COEXISTENCE OF DIFFERENT WIRELESS SENSOR
NETWORKS
MAC Protocol Interference between X-MAC and Low Power Probing
Sven Zacharias, Thomas Newe, Sinead O’Keeffe and Elfed Lewis
Electronic and Computer Engineering, University of Limerick, Limerick, Ireland
Keywords: Wireless Sensor Network, WSN, Medium Access Control, MAC, Coexistence, Channel Interference,
Competition, Interoperability.
Abstract: Wireless Sensor Networks (WSNs) are an emerging technology that will be widely deployed in the near
future. Most WSNs operate on the 2.4 GHz band of the three free ISM frequency bands. The 2.4 GHz
frequency band is already used by different wireless systems. With an increasing number of WSNs, the
scenario of different WSNs operating on the same IEEE 802.15.4 frequency channel becomes more likely.
WSN Medium Access Control (MAC) Protocols used today were not designed with this problem in mind.
To date, the research focused on interference on the Physical Layer. This work analyses the jamming
potential and the robustness of MAC Protocols, namely X-MAC and Low Power Probing (LPP), at the level
of inter-network competition for medium access, when multiple WSNs are in range of each other operating
on the same channel. The following parameters have been investigated and their effect on interference is
shown: sampling time, channel check rate and payload.
1 INTRODUCTION
A Wireless Sensor Networks (WSNs) consists of
many, theoretically up to thousands of sensor nodes.
A single sensor node, called a mote, is a small and
inexpensive device that is built from the following
main parts: one or more sensors, a data processing
unit, a wireless communication interface, and an
energy source. Today’s most suitable wireless
transfer technologies for WSNs are based on the
Institute of Electrical and Electronics Engineers
(IEEE) 802.15.4 standard (IEEE, 2003), since it
provides a simple, low-power stack for the Physical
and Data Link Layers.
2 PHYSICAL LAYER
The IEEE 802.15.4 standard can physically operate
on the three free Industrial, Scientific and Medical
(ISM) frequency bands offering 27 channels: one at
868 MHz, ten in the 915 MHz band and 16 in the 2.4
GHz band. The only frequency band available
worldwide is at 2.4 GHz, which is the most used
ISM band utilised by many technologies and
therefore the band is crowded (Zhou et al., 2006).
Possible sources of interference in the 2.4 GHz
band can be the common microwave oven and
harmonics of monitors. However, as revealed by a
technical report of the Jennic Cooperation (Jennic,
2008) investigating the effects of different
Figure 1: Channels in the ISM 2.4 GHz base band used by IEEE 802.11b and IEEE 802.15.4. Bold channels are non-
overlapping channels that are normally used. Do not scale spectral mask or output power from this drawing.
197
Zacharias S., Newe T., O’Keeffe S. and Lewis E..
COEXISTENCE OF DIFFERENT WIRELESS SENSOR NETWORKS - MAC Protocol Interference between X-MAC and Low Power Probing.
DOI: 10.5220/0003829601970202
In Proceedings of the 1st International Conference on Sensor Networks (SENSORNETS-2012), pages 197-202
ISBN: 978-989-8565-01-3
Copyright
c
2012 SCITEPRESS (Science and Technology Publications, Lda.)
interference sources, Wireless Local Area Networks
(WLANs) are the main source of interference. The
interference of WLANs based on the IEEE 802.11b
(IEEE, 2007) and the newer IEEE 802.11n (IEEE,
2009) standard on IEEE 802.15.4 have been studied
in detail (Yang et al., 2011); (Petrova et al., 2007);
(Bello and Toscano, 2009).
In order to minimise the risk of interference on
WSNs, a WSN channel outside the band of the used
WLAN channels is normally chosen, as shown for
IEEE 802.11b in Figure 1. But Petrova et al. (2007)
report that even outside of the used WLAN
channels, IEEE 802.11n interferes with WSNs.
Due to WLANs and other external factors, the
choice of non-interfered WSN channels is often
limited to four or less. The default, pre-set channel
in TinyOS and ContikiOS is 26. Hence it is likely
that many WSNs operate on channel 26.
