minimal transmission power as distance estimation for ... · minimal transmission power as distance...
TRANSCRIPT
Minimal Transmission Power asDistance Estimation
for Precise Localization in Sensor Networks
Jan Blumenthal, Dirk TimmermannUniversity of Rostock
IWCMC, Vancouver2006/07/06
2
Focus of Presentation
• Preconditions– Hundreds of sensor nodes are
randomly deployed– Position initially unknown
• Why do we need localization?– Measurement requires position– Example: Self organization, self
healing, geographic routing
• Problem Statement– Localization needs distance
information– How to measure distances ?
Flood prevention by dikeobservation
3
Distance Estimation with RSSI in Theory
• Energy of signal decreases with distance d
• Sensor node measures energy of received signal
• Compared to a reference voltage• Received Signal Strength
supported by hardware– Cheap and always available
20
4R
R SS
P G GP d
λπ
⎛ ⎞= ⎜ ⎟⎝ ⎠
[ ]2
0( ) [ ] 10 log 20 log( )4R S R SP d dBm P dBm G G dλπ
⎡ ⎤⎛ ⎞= + ⋅ − ⋅⎢ ⎥⎜ ⎟⎝ ⎠⎢ ⎥⎣ ⎦
distance d [m]R
ecei
ved
Sign
al S
tren
gth
[dB
m]
Measured RSS1
d1
Power Relations:
PTX Transmission PowerPRX Received PowerGTX Gain at SenderGRX Gain at Receiverd Distanceλ0 Wave length
20
4RX
RX TXTX
P G GP d
λπ
⎛ ⎞= ⎜ ⎟⎝ ⎠
4
Distance Estimation with RSSI in Theory II
20
4RX
RX TXTX
P G GP d
λπ
⎛ ⎞= ⎜ ⎟⎝ ⎠
0
4RX TX TX
RX
TX
RX
G G PdP
PkP
λπ
=
=
20
4R
R SS
P G GP d
λπ
⎛ ⎞= ⎜ ⎟⎝ ⎠
distance d [m]R
ecei
ved
Sign
al S
tren
gth
[dB
m]
Measured RSS1
d1
PTX Transmission PowerPRX Received PowerGTX Gain at SenderGRX Gain at Receiverd Distanceλ0 Wave length
Friis‘ Equation:
max TXd P∼
Rearrange
~
5
RSSI on Chipcon CC1010
- graph not stable- high variance
0
50
100
150
200
250
0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60 64 68 72Distance [m]
Rec
eive
d Si
gnal
Str
engt
h
6
RSSI in Theory and in Reality
Transmitter Receiver
RSSI
Message Message
RSSI in Theory
0
50
100
150
200
250
0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60 64 68 72
RSSI in Reality
Result: In Reality, distances based on RSSI are inaccurate.
Y Y
dMax. Transmission
Power
Approach: Minimal Transmission Power
8
Approach
Assumption:– Max. transmission power– Measuring RSSI is inaccurate
caused by• Measuring principle• Hardware effort
Approach:– Stepwise increasing transmission
power PTX
– In case of reception of a message the transmission power PTXspecifies a distance d
– Utilize only smallest transmission power Distance d
• Moving graph• Find crossing
with x-axisPTXMax
PTX
dMax d
• Measuring PRX• Recalculation of d
PRX
Distance d ddMax
PTXPTXMax
PTX_1
PTX_2
9
Finding the Distance
• Transmission power PTX is controlled via Special Function Register SFRTX
• SFRTX of tranceiver is transmitted with every message
• Distance d is calculated out of smallest SFRTX
Example (Scatterweb)• SFRTX tunable in range 0..100 (300m)• SFRTX(16±4) → d=2m
SFRTX=11
2m
SFRTX=14 SFRTX=16
Transmitting node (Beacon)
Receiving sensor node
10
Relation between Transceiver and SFRTX
Approximated transfer function H(x) of tranceiver (npn transistor)
Message
Transmission Power
YTransmitter
PTX
SFRTX
4TX
SFRkPTX ⋅=≈
Special Function Register to control transmitter
11
Relation between Distance and SFRTXPTX = Transmission PowerSFRTX = Transmission RegisterPRX = Received Powerd = distance
SFRTX
d
Already known:
Transmitter Y
Adjust Transmission Power via Special Function Register SFRTX
SFRTX
MessagePTX
max TXd P∼~
4TX
SFRkPTX ⋅=≈
2
2
4
TX
RXTX
SFRd
PkSFRd
=
⎟⎠⎞
⎜⎝⎛=πλ
≈
dSFRTX =≈
12
Min. Transmission Power
Message Message
• Distance d can be easly calculated from transmitted SFRTX
• Smaller estimation error compared to RSSI
Theory
d
Min. SFRTXTransmission Power
SFRTX
Y YTransmitter Receiver
d
Min. SFRTX
Reality
PTX 4TX
SFRkPTX ⋅=≈
13
Measurement Results: Scatterweb
0
5
10
15
20
25
30
35VarianzMinVarianzMaxMittelwert
Min. SFRTX of sensor nodes based on Scatterweb (indoor, ideal conditions, 40 samples each)
0 50 400100 150 200 250 300 350
Min. VarianceMax. VarianceAverage
We know, in theory: dSFRTX =≈
Distance dij between sender and receiver [cm]
Min
. Tra
nsm
issi
on P
ower
[SFR
TX]
14
Linearized Measurements
0
100
200
300
400
500
600
700
800
900
5 30 55 80 105 130 155 180 205 230 255 280 305 330 355 380
Min
. Tra
nsm
issi
on P
ower
[SFR
TX²]
Distance dij between sender and receiver [cm]
Linear Regressionf(x)=mx+n• m=1.3• n=0
• Create linear equation• Linearize graph by squaring d• Determine raising of graph• Combine final equation
