TIC-STH 2009

Cognitive Wireless Sensor Networks:

Emerging Topics and Recent Challenges
Amir Sepasi Zahmati, Sattar Hussain, Xavier Fernando Ali Grami
Department of Electrical and Computer Engineering Faculty of Engineering and Applied Science
Ryerson University, Canada University of Ontario Institute of Technology, Canada
Email: {asepasi, s34hussa, fernando}@ee.ryerson.ca Email: [email protected]

Abstract—Adding cognition to the existing Wireless Sensor to the spectrum allocation and utilization concept. A Cognitive
Networks (WSNs), or using numerous tiny sensors, similar to the Radio (CR) is an intelligent wireless communication system
idea presented in WSNs, in a Cognitive Radio Network (CRN) that is aware of its surrounding environment, and adapts
bring about many benefits. In this paper, we present an overview
of Cognitive Wireless Sensor Networks (CWSNs), and discuss the its internal parameters to achieve a reliable and efficient
emerging topics and recent challenges in the area. We discuss communication [9-10]. Through CR technology, unlicensed
the main advantages, and suggest possible remedies to overcome (secondary) users periodically monitor the spectrum for vacant
the challenges. CWSNs enable current WSNs to overcome the channels, and use the channels even though they are originally
scarcity problem of spectrum which is shared with many other owned by the licensed (primary) users. The first organization
successful systems such as Wi-Fi and Bluetooth. It has been
shown that the coexistence of such networks can significantly in the US which developed new technologies to allow multiple
degrade a WSN’s performance. In addition, cognitive technology radio systems to share the spectrum through adaptive mech-
could provide access not only to new spectrum, but also to anisms in the form of the neXt Generation communications
spectrum with better propagation characteristics. Moreover, by (XG) program was the Defense Advanced Research Projects
the adaptive change of system parameters such as modulation Agency. The US Army has also been researching the so-
type and constellation size, different data rates can be achieved
which in turn can directly influence the power consumption called Adaptive Spectrum Exploitation for real-time spectrum
and the network lifetime. Furthermore, sensor measurements management in the battlefield [4].
obtained within the network can provide the needed diversity In this study, we present an overview of Cognitive Wireless
to cope with spectrum fading at the physical layer. Sensor Networks (CWSNs), and discuss the emerging topics
Index Terms—Wireless Sensor Networks (WSN), Cognitive and recent challenges in the area. First, in Section II we
Radio, Spectrum Sensing.
explore how a CWSN works, and study the specific charac-
teristics of a CWSN. Then, we discuss CWSN’s advantages
I. I NTRODUCTION compared to the current WSNs in Section III. Section IV
Rapid advances in processing capability, memory capacity, explains CWSNs design challenges, and presents possible
and radio technology have enabled the development of dis- remedies to overcome the challenges. Finally, in Section V,
tributed networks with small, and inexpensive communication we conclude the paper.
nodes. These nodes are capable of sensing, communicating,
II. H OW A CWSN W ORKS ?
and can be deployed at a cost much lower compared to
a traditional wired sensor system. Such systems are called Similar to the existing WSNs, a CWSN consists of many
wireless sensor networks (WSNs) [1-3]. WSNs are facing tiny and inexpensive sensors where each node operates on
many challenges including the limited bandwidth assigned to a limited battery energy. In a WSN, each node either sends
them which is in general the industrial, scientific and medical and receives data or it is in idle state. However, in a CWSN,
(ISM) band. Bandwidth limitation is highlighted due to the there would be another state called sensing state where the
ever increasing demand for spectrum usage by various wireless sensor nodes sense the spectrum to find spectrum opportu-
systems such as WiFi and Bluetooth. In general, spectrum nities or spectrum holes. Fig. 1 depicts different states for
scarcity is a major concern nowadays. Note that spectrum auc- both networks. Among various tasks for each sensor node,
tion for third-generation (3G) mobile communications yielded the transmission and reception of data are the most energy
USD 17 billion, USD 34 billion and USD 46 billion in US, consuming tasks.
England and Germany respectively [4]. Spectrum sensing task in the CWSN sensing state, can be
To overcome the problem of spectrum scarcity in a WSN, performed either by a distributed or centralized scheme. In a
a new concept of CWSN has been proposed in the literature distributed scheme, each sensor competes with other sensors
[5-7]. Cognitive technology is a fundamentally new approach to access the available spectrum [5]. Thus, each sensor must
have the ability to sense the whole channel, and determine an
optimal scheme to maximize its benefits, such as the number of
978-1-4244-3878-5/09/$25.00 
c 2009 IEEE transmissions over time. However, due to the fact that CWSN