Trends like the Internet of Things and Machine
to Machine communication will lead to many
embedded wireless networks in the near future. So it
becomes more and more likely that there will be a
scenario of two WSNs operating on the same IEEE
802.15.4 channel and thus using the same Physical
Layer (maybe even the same radio chip), but
probably a different Data Link Layer.
3 MEDIUM ACCESS CONTROL
(MAC) SUB-LAYER
Since a network consists of many participants, but
the radio channel can only be used by a single
participant at any time, the task of the MAC Sub-
Layer is to avoid two or more nodes trying to
transmit at the same time (packet collisions).
In classical networks, all transfer modules are
always turned on due to the fact that energy is not
constrained, thus the communication medium can be
monitored at all times. This behaviour leads to the
simplest form of MAC Protocol, called Carrier
Sense Multiple Access (CSMA). The potential
sender listens in order to determine if the channel is
used by another device for transmitting. This check
is called Clear Channel Assessment (CCA). If the
channel is idle, the sender transmits its own
message. If the channel is used, the potential sender
performs a backoff algorithm, which means it waits
for a random time and retries. An extension of this
behaviour often used in WLANs is Carrier Sense
Multiple Access/Collision Avoidance (CSMA/CA)
with handshake. This handshake uses Request To
Send (RTS) and Clear To Send (CTS) messages
before each data transfer. It adds additional overhead
but takes into account that the receiving node can be
blocked by a device outside of the sender’s
communication range (Hidden Terminal Problem).
ContikiOS, which is the operating system used
for the following experiments, includes two network
stacks: uIP and Rime (Dunkels et al., 2007). uIP is a
small TCP/IP stack and Rime is a lightweight
communication stack designed for low power radios
used in WSNs. This work uses the Rime stack. WSN
MAC Protocols have to save energy, which is
realised by turning the radio unit off most of the
time. The channel check rate determines how often
the radio is switched on. In ContikiOS, the term
Radio Duty Cycling (RDC) Layer is used for the
lower part of the MAC Layer that manages the sleep
times of the radio. The MAC Layer Protocol in
ContikiOS can provide retransmissions (CSMA)
when the RDC Layer indicates a collision.
3.1 Radio Duty Cycling (RDC)
A huge number of RDC/MAC Protocols have been
published. Summaries and comparisons of the most
commonly used ones can be found in the literature
(Demirkol et al., 2006); (Kredo II and Mohapatra,
2007); (Roy and Sarma, 2010). A possible taxonomy
for MAC Protocols is:
Unscheduled
o Push, Sender-initiated, e.g. Low Power
Listening (LPL), X-MAC
o Pull, Receiver-initiated, e.g. Low Power
Probing (LPP)
Scheduled
o Time Division Multiple Access (TDMA)
Scheduled approaches have many advantages, but
they are mainly optimised for high traffic and
therefore adapt poorly to network changes and need
synchronisation (Cionca et al., 2008). Hence,
unscheduled protocols are used for dynamic WSNs.
In the following, the two unscheduled approaches
Push and Pull are studied in further detail by means
of X-MAC and LPP. Like most protocols, they have
been designed for stand-alone usage, so that both
coexistence with other protocols and competition for
the medium have not been considered yet.
3.1.1 X-MAC
X-MAC (Buettner et al., 2006) is a pushing protocol
based on the Low Power Listening (LPL) approach
(Moss et al., 2007), where the nodes turn off their
radios for most of the time. If a node is about to
send, it turns on its radio and sends short preambles,
called strobes, until it receives an acknowledgement;
SENSORNETS 2012 - International Conference on Sensor Networks
198
then the message is sent. Non-sending nodes wake
up for a short listening period after the sleep time in
order to monitor the channel for strobes. Due to this
behaviour, the idle listening time is reduced.
3.1.2 Low Power Probing
Low-Power Probing (LPP) (Musaloiu-E. et al.,
2008) can be roughly described as the inverse
approach to X-MAC. Instead of the sender initiating
the communication, the receiver is announcing its
ability to receive messages, basically pulling
messages. In LPP, all nodes are duty cycled and
wake up for just a short time. If a node is awake, it
sends a small packet, called probe, to signal that it is
awake and then it listens for a short time for replies.
A sending node turns its radio on and listens for the
probe of the communication partner.
