Content uploaded by Jon Mark
Author content
All content in this area was uploaded by Jon Mark on Jun 10, 2014
Content may be subject to copyright.
Exploiting Heterogeneity Wireless Channels for Opportunistic
Routing in Dynamic Spectrum Access Networks
Yongkang Liu1,LinX.Cai
1,2, Xuemin (Sherman) Shen1, and Jon W. Mark1
1Department of Electrical and Computer Engineering, University of Waterloo, Waterloo, ON N2L 3G1, Canada
{y257liu, lcai, xshen, jwmark}@bbcr.uwaterloo.ca
2Department of Electrical and Computer Engineering, McMaster University, Hamilton, ON L8S 4L8, Canada
Abstract— In this paper, we exploit the heterogeneity of wire-
less channels and propose an efficient opportunistic cognitive
routing (OCR) scheme for dynamic spectrum access (DSA)
networks. We first introduce a novel routing metric by jointly
considering physical characteristics of spectrum bands and di-
verse activities of primary users (PU) in each band. To effectively
explore the spectrum opportunities, a proper channel sensing
sequence for fast and reliable message delivery is determined
by secondary users (SU) in a distributed way. We then develop
a greedy forwarding scheme that SUs can select the next hop
relay based on the geometry information and channel access
opportunity of their one hop neighbors. For the proposed OCR,
as routing control messages are locally exchanged, SUs can
efficiently make the routing decision and opportunistically access
the available channels. We further evaluate the performance of
OCR via extensive simulations. It is shown that our proposed
scheme outperforms existing opportunistic routing schemes in
DSA networks by exploiting the heterogeneity of spectrum bands
for opportunistic channel access.
I. INTRODUCTION
Dynamic spectrum access (DSA) network has emerged as
a promising solution to improve the spectrum utilization by
enabling secondary users (SU) to opportunistically share the
unoccupied spectrum resources with primary users (PU). The
DSA network differs from other traditional wireless networks
in that there is no statically allocated spectrum, and SUs
need to dynamically switch among various frequency bands
for opportunistic transmissions. The highly dynamic network
resources and the inherent multi-channel structure make the
protocol design for a DSA network very challenging.
Most research activities in a DSA network focus on how
to efficiently and accurately detect the spectrum opportunities
for medium access in the physical and MAC layers with
or without user cooperations [1], [2], [3]. Recently, routing
in a DSA network has attracted considerable attentions. By
incorporating spectrum sensing operations and utilizing ge-
ometry information, some spectrum-aware routing protocols
have been proposed for joint channel assignment and route
establishment [4], [5]. However, these routing protocols needs
to maintain a pre-defined end-to-end route in the routing table,
and thus may not adapt well to provide a reliable route when
the PU activities are highly dynamic. On the other hand,
opportunistic routing is considered as an efficient approach
for wireless networks by allowing senders at each hop to
make local routing decisions in an opportunistic way. As
most opportunistic routing protocols are studied in a single
channel scenario, how to extend opportunistic routing in a
multi-channel DSA network is still an open research issue. In
addition, different spectrum bands in various primary networks
may exhibit different propagation characteristics that result
in different achievable data rates, communication coverages,
etc., which may have significant impacts on the DSA network
performance and thus should be taken into consideration in
the routing protocol design.
In this paper, we exploit the heterogeneity wireless channels
in the design of opportunistic cognitive routing (OCR) scheme.
In specific, we first introduce a novel metric to effectively
characterize the capability of an SU to forward data over a
channel, considering diverse usage patterns and propagation
characteristics in different channels. Based on the metric,
each SU can determine a proper spectrum sensing sequence
to effectively explore the spectrum opportunities for data
forwarding. To further improve the reliability and efficiency
of data delivery, each relay candidate in the same channel
can distributively determine its relaying capability based on
its location and transmission history. An SU with a higher
distance gain (i.e., closer towards the destination) and success
transmission capability will be more likely to be selected as
the next hop relay. By determining the best relay at each
hop, the proposed OCR adapts well to the network dynamics,
e.g., PUs’ activities, user mobility, etc., in a multi-hop multi-
channel DSA network.