978-1-4244-3878-5/09/$25.00 ©2009 IEEE 593

 are TV broadcasting services and wireless microphones, and
the adopted values are: CDT ≤ 2 sec, CMT = 2 sec, PDT=
-107 dBm (for wireless microphones), PDT = -116 dBm (for
TV broadcasting), PD = 90% and PFA = 10%.
We consider the basic 802.15.4 standard, which has been
used for a number of WSN applications, and present the
cognitive features for its physical and MAC layers. The
802.15.4 standard [13] defines 16 channels, each of 2 MHz
bandwidth, in the 2.4 GHz band. While the 2.4 GHz band is
getting crowded, secondary operation in other spectrum bands
would be inevitable. Reference [5] defines the physical layer
of a CWSN, similar to the 2.4 GHz physical layer in 802.15.4,
Fig. 1. Different states in a WSN vs a CWSN but shifted to a different center frequency. In this way, by using
the same channel bandwidth and QPSK modulation and other
similar physical layer parameters as in the 2.4 GHz mode, it is
sensor nodes are mostly low-powered with limited capabilities, shown that CWSN could achieve the same rate with extended
it may not be feasible to deploy the full functionalities of a dis- range, depending on the center frequency used [5].
tributed scheme in these networks. Thus, in many applications Generally, a CWSN supports much lower data rate and
a centralized scheme is preferred [8]. In a centralized scheme, much smaller transmit power compared to a WLAN [14]. As
spectrum opportunities are detected by a single entity called a result, sensing duration may not be an overhead issue, and
network coordinator [7]. Fig. 2 compares a distributed sensing most of the time, the sensors would be in the idle mode.
scheme with a centralized method. Moreover, the network However, for high throughput networks, the total sensing
coordinator broadcasts a channel switch command to indicate duration would be a critical issue.
an alternate available channel. The alternate channel could
be another licensed channel or an unlicensed channel in the III. CWSN A DVANTAGES
ISM band. The broadcast message could be retransmitted by Adding cognition to a WSN provides many advantages.
multiple nodes to reliably deliver the message. Typically, there Sensor nodes in a CWSN can measure and provide accurate
exist two traffic load configurations in a CWSN: information at various locations within the network. Measure-
• Regular status report: each sensor sends regular status ments made within the network provide the needed diversity
update to the coordinator. The information in such status to cope with multi-path fading. In addition, a CWSN could
updates depends on the particular application. provide access not only to new spectrum (rather than the
• Control commands: control messages are sent by the worldwide available 2.4 GHz band), but also to the spectrum
coordinator. For example, in a heat control application, with better propagation characteristics [5]. The path loss
the coordinator sends commands to switch on/off the decreases as the operating frequency decreases. Therefore, if
heaters. the transmission power of secondary user remains the same, its
Apparently, there would be an overhead of message ex- transmission range increases at lower frequencies [15]. Hence,
change in this scheme as the coordinator should send the a channel decision of lower frequency leads to the following
results of spectrum sensing into a sensor whenever it receives advantages in a CWSN.
transmission request. • Higher transmission range.
To make spectrum sensing task easier, primary users may • Fewer sensor nodes required to cover an specific area.
transmit a pilot signal periodically on a spectrum band if that • Lower energy consumption.
band is occupied by itself. By detecting the presence of such The higher communication range provides CWSNs with the
a pilot signal, secondary nodes can determine if that particular smaller number of hops needed per route. Thus, the average
subcarrier is available or not. end-to-end delays would be also smaller. Moreover, in future
Similar to a CR network, a secondary user must be able applications such as upcoming smart dust networks, oppor-
to detect the presence of a primary signal stronger than the tunistic use of spectrum would be inevitable. The growing
Primary Detection Threshold (PDT) within the Channel Detec- number of such short range nodes distributed in a region
tion Time (CDT). In addition, spectrum opportunity detection which use different radio technologies at unknown locations
must be performed with a success probability greater than or cause the spectral environment to be sharply variable, even
equal to the Probability of Detection (PD) and a probability of over short distances and brief time periods. Hence, time-
false alarm lower than or equal to the maximum Probability limited measurements carried out at a distance from the local
of False Alarm (PFA). Furthermore, the secondary system has area under study cannot characterize the local conditions [4].
to vacate the channel within the Channel Move Time (CMT) Therefore, it is expected that CWSNs to be widely used in the
once a primary signal is detected [5]. The values of the above future. The performance gains are obtained at the cost of a
parameters are set according to each specific application. For slight increase in the protocol complexity and network control
instance, in IEEE 802.22 standard [11], the primary signals overhead. Table I briefly compares WSNs and CWSNs.