Additionally, LPP simplifies routing. The data is
pulled hop by hop to the base station instead of
pushed for which the sender needs to know an
address of a node closer to the base station.
Figure 2 shows the principle of both approaches.
3.2 Competition between MAC
Protocols
The interference between different WSNs has been
studied only recently: Bello and Toscano (2009)
show the interference of WSNs operating on
adjacent channels. Bertocco et al. (2008) investigate
interference of two WSNs. They evaluate a ZigBee
WSN interfered by Bluetooth, WLAN and another
ZigBee WSN on the same channel and show how
CSMA/CA affects the Packet Error Rate (PER).
Boano et al. (2010) investigate the factors
influencing the robustness against interference of
WSNs and use a “Semi-Periodic Interferer” to
simulate a WSN on the same channel. Finally, they
present an enhanced, more robust version of X-
MAC. The security aspect of an intended jamming
of a WSN has also been studied in detail (Xu et al.,
2005).
4 EXPERIMENTAL SETUP
All experiments are conducted using TelosB sensor
nodes (MEMSIC, 2010) in a normal office
environment, thus the environment is full of WLAN
signals, laptop/desktop computers and other
electronic and metallic equipment. Channel 26 is
chosen, since this channel is less interfered with by
WLANs, as previously mentioned. To be sure that
missed packets are not due to external interference,
interfered and non-interfered trials are done in
alternating order. Also the received signal strength
indicator (RSSI) is logged for every packet to ensure
a good connection. Additionally, to rule out external
factors and gain better insight into the causes for the
results, all experiments have been simulated in
Cooja. The “Unit Disk Graph Medium” and a
random start up time of maximal 3 s (which is
comparable to the start of nodes by hand) have been
used. Due to the limited space and the fact that the
simulation corresponds to the experiments, only the
experimental results are presented here.
The software for the experiments is written in
ContikiOS version 2.4. X-MAC is used in two
versions, with and without CSMA. CSMA supports
a four packet big buffer and tries a single
retransmission. LPP provides a packet buffer size of
four packets. In the following setup, the observed
network will be called collector and the interfering
network will be called jammer.
Figure 2: (a) In X-MAC, the sender actively tries to establish the communication by sending strobes. (b) In LPP, the
receiver is pulling data with the help of probes. (c-f) The standard timing of X-MAC and LPP with a channel check rate of 4
Hz: (c) X-MAC listens for 6 ms at the beginning of each cycle and then sleeps for 244 ms. (d) If X-MAC wants to transmit,
it sends strobes until it receives an acknowledgement or reaches 50 reiterations. (e) LPP sends a probe every 250 ms waits
for incoming data and then sleeps. (f) A potential sender wakes up and listens for an incoming probe for a maximum of 1s.
COEXISTENCE OF DIFFERENT WIRELESS SENSOR NETWORKS - MAC Protocol Interference between X-MAC and
Low Power Probing
199
The senders of the collector network send a
unicast message to the base station every five
seconds, 100 messages in total. The packet has a
payload of 4 bytes consisting of an unsigned integer
counter and a timestamp. Additionally, data from
lower layers are added, and thus the packet has a
total size of 36 bytes for X-MAC and 21 bytes for
LPP. This is considerably less than the 127 bytes
allowed for a single IEEE 802.15.4 data frame,
hence every message is sent in a single frame. The
message is only slightly longer than an X-MAC
strobe of 28 bytes or a LPP probe of 13 bytes. This
overhead is realistic for a simple temperature
measurement system for example. The default
values were used for the on- and off-times of the
radios, the number of retries and all other
parameters. Thus, both protocols have comparable
energy consumptions while being idle. The channel
is checked at a rate of 4 Hz, thus cycles of 250 ms
occur, the resulting standard timing is illustrated in
Figure 2 (c-f).
The nodes were placed close (< 20 cm) to the
base station on a desk, in direct line of sight and thus
the distance or orientation of the nodes is not
affecting the transfer. The collector network builds a
single-hop star topology and the jammer network
consists of a sender and a base station. All nodes are
in communication range of each other.
The base stations in these experiments are duty
cycled as well, since they would just be a hop closer
to the real base station in a multi-hop network. They
do not send data messages. All nodes are started
manually, thus the start times vary slightly. Every
trial is conducted ten times in order to provide a
reliable data set.
