Title: Experimental%20Study%20of%20Concurrent%20Transmission%20in%20Wireless%20Sensor%20Networks
1Experimental Study of Concurrent Transmission in
Wireless Sensor Networks
- Dongjin Son, Bhaskar Krishnamachari (USC/EE),
- and John Heidemann (USC/ISI)
2Motivation
- Prior work
- Understanding wireless propagation essentials
- Zhao, Ganesan, Aguayo, Cerpa, Woo, Lal, Zuniga,
Son, etc. - Only few consider concurrent packet transmission
- Whitehouse, Jamieson, Kochut
- Concurrent transmission is endemic in dense
networks - Applications
- Event detection and target tracking
- Code distribution and flooding for route discovery
3Research goals
- Understanding concurrent packet transmissions
! - Systematic experimental study
- Single and multiple interferers
- Develop a better interference model
-
4Main findings
- Single Interferer effects
- Capture effect is significant
- SINR threshold varies due to hardware
- SINR threshold does not vary with location
- SINR threshold varies with measured RSS
- Groups of radios show 6 dB gray region
- New SINR threshold (simulation) model
- Multiple interferer effects
- Measured interference is not additive
- Measured interference shows high variance
- SINR threshold increases with more interferers
5Main findings
- Single Interferer effects
- Capture effect is significant
- SINR threshold varies due to hardware
- SINR threshold does not vary with location
- SINR threshold varies with measured RSS
- Groups of radios show 6 dB gray region
- New SINR threshold (simulation) model
- Multiple interferer effects
- Measured interference is not additive
- Measured interference shows high variance
- SINR threshold increases with more interferers
6Part I Single interferer
- Main research questions
- Does concurrent transmission imply a collision ?
- Can we identify a constant SINR threshold
- (SINR?) for capture?
- Experiments
- Two concurrent senders
- varying transmitter hardware and power
7Methodology
Mica2
PC104
Sender1 (SRC1)
Receiver
Sender2 (SRC2)
Synchronizer (Sync)
Time
Sync
Synchronizes the clocks of both senders
8Methodology
Sender1 (SRC1)
Receiver
Sender2 (SRC2)
Synchronizer (Sync)
Time
Sync
SRC1
Measure an ambient Noise (N)
Measure the RSS of Sender1 (S1)
9Methodology
Sender1 (SRC1)
Receiver
Sender2 (SRC2)
Synchronizer (Sync)
Time
Sync
SRC1
Sync
SRC2
Measure the ambient Noise (N)
Measure the RSS of Sender2 (S2)
10Methodology
Sender1 (SRC1)
Receiver
Sender2 (SRC2)
Synchronizer (Sync)
Time
Sync
SRC1
Sync
SRC2
Sync
SRC1
SRC2
Test the delivery of the senders packet
under the CTX
11Methodology
vary Tx power, hardware, location
Sender1 (SRC1)
Receiver
- Stronger packet ? Signal
- Weaker packet ? Interference
Sender2 (SRC2)
Synchronizer (Sync)
Time
Sync
SRC1
Sync
SRC2
Sync
SRC1
SRC2
Test the delivery of the senders packet
under the CTX
epoch
Repeat this epoch and measure PRR
12Power and PRR based regions
White gt 90 PRR
Black lt 10 PRR
Gray 1090 PRR
- Black-Gray-White due to power change
- Prior work (Zhao, woo etc) use a distance based
definition - SINR threshold (SINR?)
- SINR (Signal-to-interference-plus-noise) value
which ensures reliable packet reception
13Capture effect
White
Gray
Black
Gray
White
- Finding Capture effect is significant SINR?
is not constant - Concurrent packet transmission does not always
means packet collision (capture effect recently
studied by Whitehouse et al.) - Systematically study capture effects and quantify
the SINR? value
14Modeling SINR to PRR relationship
- Regression model for simple description of
experimental data
f frame size of the packet in bytes l preamble
size in bytes - Model based on the link layer
model by Zuniga and Krishnamachari
? ß0 changes the shape (ß0 is set to 2.6
based on the empirical data) ? ß1 changes
the location
ß01
ß02
ß03
-1 0 1 2
ß0,ß1
ß1
ß0,ß1
ß0,ß1
ß0,ß1
ß0,ß1
ß0,ß1
15Transmitter hardware effect
- How much SINR threshold change does transmitter
hardware can make ? - Does hardware variation dominate other effects?
- E.g., compared to the location effect
- Experiments
- Hold location constant
- Swap one of the transmitter hardware
16Does transmitter hardware affect SINR? ?
- Vary transmitter hardware (SRC1-SRC2, SRC1-SRC3)
- while keeping the same receiver
SRC1 (with SRC2)
SRC2 (with SRC1)
SRC1 (with SRC3)
SRC3 (with SRC1)
-1.7 dB
1 dB
5.3 dB
3.4 dB
Finding SINR? changes with different
transmitter hardware
17Signal strength effect
- Is SINR threshold constant at different
- signal (or interference) strength level?