The main contribution of the paper is three-fold. First,
we introduce a novel metric to effectively characterize the
capability of an SU for data forwarding in each channel, taking
into consideration different channel characteristics and channel
usage patterns in a primary network. Second, we develop a
distributed scheme to select the best relay in an available
spectrum band, by balancing the data forwarding distance gain
and the success transmission probability. Thirdly, we evaluate
the performance of the proposed OCR and demonstrate that
the proposed OCR outperforms existing opportunistic or geo-
graphical routing protocols by jointly exploring heterogeneity
wireless channels and geometry information of SUs.
The remainder of the paper is organized as follows. The
system model is introduced in Section III. A multi-channel
opportunistic cognitive routing scheme is proposed in Sec-
tion IV, followed by performance evaluation in Section V.
Concluding remarks are given in Section VI.
978-1-61284-231-8/11/$26.00 ©2011 IEEE
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE ICC 2011 proceedings
II. RELATED WORK
Most research efforts in DSA networks have been put on
issues in the physical layer spectrum sensing, or the link
layer medium access control and resource management [1].
Recently, researchers have recognized that these designs would
not harvest the expectable benefits without supports from
the upper layers. As a DSA network usually coexists with
multiple primary networks, the routing scheme design is of
critical importance when the end-to-end path performance is
considered.
In a DSA network, the route for a source-destination pair
is comprised of spectrum opportunities over multiple hops
via spectrum sensing operations. In [4], a routing scheme is
proposed for the joint channel-route design with a layered
graph model. The solution does not address reliability issues
against the route dynamics because any PU activity along the
route may interrupt the ongoing DSA transmission. In [6],
a probabilistic path metric is proposed for the source-based
path selection, and multiple channels are bound to make the
estimated capacity of DSA link satisfy a required demand.
Geographic routing is considered efficient for wireless net-
works due to its low complexity and scalability under dynamic
wireless conditions [7], [8]. In [5], geographic routing is
applied in the DSA networks. Based on the path pool including
those paths optimized according to geographic forwarding
rules in each channel, the end-to-end route consists of the links
not affected by PU. As the channel condition is highly dynamic
in a DSA network due to PU activities, frequent route updates
may significantly reduce efficiency of the routing protocol,
especially when the transmission delay is comparable to the
short holding window of the spectrum opportunities [9].
On the other hand, opportunistic routing has been proposed
for mobile ad hoc wireless networks as it adapts well in a
dynamic wireless environment [10], [11]. Instead of using a
pre-defined end-to-end path, the forwarding decision is made
in an opportunistic way by the sending node at each hop [12],
based on the instantaneous local link capacity or queuing
information, etc. To the best of our knowledge, most existing
opportunistic routing protocols have been proposed for a single
channel scenario. How to exploit the heterogeneity wireless
channels in the design of multi-channel opportunistic routing
in a DSA network remains an open issue.
III. SYSTEM MODEL
In this paper, we consider a densely deployed infrastructure-
less multi-hop DSA network consisting of CNchannels,
C1,C
2,··· ,C
N, as shown in Fig. 1. Secondary users are
randomly distributed over the network. Each SU is equipped
with one half-duplex cognitive transceiver which can switch
to one channel at one time for spectrum sensing and op-
portunistic transmissions. The sensing distance is two times
the transmission distance. In the DSA network, SUs have the
location information of the destination D, i.e., the gateway,
and themselves through GPS or other localization services.