4.1 Network Scalability
The scalability of a non-interfered collector network
is tested in order to provide a comparison between
high traffic and interference.
Table 1: Packet Error Rates (PERs) [%] of the scalability
experiment.
Scalability (sender)
3 5
X-MAC
mean
3.93 50.58
median
0.67 49.30
stdev
10.14 5.52
X-MAC/
CSMA
mean
3.47 15.44
median
0.17 17.00
stdev
5.36 5.90
LPP
mean
8.20 11.38
median
0.17 7.70
stdev
16.46 12.20
The resulting mean, median and standard
deviation of the PERs over ten repeats for this
experiment are shown in Table 1. Even with only
three sending nodes, some packets are lost due to
collisions. The PERs of all protocols increase due to
the traffic of two additional senders. The lost packets
almost always belong to a single node. If a deadlock
between nodes occurs due to bad timing then it stays
unresolved until the end of the experiment due to the
fixed duty cycling. This results in a constant, high
packet loss for a single node.
4.2 Interference Experiments
For all of the following interference experiments,
this work uses a collector network (consisting of a
base station and three sending nodes) and a jammer
network (consisting of a sender and a base station).
Experiment 1) Scalability versus Interference: This
experiment compares the effects of having two
additional senders in a network (Table 1) with two
interfering nodes using a different protocol. X-MAC
and X-MAC/CSMA are interfered by LPP and LPP
is interfered by X-MAC. The parameters used
(sampling time, channel check rate and payload) are
exactly the same for the collector and jammer
network. The effect of the interference, shown in
Table 2, is an increased PER that is lower than the
PER of the scaled up networks for X-MAC and LPP
and roughly the same for X-MAC/CSMA. The X-
MAC collector network is losing almost all packets
or packets at regular intervals from a sender. LPP is
almost unaffected.
Experiment 2) Decreased Sampling Time: To
increase the effect of interference the number of sent
packets on the jammer network is raised by setting
the sampling interval to 1 s. The packets contain the
same payload as the packets of the collector
network. Both networks use the same channel check
rate. As shown in Table 2, this only leads to an
increased PER for X-MAC.
Experiment 3) Decreased Channel Check Rate: To
investigate the effect of the channel check rate, the
sampling rate is set to 2 Hz for the jammer network.
This results in fewer data transfer contacts, but
increases the number of strobes/probes needed to
synchronise (the channel is checked twice in a
second and each second, a packet is sent). The PER
is not changing considerably (see Table 2), which is
due to the fact that the actual message is similar in
length compared to the strobes/probes.
Experiment 4) Increased Payload: The payload sent
by the jammer network is enlarged to see the effect
of increased packet length. A longer packet results in
SENSORNETS 2012 - International Conference on Sensor Networks
200
a longer continuous blocking of the channel. The
payload is increased to 42 bytes. The X-MAC and
X-MAC/CSMA collector is interfered by LPP with a
59 byte packet and LPP by X-MAC with a 74 byte
packet. The effect on the PER does not considerably
differ from Experiment 2, only X-MAC/CSMA
shows a higher PER, as shown in Table 2. This
shows that the payload and thereby the packet length
has a small effect due to the overhead of the strobes
and probes. X-MAC can send up to a maximum of
50 strobes to establish the data exchange. LPP
always sends a probe at each duty cycle. Compared
to these strobes/probes, the data packet length is
almost negligible.
Experiment 5) Pobes versus Data Packets: To show
that the overhead traffic generated by probes is more
relevant than the data packets, the sampling time is
set up to 25 s. Despite the sampling time change,
LPP still generates the same amount of probes. The
resulting PER, shown in Table 2, is still as high as in
Experiment 2 and Experiment 4 while the sent
application data decreased by the factor of five. This
shows that the overhead is more important than the
actual data sent.
Experiment 6) LPP jammed by X-MAC/CSMA: In all
the experiments described so far, LPP is jammed by
X-MAC without CSMA. In this experiment, LPP is
jammed by X-MAC/CSMA. The retransmissions
caused by CSMA generate additional traffic on this
jammer network. As shown in Table 2, this
additional traffic slightly increases the PER.