Theory:
dmSFR
dSFR
TX
TX
⋅=
≈2
≈
15
Measurement Results: Scatterweb
0
5
10
15
20
25
30
5 25 45 65 85 105 125 145 165 185 205 225 245 265 285 305 325 345 365 385
Distance d ij between sender and receiver [cm]
Min
. tra
nsm
issi
on p
ower
[SFR
TX]
Average min. transmission power
Average min. transmission power plus standard deviation
Average min. transmission power minus standard deviation
Regression 1.3TX ijSFR d= ⋅
dSFRTX ⋅= 3.12
3.1
2TXSFRd =dSFRTX ⋅= 3.1
Example Application
17
Example Application with Scatterweb
Localization Algorithm:Weighted Centroid Localization (WCL)
– simple and fast calculation– small memory footprint– acceptable positioning error
Jan Blumenthal, Frank Reichenbach, Dirk Timmermann: Precise Positioning with a Low Complexity Algorithm in Ad hoc Wireless Sensor Networks, PIK - Praxis der Informationsverarbeitung und Kommunikation, Vol.28 (2005), Journal-Edition No. 2, S.80-85, ISBN: 3-598-01252-7, Saur Verlag, Deutschland, June 2005
Jan Blumenthal, Frank Reichenbach, Dirk Timmermann: Position Estimation in Ad hoc Wireless Sensor Networks with Low Complexity (Slides), Joint 2nd Workshop on Positioning, Navigation and Communication 2005 (WPNC 05) & 1st Ultra-Wideband Expert Talk 2005 (05), S.41-49, ISBN: 3-8322-3746-1, Hannover, Deutschland, March 2005
References:
Task:Determine position of moving sensor node
Immobile beacons
Mobile sensor node
Sensor network
18
Weighted Centroid Localization (WCL)
Approach:- Positioning by centroid determination
Pi´ (CGLCD)- Improved precision by weighting
measured distance dix using wij()
( )1
1
( , )''( , )
b
ij jj
i b
ijj
w B x yP x y
w
=
=
⎛ ⎞⋅⎜ ⎟
⎝ ⎠=⎛ ⎞⎜ ⎟⎝ ⎠
∑
∑ wij = Weight of distance Bj and node ib = Number of beaconsBj(x,y)= Position of beacons j
1
1'( , ) ( , )n
i jj
P x y B x yn =
= ∑
WCL
CGLCD ''( , )iP x y'( , )iP x y
1B 2B
3B4B
4id3id
2id1id
( , )iP x y
Immobile beacons
Mobile Sensor Node
19
Weight-based on Distances
( )1
ij g
ij
wd
=
Weight wij() depends on measured distance between beacon and sensor node.
Definition of Weight:
How to determine dij?
( )1
1
( , )''( , )
b
ij jj
i b
ijj
w B x yP x y
w
=
=
⎛ ⎞⋅⎜ ⎟
⎝ ⎠=⎛ ⎞⎜ ⎟⎝ ⎠
∑
∑
Optimal: g=3
20
Operations on Beacons
Max. transmission power sent ?
Increase round number
Send message with own position, transmission power and round number
Increase transmission power and wait one time slot
no
yes
Adjust transmitter with current transmission power
Reset transmission power
• Beacons are sensor nodes with already known position• Send out messages with own position and currently adjusted SFRTX
21
Operations on Sensor Nodes
Higher round number ?
New min. trans-mission power ?
Store beacons’ position, round number and transmission powerReject message
no
yes
no
yes
Waiting for messages
New beacon position received ?yes
no
Beacon message received ?
yes
no
• Sensor nodes do not know own position
• Calculate distance and positions out of received messages
22
Reduce high Oscillating Measurements
Control Loop
Calculate averaged min. transmissionpower e.g.,
PTX aPTX
0.25 0.75TX TX TXaP aP P= ⋅
Higher round number ?
New min. trans-mission power ?
Store position, round number and transmission powerReject message
no
yes
no
yes
Waiting for messages
New beacon position received ?yes
no
Beacon message received ?
yes
no
23
Demonstrator Show• 4 beacons at all corners, one unknown node was moved• Localization error: 0.3m• Notation:
– P12 : current transmission power is 12– aP10: averaged transmission power is 10
Scatterweb
SpyGlass (University of Luebeck) Field size = 2x2 m
http://rtl.e-technik.uni-rostock.de/~bj/movies/PositionEstimationUsingMinimalTransmissionPower.mpg
24
Discussion
Remarks• based on circular transmission range• concurrent channel access through hidden terminal
problem• numerous sources of interferences (sensor nodes,
steel girders, obstacles)• noticeable delays caused
– increasing transmission power– round counting
Advantages of min. Transmission Power• simple distance estimation• easy to implement• more accurate than RSSI
25
Conclusion
Essentials • Wireless sensor networks require localization of
sensor nodes• Most distance estimations are inaccurate especially in
indoor use
New approach to estimate a distance• Minimal transmission power
Proof of concept• Higher resolution and smaller variances than RSSI• Example application combined with Weighted
Centroid Localization (WCL)