We assume the channel hopping sequence of the destination
Channel N
Channel 2
Channel 1
S
D
S
D
S
D
r
rD
d
SD
d
PUSU
Fig. 1. Multi-hop DSA Network
gateway is known to all its one-hop neighbors so that there
is no transceiver synchronization problem between the SUs
and the destination gateway. As the frequency channels may
span over a wide spectrum range, different channels exhibit
different propagation characteristics, which result in various
transmission distances, and diverse achievable data rates in
these channels. As shown in Fig. 1, the SU’s communication
coverage differs in each channel, which results in different sets
of neighboring SUs and PUs. On the one hand, with more SUs
in the neighborhood, it is very likely to find a better relay. On
the other hand, more PUs in the communication coverage may
reduce the transmission opportunity of the tagged SU as more
PU activities may be detected. An independent and identically
distributed two-stage ON/OFF model is applied to all PUs in
each channel. SUs distributively trace the ON/OFF parameters
by periodic sensing. How to assure accurate channel sensing
and parameter estimation has been extensively studied in [1],
which is beyond the scope of this paper.
IV. OPPORTUNISTIC ROUTING IN DYNAMIC SPECTRUM
ACCESS NETWORKS
In this paper, we first introduce a channel sensing metric
that characterizes the dynamic properties of DSA networks,
based on which a secondary sender can determine a proper
spectrum sensing sequence and probe the spectrum opportu-
nities to broadcast a Route REQuest (RREQ) message. We
then propose a novel routing metric for selecting the best
SU that has the highest opportunistic relay capability, using a
distributed medium access mechanism. Basically, the available
neighboring SUs operating in the same channel as the sender
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE ICC 2011 proceedings
receive the RREQ message and determine their capability to
forward the data towards the destination. An SU with a higher
forwarding capability will select a smaller backoff window
and transmit a Route REPly (RREP) message, and thus is
more likely to be selected as a relay in the next hop. After the
successful exchange of RREQ and RREP messages, the sender
will forward the data to the selected SU. Otherwise, the sender
will attempt to re-broadcast the RREQ in the next channel. The
procedure repeats until the data message successfully reaches
the destination. The pseudo codes of the sender and receiver
algorithms are tabulated in Algorithms 1 and 2. More details
of the metric design and opportunistic routing protocol will be
presented in the following subsections.
Algorithm 1 Sender
1: if (Source user Shas a packet for Destination user D)then
2: Scalculates γk, determines the spectrum sensing sequence
and starts sensing in the first channel;
3: if Channel is sensed idle then
4: Broadcast an RREQ message;
5: if Receive an RREP message correctly then
6: Transmit data to the SU that responds with an RREP in
current channel;
7: else if Receive multiple RREP messages then
8: Retransmit an RREQ to notify collided SUs;
9: Go to Step 5;
10: end if
11: else
12: Select the next channel and start sensing;
13: end if
14: Update γkand reordering the spectrum sensing sequence;
15: Go to Step 3;
16: end if
A. Sensing Sequence
To effectively probe the spectrum opportunities in a highly
dynamic DSA network, it is critical to determine a set of chan-
nels with an appropriate spectrum sensing sequence for each
SU. As radio waves at different frequency bands propagate
differently, various channels may exhibit diverse propagation
characteristics. For example, some frequency bands can sup-
port very high data rate at very short distance, e.g., millimeter
wave bands, while others are used for low rate communications
over medium and long ranges. In addition, the activities of SUs
in DSA network is heavily dependent on those of PUs. If the
channel is occupied by a large number of active PUs, it is
less likely that an SU can find an opportunity in this channel.
Recognizing the diversity in different channels, we propose a
new metric to characterize the forwarding capability in each
channel, which is given by
γk=αTOF F
TON +TOF F
Rk+(1−α)Dk(1)
where TOF F and TON are the average time durations that
the channel is idle and busy, respectively, TOFF
TON +TOFF is
the probability that the channel is sensed idle, Rkis the
achievable data rate in channel kwith the corresponding
transmission distance Dk, normalized by maxkDkfor k∈
Sensing Relay Selection Data Transmission
One Hop T ransm issi on
RREQ RREP
CH 1
CH 2
CH 3
CH 4
CH 5
SU
sensin g
Sender Transmission
(RREQ, Data packet)
Receiver Feedback
(RREP, Data ACK)
Fig. 2. Relay Selection
CN, and αis a weighting factor. Each SU updates TOFF
and TON via periodic sensing. In each channel, the SU may
achieve different data rates at different distances with adaptive
modulation technique. Generally, a fewer number of hops
is required to forward the data with a larger transmission
distance. In this paper, we choose the largest Dkto calculate
γkand sort channels in the descending order of γkfor channel
sensing. How to choose the most proper modulation scheme
to balance Dkand Rkis under further investigation. SUs
select the channel with the highest forwarding capability (i.e.,
more available network resources and a longer transmission
distance), and attempts data communication when opportunity
appears. We can balance the available channel resources and
the transmission coverage by adjusting the value of α.The
impacts of different αon the routing performance will be
studied in Section V.