594
 Fig. 2. Distributed and centralized sensing schemes in a CWSN

IV. D ESIGN C HALLENGES AND P OSSIBLE R EMEDIES TO solution, the allocated power on all sub-carriers needs to
OVERCOME THE C HALLENGES be decreased.
• Region (3): In this case, data rate does not reach the
In a CWSN, energy efficiency is a key design factor.
minimum constraints. Hence, the allocated power should
Reference [6] investigates a resource allocation problem from
be increased to achieve the data rate requirement under
energy efficiency standpoint. A fully distributed channel se-
the maximal power bound.
lection and power allocation scheme has been proposed to
• Region (4): In this case, the total energy consumption has
minimize the energy per bit over all sub carriers, subject
already exceeded the maximal power bound, but the data
to the required data rate and power constraints. Because the
rate requirement is still not met. Therefore, there is no
optimization problem is defined subject to two constraints, the
feasible solution in this region.
solution space would include four regions. Fig. 3 depicts these
regions. Another challenge which causes an end-to-end delay in
a CWSN, is the sensing duration time. During this interval
• Region (1): The first partition satisfies the data rate and
all traffic is suspended, and spectrum sensing is performed.
power constraints. Hence, the optimal solution lies in this
Particulary, in the cases where the coordinator is the only
region.
sensing-capable node and also the source of the application
• Region (2): The data rate and power allocation both
commands, primary user detection and channel allocation
exceed the upper bounds in this case. To get a feasible
would directly impact the QoS. Reference [5] shows that this
increases the average end-to-end delay.
TABLE I Another challenge which is raised by the multi-channel
CWSN S VS WSN S
nature of CWSNs is the multi-dimensional nature of the
Parameter WSN CWSN resource allocation optimization problem in these networks
Higher transmission range X [6]. In addition to finding the optimal subcarrier set and power
Fewer sensor nodes required X allocation which is obtained individually for each sensor,
Lower energy consumption X
Less end-to-end delay X
another challenging problem would be to consider the co-
More accuracy X channel interference when multiple new users decide to use the
Less protocol complexity X same frequency. Hence, a distributed power control scheme is
Less network control overhead X needed in these cases to manage the co-channel interference.

