
• BESTnet v. 1.1
– Centralized multi-hop sensor network
– 10 motes (6 sensor nodes 4 relay nodes) distributed to 3 desks
• Message packet
Destination address
2 bytes
Active message handler ID 1 byte Source mote ID 2 bytes
Group ID 1 byte Last sample no. 2 bytes
Message length 1 byte ADC channel 2 bytes
Payload Up to 29 bytes Data array 102bytes

Challenges of Mote Sensor Networks in
Lighting Application
• Mote sensor nodes:
– Small individual sensing scope
– Relatively less accurate sensors
– Low robustness to local disturbances
• Sensor networks:
– Massive information
– Data packet collision and interference
– Nontrivial to drive a mathematical model
• Lighting environment:
– Highly discontinuous state
– Illuminance variation

How Fuzzy Validation and Fusion
Overcome the Challenges?
• Fuzzy logic
– Easy to implement
– No complicated mathematic model involved
– Work well for sensor fusion
• Sensor validation and fusion
– Isolate faulty sensor data
– Robust to individual local bias
– Boost accuracy
– Extract pertinent local and global information

Mote-FVF
• Mote-FVF (Fuzzy Validation and Fusion)
algorithm
– Validation
• Majority voting
• Dynamic centered validation curve generating
• Confidence value assigning
– Fusion
• Weighted averaging
– Prediction
• Exponential weighted moving average time series
predicting

Mote-FVF
Validation
• Fuzzy rules for validation curve:
Def: Cor(x) – the majority voting result
Var(x) – the difference between Cor(x) and predicted reading
IF Var(x) small THEN move toward the Cor(x) a small amount
IF Var(x) medium THEN move toward the Cor(x) a medium amount
IF Var(x) large THEN move toward the Cor(x) a large amount
Sensor confidence ()
1
0.8
0.6
0.4
0.2
0
x ( ) cor x x ˆx
Fuzzy validation gate
Measurement

Mote-FVF
Fusion
• Weighted average of sensor readings and the
predicted value
Def:
x
z
f
− fused value
−
th
i reading from the i sensor
th
( zi
) − confidence value of the i sensor reading
σ
α − adaptive parameter
ω −scaling factor
x
f
=
n
z
i
1
n
∑
∑
1
σ( z)
+ αxˆ
i ω
σ( z ) + α
i ω

Mote-FVF
Prediction
• Fuzzy rule for adaptive parameter ( α )
– IF change of readings small THEN α large,
– IF change of readings medium THEN α medium,
– IF change of readings large THEN α small.
• Time series predictor
xˆ( k + 1) = α xˆ( k) + (1 −α
) x ( k)
f

Mote-FVF
Implementation
Fused value
True illuminance
Sensor data
(one color per node)

Comparison of Fusion Algorithms
True illuminance & sensor readings
Mote-FVF algorithm (with median
value majority voting approach)
Mote-FVF algorithm (with Gaussian
correlation majority voting scheme)
Gaussian correlation majority voting
Mote-FVF algorithm without
majority voting

Mote-based Actuation
• Hardware setup
– Four 3V power port on prototyping board as 4 digital
output (16 states defined)
– 4 bit digital-to-analog converter
– Current-to-voltage amplifier for D/A converter
• Challenges
– Low actuation state resolution
– Reliability of actuating command
– Nonlinear behavior of dimmable ballast

Conclusion
• Sensor aspect
– Photoconductors don’t have linear behaviors
– Accelerometers could be fused with other alternative
occupancy sensors
• Sensor validation and fusion aspect
– Mote-FVF algorithm is capable of extracting pertinent
information and rejecting failures
– Simple enough to be embedded in the programmable
motes for intra- and inter-network data fusion
– Works for both redundant and disparate sensor fusion
• Mote actuation aspect
– States of mote actuation could also be fused to boost the
overall efficiency and accuracy

Future Research
• Develop method for distinguishing packet loss
from packet delay and determine optimal time
between iterations of mote-FVF
• Develop self-calibration algorithm
• Implement mote-FVF with more sensor information
(occupancy sensor, actuation state, etc.)
• Distribute mote-FVF algorithm to motes and
implement decentralized sensor network
• Derive statistical model, implement and compare
mote-FVF with model-based approach.

Majority Voting
(Correlation among Sensor Readings)
• Median value approach
– Filter out readings out of the physical limitations of sensor
– Take median of the remaining readings
• Gaussian correlation approach
– Filter out readings out of the physical limitations of sensor
– Generate Gaussian functions centered at each reading
– Take the reading corresponding to the maximum of the
normalized summation of all Gaussian functions

Illuminance-temperature
Interference
illum. (lux) sensor readings mapped temp. true emp. () start time end time
256 start 200.09 24.9 24.4 01:55:50 01:56:15
300 184.75 22.5 24.4 01:56:30 01:57:00
400 167.59 19.6 24.6 01:57:30 01:58:00
500 159.32 18.2 24.8 01:58:50 01:59:20
600 153.49 17.2 25.0 02:00:05 02:00:35
700 147.4 16.2 25.2 02:01:00 02:01:30
800 142.51 15.3 25.4 02:01:55 02:02:25
900 139.61 14.8 25.6 02:03:00 02:03:30
219 217.79 27.6 25.4 02:04:00 02:04:30
900 end 139.7 14.8 25.6 02:04:40 02:05:10
illum. (lux) sensor readings mapped temp. true emp. () start time end time
300 end 192.66 26.7 25.2 02:14:50 02:15:20
261 127.34 15.1 25.2 02:14:10 02:14:40
300 131.76 16.0 25.3 02:13:30 02:14:00
400 135.14 16.7 25.4 02:12:20 02:12:50
500 140.02 17.6 25.5 02:10:35 02:11:05
600 147.91 19.1 25.5 02:09:35 02:10:05
700 156.47 20.7 25.4 02:08:50 02:09:20
800 174.46 23.8 25.4 02:08:10 02:08:35
900 185.79 25.7 25.2 02:07:10 02:07:40
219 start 174.71 23.9 25.1 02:06:25 02:06:55

Mote-based Actuation Architecture
Actuating
command
(state)
Mote
Radio receiver
Prototype board
16 states (levels)
between 0~10V
120V Power
Line
4 bit
digital-to-analog
converter
+5V
Regulator
Transformer
Optional
amplifier
0~10V
-15V
+15V
Regulator
Regulator
Rectifier
Dimmable
ballast