Algorithm 2 Receiver
1: Listening on the current channel;
2: if An RREQ message is received then
3: Calculate νrand select a backoff timer;
4: while Backoff timer !=0do
5: if Overhear other RREP messages then
6: Stop the backoff timer and go to Step 1;
7: end if
8: end while
9: Send an RREP message to Sender and go to Step 1;
10: end if
B. Relay Selection
Once a sender discovers a spectrum opportunity in one
channel via spectrum sensing, it broadcasts a RREQ message
over the channel. All SUs operating over the same channel and
in the transmission range of the sender are relay candidates
which can help forward the data in the next hop. To facilitate
opportunistic routing in DSA network, each SU needs to
respond a RREP message to the sender, so that the sender
can select the best candidate to relay the message towards the
destination.
In our proposed protocol, an SU, e.g., SU r, estimates its
relay capability by calculating
νr=βPsr+(1−β)dSD
drD
,(2)
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE ICC 2011 proceedings
Number of SUs 200
Number of Pus per channel 4
Channel number 6
Delay Threshold 150 ms
λON 0.08 /ms
λOF F 0.04 /ms
Source-destination distance 360 m
Spectrum Sensing Time 1ms
A backoff mini-slot 4μs
TAB LE I
SIMULATION PARAMETERS
where the successful transmission probability Psrindicates
the opportunistic transmission capability of user rin a multi-
hop DSA network, which is obtained from the transmission
history of SU r,drD is the distance between SU rand the
destination gateway, and dSD is the distance between the
source and destination, as shown in Fig. 1. βis a weighting
factor to balance the user transmission capability and the
forward distance gain. SU ris qualified in the relay candidate
selection only when νris within [νmin,ν
max]indicated in
the RREQ message. To reduce the possible collisions among
multiple relay candidates, we use a contention based MAC for
relay selection process. As shown in Fig. 2, each SU selects
a backoff timer Wr∈[0, W] based on the estimated νr.The
backoff timer reduces by one for every idle mini-slot and an
SU can transmit only when its backoff timer reaches zero.
Therefore, the SU with the highest νrwill transmit first and
thus be selected as the next hop relay. Other SUs stop their
own timer upon overhearing a RREP from other SUs. It is
also possible that multiple SUs may select the same backoff
timer which causes collisions. If multiple copies of RREP are
received at the sender, the sender will re-transmit a RREQ
message, and only those collided SUs in the previous round
contention will enter the second round contention by randomly
selecting a backoff timer in [0, W]. The process repeats until
only one relay is selected. After successfully delivering the
data to the next hop, the sender switches to the listening state
over the same channel.
V. P ERFORMANCE EVA L U AT I O N
We evaluate the performance of the proposed routing proto-
col under the network parameter settings in Table I if no other
specification is made in the individual study. The simulation
model is built in C++ using the reference of the Cognitive
Radio Cognitive Network Simulator [13]. The heterogeneity
wireless channel patterns, such as the transmission distance
and the transmission data rate, are listed in Table II. PUs and
SUs are randomly distributed in an area of 300×300 m2.One
connection is initiated in the DSA network, and the distance
between the source-destination pair is arbitrarily set to 360 m.