4.3 Discussion
The PER is not equally distributed among the
repetitions of the experiments. In the results of the
single trials three different cases can be identified:
full loss of all 100 packets of a single sender,
packets loss occurring in regular intervals, or no
packet loss. These patterns match with patterns
shown in Cooja simulations, so it can be asserted
that the packet errors are not caused by external
interference. Since theses packet losses are due to
collisions between the jammer and collector
network, timing is a vital key factor.
For X-MAC and X-MAC/CSMA, the
interference has roughly the same effect as the
increased traffic of more communication partners
using the same MAC Protocol.
LPP seems not to be affected by the jammers.
This result matches with Boano et al. (2010), since
the latter shows that a “Packet Queue with Fast
Drain” helps to avoid interference. LPP is using a
similar concept by default, since it has an included
buffer, which is transmitted in a row when the
handshake succeeds. But LPP seems to fail in
scaling to more nodes and jams other WSNs with its
permanent probing. Since the major traffic is
generated by the overhead in the network, the actual
data sent has a small effect on the PERs.
5 CONCLUSIONS
To date, robustness against interference with other
technologies has been widely investigated, but the
interference between WSNs operating on the same
channel has not been focused on to date. In WSN
MAC Protocol design, the assumption of an isolated
WSN is still dominant. This work gives a first
attempt to direct attention to the problem of inter-
WSN interference.
The parameters: sampling time, channel check rate
and payload of the jamming network have been
experimentally investigated. For the setup being
used, their effects on the PER are surprisingly small,
since most of the traffic is overhead. The results
show that the packet errors caused by a second
network using a different protocol are roughly the
same size as the packet errors caused by traffic of
additional nodes on the same network.
In this work, only the parameters of the
Table 2: Packet Error Rates (PERs) [%] of the interference experiments.
Jammer network (sampling interval, channel check rate)
5 s, 4 Hz 1 s, 4 Hz 1 s, 2 Hz
1 s, 4 Hz,
payload
25 s, 4 Hz,
probes
1 s, 4 Hz
CSMA
Experiment number 1 2 3 4 5 6
X-MAC
mean
38.77 53.10 40.23 50.67 50.13 -
median
36.83 47.33 45.67 44.00 53.67 -
stdev
20.55 26.47 14.30 26.09 16.03
X-MAC/
CSMA
mean
18.10 19.17 17.13 28.80 21.37 -
median
15.17 15.00 10.83 20.83 22.33 -
stdev
15.26 16.99 15.36 23.74 7.70
LPP
mean
0.17 0.23 6.63 4.50 - 5.87
median
0.17 0.17 0.00 0.33 - 0.33
stdev
0.17 0.26 19.46 12.83 - 16.39
COEXISTENCE OF DIFFERENT WIRELESS SENSOR NETWORKS - MAC Protocol Interference between X-MAC and
Low Power Probing
201
interfering network have been changed. Valuable
insight into the problem could be provided in future
studies by investigating the parameters of the
collector network and their effect on its vulnerability
to interference. A larger setup including multi-
hopping would exacerbate the interference problem.
The conduction of the experiments in a shielded
environment would exclude external interference
and thereby deliver less noisy measurements.
A modification of the MAC Protocol duty
cycling supporting variable timings could avoid the
deadlock. The authors believe that the problem of
inter-WSN interference will gain more research
interest in the near future.
ACKNOWLEDGEMENTS
The authors wish to thank the following for their
financial support: The Embark Initiative and Intel,
who fund this research through the Irish Research
Council for Science, Engineering and Technology
(IRCSET) postgraduate Research Scholarship
Scheme. Thanks to the anonymous reviewers for
their valuable comments and suggestions.
REFERENCES
Bello, L. L. and Toscano, E., 2009. Coexistence issues of
multiple co-located IEEE 802.15.4/zigbee networks
running on adjacent radio channels in industrial
environments. Industrial Informatics, IEEE
Transactions on, 5(2):157 –167.
Bertocco, M., Gamba, G., and Sona, A., 2008. Is
CSMA/CA really efficient against interference in a
wireless control system? An experimental answer. In
Emerging Technologies and Factory Automation,
2008. ETFA 2008. IEEE International Conference on,
pages 885–892. IEEE.