595
 Lifetime maximization or energy efficiency at the physical the sensing and operational functions in such cases, unique
layer is another design challenge. For instance, by adapting advantages, especially to low power mobile applications would
the modulation type and constellation size and channel coding be provided.
rate, different data rates can be achieved which will directly
V. C ONCLUSION
influence the power consumption of each node, and in turn
will affect the lifetime of the whole network. Cognitive Radio is a recently presented technology that can
A possible solution to extend a CWSN’s lifetime is to potentially improve the efficiency of spectrum utilization. In
perform spectrum sensing task by a subset of sensor nodes. this paper, we presented Cognitive Wireless Sensor Networks
The network coordinator could perform spectrum sensing (CWSNs), and discussed how they work. We further investi-
within the network or a number of sensors could be employed gated the advantages brought about with adding cognition to
and form a distributed scheme in a large network [4]. Due the current WSNs. Finally, some of the main challenges and
to the fact that spectrum sensing is a repetitive process, possible remedies to overcome them were studied.
and it consumes extra energy from battery powered sensors, R EFERENCES
performing spectrum sensing in a small portion of the nodes
[1] W. B. Heinzelman, A. Chandrakasan, and H. Balakrishanan, “An
saves the energy consumption of the whole network. Application-Specific Protocol Architecture for Wireless Microsensor Net-
Sensing frequency could be either fixed or variable. In a works,” IEEE Trans. Wireless Commun., vol. 1, no. 4, pp. 660-70, Oct.
fixed scheme, sensing frequency is a predefined parameter 2002.
[2] A. S. Zahmati, B. Abolhassani, A. A. B. Shirazi, and A. S. Bakhtiari,
which is scheduled at a known periodic interval within a “An Energy-Efficient Protocol with Static Clustering for Wireless Sensor
given duration. On the contrary, in a variable scheme, sensing Networks,” International Journal of Electronics, Circuits and Systems,
frequency adaptively varies according to the changes in the vol. 1, no. 2, pp. 135 - 138, 2007.
[3] N. M. Moghaddam, A. S. Zahmati, and B. Abolhassani, “Lifetime En-
channel environment. Sensing duration is determined by the hancement in WSNs using Balanced Sensor Allocation to Cluster,” IEEE
accuracy of the spectrum sensing algorithm and the required International Conference on Signal Processing and Communications
probability of detection, while the sensing frequency is set in (ICSPC 2007), pp. 101 - 104, Nov. 2007.
[4] N. S. Shankar, C. Cordeiro, K. Challapali, “Spectrum Agile Radios: Uti-
order to satisfy the channel detection time parameter. Since lization and Sensing Architectures,” First IEEE International Symposium
sensing duration schedule could be set well ahead of time, a on New Frontiers in Dynamic Spectrum Access Networks, DySPAN 2005,
possible remedy to avoid overlap transmission within the in- pp. 160-169, Nov. 2005.
[5] D. Cavalcanti, S. Das, W. Jianfeng, and K. Challapali, “Cognitive Radio
terval is to broadcast the schedule by the network coordinator. based Wireless Sensor Networks,” Proceedings of 17th International
Therefore, all nodes can adjust their transmissions according Conference on Computer Communications and Networks, 2008. ICCCN
to the schedule. ’08, pp. 1 - 6, Aug. 2008.
[6] S. Gao, L. Qian, and D. R. Vaman, “Distributed Energy Efficient
In addition, to prevent control commands lost, due to Spectrum Access in Wireless Cognitive Radio Sensor Networks,” Wireless
interference from primary users, some time-out mechanism Communications and Networking Conference 2008, WCNC ’08, pp. 1442
could be implemented to deal with such situation. For ex- - 1447, Mar. 2008.
[7] S. Gao, L. Qian, D. R. Vaman, and Q. Qu, “Energy Efficient Adaptive
ample, nodes can use a timer to proactively check for the Modulation in Wireless Cognitive Radio Sensor Networks,” IEEE Inter-
presence of the coordinator’s message [5]. Finally, in some national Conference on Communications 2007, ICC ’07, pp. 3980 - 3986,
applications, operational networks need to be mobile while the Jun. 2007.
[8] S. Byun, I. Balasingham, and X. Liang, “Dynamic Spectrum Allocation
sensing modules does not need to be mobile. By separating in Wireless Cognitive Sensor Networks: Improving Fairness and Energy
Efficiency,” 68th IEEE Vehicular Technology Conference, VTC’08, pp. 1
- 5, Sept. 2008.
[9] J. Mitola, “Cognitive Radio: An Integrated Agent Architecture for
Software Defined Radio,” Ph.D. dissertation, Royal Inst. Technology,
Stockholm, Sweden, 2000.
[10] S. Haykin, “Cognitive Radio: Brain-Empowered Wireless Comm.,”
IEEE J. Selected Areas in Comm., vol. 23, no. 2, pp. 201- 220, Feb.
2005.
[11] IEEE 802.22 draft standard, “IEEE P802.22TM/D0.4 Draft Standard for
Wireless Regional Area Networks,” http://www.ieee802.org/22/, Novem-
ber 2007.
[12] C. Cordeiro, M. Ghosh, D. Cavalcanti, and K. Challapali, “Spectrum
Sensing for Dynamic Spectrum Access of TV Bands,” CrownCom 2007,
Orlando, Florida, August 2007.
[13] IEEE 802.15.4 Standard, “Wireless medium access control (MAC) and
physical layer (PHY) specifications for low-rate Wireless Personal Area
Networks (LR-WPANs),” 2003 Edition.
[14] A. Willig, “Recent and Emerging Topics in Wireless Industrial Com-
munications: A Selection,” IEEE Transactions on Industrial Informatics,
vol. 4, no. 2, pp. 102 - 124, Feb. May 2008.
[15] I. F. Akyildiz, W. Lee, M. C. Vuran, and S. Mohanty, ”NeXt Gener-
ation/Dynamic Spectrum Access/Cognitive Radio Wireless Networks: A
Survey,” Computer Networks, vol. 50, pp. 2127-2159, Sep. 2006.

Fig. 3. Partitions of the constraint space

596