We use a saturated flow with a package size of 500 bytes,
and packet delay bound of 150 ms. We model the PU activity
channel PU Tx range SU Tx range SU data rate
CH1, CH2 50 m20 m6Mbps
CH3, CH4 100 m60 m4Mbps
CH5, CH6 500 m100 m0.5Mbps
TAB LE I I
CHANNEL USAGE PATTERN
1 2 3 4 5 6 7 8 9 10
10
20
30
40
50
60
70
80
90
Number of PUs per channel
End−to−End Packet Delay (ms)
GPSR+
OCR alpha 0.6
OCR alpha 0.9
Fig. 3. Comparison of End-to-End Delay under Different PU Activities
as an exponential ON-OFF process with parameters λON and
λOF F . SUs individually estimate the ON/OFF durations in
each channel. The channel switch time plus the minimum
sensing duration with energy detection is 1ms. Each mini
slot is 4 μs. We run each experiment for 120 s, and the first
20 s in each trial run is for SUs to trace PU activities in the
channels. We repeat them 50 times to calculate the average
value.
Our OCR scheme is compared with the ”GPSR+” protocol
which is an extensive version of the classical GRSR pro-
tocol [7] in a multi-channel DSA network. In GPSR+, the
sender senses the channels in descending order by the channel
maximum transmission distance, and greedy forwarding is
undertaken by the sender in the relay selection. The relay
at the next hop is determined by the sender with the closest
position to the destination among all relay candidates in the
same channel as the sender.
We first evaluate the end-to-end delay performance of the
proposed OCR scheme under different PU activities. We set
the weighting factor αto be 0.6and 0.9, respectively, and
compare the performance with GPSR+. GPSR+ prefers the
channel which has the larger SU transmission range in the
sensing sequence. As shown in Fig. 3, it can be seen that
our proposed OCR scheme significantly outperform GPSR+
in terms of end-to-end packet delay, especially in the case
α=0.6. Using a greedy forwarding algorithm to select a relay
with the highest distance gain, GPSR+ can achieve shortest
path with minimum hops in most cases. However, in a DSA
network, more PUs’ activities may be detected over a longer
communication coverage, and less transmission opportunity
can be exploited for SUs’ transmissions. In our proposed
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE ICC 2011 proceedings
0 5 10 15 20
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Number of End−to−End Path Hops
CDF
GPSR+, PU 2
GPSR+, PU 8
OCR alpha 0.6, PU 2
OCR alpha 0.6, PU 8
Fig. 4. CDF of Path Hop Counts under Different PU activities
100 150 200 250 300 350
30
35
40
45
50
55
60
65
70
75
80
Number of SUs
End−to−End Packet Delay (ms)
GPSR+
OCR alpha 0.6
Fig. 5. Comparison of End-to-End Delay under Different SU Densities
OCR, by jointly considering the channel characteristics and
PU activities, SUs can efficiently exploit the channel access
opportunities for data forwarding. By well balancing the
two factors, a much lower delay can be achieved. Optimal
parameter setting is beckon for further investigation.
We then study the impacts of heterogeneity wireless chan-
nels on the performance of the proposed routing scheme. With
a shorter communication coverage, there is less likely that an
SU will be interfered by a PU, and thus the SU is able to
explore more spectrum opportunities for data transmissions.
Fig. 4 shows the cumulative density function of the hop counts
for the end-to-end transmission in the DSA network. There is
a marked shift of the average hop count to the higher value
which means the transmission distance per hop shrinks due to
PU activity.
We also investigate the performance under different node
density. The number of PU is 4in each channel. As shown
in Fig. 5, the delay drops drastically as the number of SUs
increases from 100 to 200. Then the decrease rate slows down
as the SU number further increases above certain level like
250 SUs here. In a sparse network condition, more SUs can
improve the network connectivity, and it is more likely to find a
proper relay, thus the number of channel switching and sensing
can be significantly reduced and a better delay performance
can be achieved. When the network is sufficiently dense, the
performance improvement mainly relies on the diversity in the
opportunistic relay selection. In this case, further increasing
the node density of DSA networks has little impact on the
delay performance of SUs.