Boano, C. A., Voigt, T., Tsiftes, N., Mottola, L., Römer,
K., and Zuniga, M. A., 2010. Making sensornet mac
protocols robust against interference. In 7th European
Conference on Wireless Sensor Networks, Coimbra,
Portugal.
Buettner, M., Yee, G., Anderson, E., and Han, R., 2006.
X-MAC: A short preamble MAC protocol for duty-
cycled wireless sensor networks. In Proceedings of the
4th International Conference on Embedded Networked
Sensor Systems, Boulder, Colorado, USA, pages 307–
320.
Cionca, V., Newe, T., and Dadarlat, V., 2008. TDMA
protocol requirements for wireless sensor networks. In
Sensor Technologies and Applications, 2008.
SENSORCOMM ’08. Second International Conference
on, pages 30 –35.
Demirkol, I., Ersoy, C., and Alagoz, F., 2006. MAC
protocols for wireless sensor networks: a survey.
Communications Magazine, IEEE, 44(4):115 – 121.
Dunkels, A., Österlind, F., and He, Z., 2007. An adaptive
communication architecture for wireless sensor
networks. In Proceedings of the Fifth ACM
Conference on Networked Embedded Sensor Systems
(SenSys 2007), Sydney, Australia
Kredo II, K. and Mohapatra, P., 2007. Medium access
control in wireless sensor networks. Computer
Networks, 51(4):961–994.
IEEE Computer Society. 2003. IEEE Standard
802.15.4™-2003 - Part 15.4: Wireless Medium Access
Control (MAC) and Physical Layer (PHY)
Specifications for Low-Rate Wireless Personal Area
Networks (LR-WPANs)
IEEE Computer Society. 2007. IEEE Standard 802.11™-
2007 - Part 11: Wireless LAN Medium Access Control
(MAC) and Physical Layer (PHY) Specifications
IEEE Computer Society. 2009. IEEE Standard 802.11n™-
2009 – Part 11: Wireless LAN Medium Access Control
(MAC) and Physical Layer (PHY) Specifications,
Amendment 5: Enhancements for Higher Throughput
Jennic Cooperation, 2008. Co-existence of IEEE 802.15.4
at 2.4 ghz. Application Note, Revision 1.0.
MEMSIC Inc., 2010. TelosB Mote Platform Datassheet.
Moss, D., Hui, J., and Klues, K., 2007. Tep 105-low
power listening. http://www.tinyos.net/tinyos-
2.x/doc/pdf/tep105.pdf - Retrieved Sept. 2011.
Musaloiu-E., R., Liang, C.-J. M., and Terzis, A., 2008.
Koala: Ultra-low power data retrieval in wireless
sensor networks. In Proceedings of the 7th
International Conference on Information Processing
in Sensor Networks, IPSN ’08, pages 421–432,
Washington, DC, USA. IEEE Computer Society.
Petrova, M., Wu, L., Mahonen, P. and Riihijärvi, J., 2007.
Interference measurements on performance
degradation between colocated IEEE 802.11 g/n and
IEEE 802.15.4 networks. In Networking, 2007.
ICN’07. Sixth International Conference on, pages 93–
93. IEEE.
Roy, A. and Sarma, N., 2010. Energy saving in MAC
layer of wireless sensor networks: a survey. In
National Workshop in Design and Analysis of
Algorithm (NWDAA), India. Tezpur University.
Xu, W., Trappe, W., Zhang, Y., and Wood, T., 2005. The
feasibility of launching and detecting jamming attacks
in wireless networks. In Proceedings of the 6th ACM
International Symposium on Mobile ad hoc
Networking and Computing, MobiHoc ’05, pages 46–
57, New York, NY, USA. ACM.
Yang, D., Xu, Y., and Gidlund, M., 2011. Wireless
coexistence between IEEE 802.11- and IEEE
802.15.4-based networks: A survey. International
Journal of Distributed Sensor Networks, Article ID
912152:17.
Zhou, G., Stankovic, J. A., and Son, S. H., 2006. Crowded
spectrum in wireless sensor networks. In Proceedings
of Third Workshop on Embedded Networked Sensors
(EmNets).
SENSORNETS 2012 - International Conference on Sensor Networks
202