- I.e., Can we always identify a constant SINR
threshold for the same hardware pair ? - Experiments
- Hold location and use the same transmitter pair
- Vary transmission power of both transmitters
18Does signal strength level affect SINR??
- Same transmitter hardware, but vary both sender
and interferers transmission power levels.
Finding SINR? changes at different signal
strength levels
19Implications of findings
(4.6 dB)
Signal strength Hardware
Hardware
(dB)
Signal strength
- Protocols based on constant SINR threshold
assumption will fail - Power control protocol and capture-aware protocol
- should consider variable SINR?
- New interference model is necessary
20Part II Multiple interferers
- Main research questions
- Textbook says Interference is additive,
- How about the reality with low-power RF
transceiver ? - Experiments
- Empirically test the additive signal strength
assumption - Varying the number of interferers and Tx power
21Methodology
Mica2
PC104
Sender
Receiver
Interferer1 (IFR1)
Interferern (IFRn)
Synchronizer (Sync)
Time
Sync
Sender
Measure an ambient Noise (N)
Measure the RSS of Sender (S)
22Methodology
Sender
Receiver
Interferer1 (IFR1)
Interferern (IFRn)
Synchronizer (Sync)
Time
Sync
Sender
Sync
IFR1
Measure an ambient Noise (N)
Measure the RSS of Interferer1 (I1)
23Methodology
Sender
Receiver
Interferer1 (IFR1)
Interferern (IFRn)
Synchronizer (Sync)
Time
Sync
Sender
Sync
IFR1
Sync
IFRn
Measure the ambient Noise (N)
Measure the RSS of Interferern (In)
24Methodology
Sender
Receiver
Interferer1 (IFR1)
Interferern (IFRn)
Synchronizer (Sync)
Time
Sync
Sender
Sync
IFR1
Sync
IFRn
Sync
IFR1
IFRn
Measure the Joint Interference
25Methodology
Sender
Receiver
Interferer1 (IFR1)
Interferern (IFRn)
Synchronizer (Sync)
Time
Sync
Sender
Sync
IFR1
Sync
IFRn
Sync
IFR1
Sync
Sender
IFR1
IFRn
IFRn
Test the delivery of the senders packet
26Joint interference (JRIS) estimators
Time
IFR1
IFR1
Jointly Measured
Independently Measured
IFR2
IFR2
IFR3
IFR3
RIS
RIS
RIS
IFR1
IFR2
IFR3
measured
JRIS(m)
Average of the actual joint interference
measurements
expected
JRIS(e)
Summation of independent interference measurement
Direct measurement!
Textbook prediction!
27Does joint interference show additivity?
RIS (dBm)
Individual RIS of IFR1 and IFR2 (dBm)
Comparison between JRIS(e) and JRIS(m) when two
interferers (IFR1 and IFR2) have equivalent RISs
at the receiver
- Finding Measured interference is not additive
- JRIS(e) is higher than JRIS(m)
- Additive behavior is different at different
signal strength levels
28Joint Interference and SINR?
4 Interferer
3 Interferer
2 Interferer
-64.1 dBm
-68.8 dBm
-64.1 dBm
-68.8 dBm
1 Interferer
-73 dBm
-73 dBm
JRIS(e)
JRIS(m)
SINR threshold measurements with different number
of interferers
- Finding SINR threshold increases with more
interferers - SINR threshold changes with different number of
interferers which changes the joint received
interference strength
29Potential of capture-aware MAC
- Compare the number of CTXable (Concurrently
Transmittable) links - Methodology
- Trace-based Simulation
- Uses real measured RSS
- Without Tx power control
- Assume red link Tx, who can CTX together?
- Observation
- More available links for the capture-aware medium
access
RTS/CTS based
CTXable links with RTS/CTS based MAC
Capture-aware
CTXable links with capture-aware MAC
30Generalized for all links in the testbed
- The number of CTXable links comparison between
traditional and capture-aware MAC
Capture-aware
Capture-unaware
Finding Capture-aware MAC shows about 3 times
more CTXable links on average
31Conclusion
- Experimental results show
- the significance of capture effects as Tx power
varies - some of the theoretical assumption does not hold
for the measurements - (1) SINR threshold varies (not constant)
- (2) Multiple interference worse than addition
(not additive) - better understanding of single and multiple
interference on - packet delivery
- Experimental results imply
- need better SINR threshold simulation models
- more efficient use of wireless channel is
possible with better understanding of concurrent
packet transmission - E.g.,) Capture-aware medium access protocol
USC ANRG http//ceng.usc.edu/anrg I-LENSE
http//www.isi.edu/ilense