VI. CONCLUSIONS
In this paper, we have proposed an opportunistic routing
scheme for multi-channel DSA networks, by jointly exploiting
heterogeneity wireless channel characteristics and geometry
information of SUs. By taking spectrum opportunity and
selecting the best relay at each hop, our proposed routing
scheme can adapt well to the network dynamics and achieve
better performance than existing cognitive routing protocols.
Efficient opportunistic routing with QoS provisioning for mul-
timedia flows is currently under investigation.
ACKNOWLEDGEMENT
This work has been supported by research grants from
the Natural Sciences and Engineering Research Council
(NSERC) of Canada.
REFERENCES
[1] H. Kim and K. G. Shin, “Efficient Discovery of Spectrum Opportunities
with MAC-Layer Sensing in Cognitive Radio Networks,” IEEE Transac-
tions on Mobile Compututing, vol. 7 no.5, pp. 533–545, May 2008.
[2] L. X. Cai, Y. Liu, X. Shen, J. W. Mark, and D. Zhao, “Distributed QoS-
aware MAC for multimedia over cognitive radio networks,” Proc. IEEE
Globecom’10, Miami, USA, December 6–10, 2010.
[3] A. Alshamrani, X. Shen, and L. Xie, “A Cooperative MAC with Efficient
Spectrum Sensing Algorithm for Distributed Opportunistic Spectrum
Networks”, J. Communications, vol. 4, no. 10, pp. 728–740, 2009.
[4] C. Xin, B. Xie, and C.-C. Shen, “A Novel Layered Graph Model
for Topology Formation and Routing in Dynamic Spectrum Access
Networks,” Proc. IEEE DySPAN’05, 2005.
[5] K. R. Chowdhury, and M. D. Felice, “SEARCH: A Routing Protocol for
Mobile Cognitive Radio Ad-hoc Networks,” Computer Communications
(Elsevier), vol. 32, no. 18, pp. 1983–1997, 2009.
[6] H. Khalife, S. Ahuja, N. Malouch, and M. Krunz, “Probabilistic Path
Selection in Opportunistic Cognitive Radio Networks,” Proc. IEEE
Globecom’08, pp. 4861–4865, 2008.
[7] B. Karp and H. T. Kung, “GPSR: Greedy Perimeter Stateless Routing for
Wireless Networks,” Proc. ACM Mobicom, 2000.
[8] A. Abdrabou and W. Zhuang, “Statistical QoS routing for IEEE 802.11
multihop ad hoc networks,” IEEE Transactions on Wireless Communica-
tions, vol. 8, no. 3, pp. 1542–1552, March 2009.
[9] H. Khalif, N. Malouch, and S. Fdida, “Multihop Cognitive Radio Net-
works: To Route or Not to Route,” IEEE Network Magazine, vol. 23, no.
4, pp. 20–25, 2009.
[10] S. Biswas, R. Morris, “ExOR: Opportunistic Multi-hop Routing for
Wireless Networks,” ACMSIGCOMM Computer Communication Review,
vol. 34, no. 1, pp. 133–144, 2005.
[11] L. X. Cai, H. Hwang, X. Shen, J. W. Mark, and L. Cai, “Optimizing
Geographic Routing for Millimeter-wave Wireless Networks with Direc-
tional Antenna,” Proc. Broadnets’09, Madrid, Spain, September 14–16,
2009.
[12] Y. Bi, L. X. Cai, X. Shen, and H. Zhao, “Efficient and Reliable Broadcast
in Inter-Vehicle Communications Networks: A Cross Layer Approach,”
IEEE Transactions on Vehicular Technology, vol. 59, no. 5, pp. 2404–
2417, 2010.
[13] Michigan Tech. Univ., “Cognitive Radio Cognitive Network Simulator,”
Available at http://stuweb.ee.mtu.edu/ ljialian/index.htm.
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE ICC 2011 proceedings