Alexandria Engineering Journal (2013) 52, 665–670

Alexandria University

Alexandria Engineering Journal

www.elsevier.com/locate/aej
www.sciencedirect.com

ORIGINAL ARTICLE

Channel assignment with closeness multipath routing

in cognitive networks
Islam Beltagy a, Moustafa Youssef b, Magdy Abd El-Azim a,*
,
Mohamed El-Derini a

a
Department of Computers and Systems Engineering, Alexandria University, Egypt
b
Department of Computer Science and Engineering, EJUST, Egypt

Received 7 July 2012; revised 21 July 2013; accepted 22 July 2013

Available online 24 September 2013

KEYWORDS
Cognitive networks;
Channel assignment;
Multipath routing;
Routes closeness
Abstract Routing in cognitive networks is a challenging problem due to the primary users’ (PU)
activities and mobility. Multipath routing is a general solution to improve reliability of connections.
Routes closeness metric was proposed for multipath routing in cognitive networks; however, the
proposed technique supports only one channel [4]. This work proposes a multichannel assignment
technique for multipath routing using routes closeness as the routing metric. It relies on the nodes of
the different paths to early detect the existence of PUs and notify nodes on other routes to avoid
using the PU’s channel that is going to be interrupted. In case the field has PUs occupying all chan-
nels, channels assigned to nodes based on how far the nodes are from the PU. Simulation results
show the effectiveness of the channel assignment technique in increasing end-to-end throughput
and decreasing delay.
ª 2013 Production and hosting by Elsevier B.V. on behalf of Faculty of Engineering, Alexandria
University.

1. Introduction cally access the spectrum when no active PUs is in range on

Cognitive network is an emerging area for wireless technology
that has the potential of increasing spectrum utilization. The
network consists of primary users (PU), who are licensed to
use the spectrum, and secondary users (SU) who opportunisti-
* Corresponding author. Tel.: +20 1223338246.
E-mail address: magdy@alexu.edu.eg (M.A. El-Azim).
Peer review under responsibility of Faculty of Engineering, Alexandria
University.
Production and hosting by Elsevier
1110-0168 ª 2013 Production and hosting by Elsevier B.V. on behalf of Faculty of Engineering, Alexandria University.
http://dx.doi.org/10.1016/j.aej.2013.07.006
this spectrum. The communication starts by channel sensing,
where each SU senses the spectrum to determine the available
set of channels to be used. Then, for any two adjacent SUs to
communicate, they need to be both tuned to the same channel.
Selecting channel for communication between adjacent SUs
from the set of available channels is called channel assignment.
Routing is the discovery and maintenance of multihop routes
between source and destination SUs [1].
Beltagy et al. [4] proposed a multipath routing protocol for
cognitive networks using routes closeness as the routing met-
ric. It discovers multiple paths between source and destination
then select the most non-close routes. Closeness of two routes
is a measure of how likely that a single PU can interrupts both
routes at the same time. Closeness routing metric is explained
666
in greater details in Section 2. Beltagy et al. [4] showed that this
routes selection enhances connections stability. In [4], the rout-
ing protocol is limited to a single data channel. That is, all PUs
and SUs are sending their data over the same channel and SUs
should stop communication once this channel is occupied.
However, the typical setting is that multiple PUs are occupying
different channels and SUs can switch from a channel to
another according to PUs existence. This work proposes a
channel assignment technique to be used on top of the routes
closeness routing protocol to support multichannel
assignment.
Channel assignment techniques can be categorized using
the assignment’s level of granularity. The channel assignment
decisions can be per packet, per link, per flow or per segment
[5].
 Packet-based: Different channel assignment decision can be
made for different packets at the same node. Frequent
channel switching results in significant channel switching
delay which is not practical [3].
 Link-based: A link between two nodes keeps using the same
channel for a period of time. A node connected to other two
nodes on two different channels is expected to encounter a
significant switching delay.
 Flow-based: All packets of a certain source-destination con-
nection are assigned to the same channel. Flow-based chan-
nel assignment is not practical in cognitive networks
because of the spacial variation of channels availability
because of primary users’ activities.
 Segment based: It is a compromise between link-based and
flow-based channel assignment. The flow is partitioned into
few number of segments - hopefully one segment. All links
belong to the same segment are assigned to the same chan-
nel. This technique is more suitable for channel assignment
in cognitive networks [5].
This paper proposes a segment based channel assignment
technique on the top of the routes closeness routing protocol.
The channel assignment technique is two steps: (1) early detec-
tion of PUs and nodes notifications, (2) channel assignment at
nodes. The first step is, once a SU detects the existence of a
PU, it notifies the source node. Then, the source node notifies
all the nodes between the source-destination pair. The second
step is that each node selects the least likely to be interrupted
channel. It is the channel that either has no PU on the field or
its PU is far from the node. The routes closeness value deter-
mined through routing phase is used as an indication for
how far the PU is from the node.
The rest of the paper is organized as follows: related work is
Section 2, then Section 3 presents the problem description and
proposed channel assignment technique. Section 4 is the per-
formance evaluation, followed by the conclusion on Section 5.
2. Related work
Research in cognitive networks spans many layers of the net-
work protocol stack including physical layer [2], MAC layer
[7], routing layer and transport layer [12]. This paper’s contri-
bution is in the routing layer.
Depending on the PU’s activity, the routing protocol is se-
lected. For low PU activity, classical routing protocols can be
I. Beltagy et al.
used because the network topology is relatively static, while in
case of highly active PUs, ad-hoc routing protocols are more
appropriate [12]. This paper assumes highly active and mobile
PUs. The used routing protocol is an adaptation of the ad-hoc
on-demand distance vector routing protocol (AODV) [14].
Routing protocols are also classified according to spectrum
awareness [12]. It could be a fully spectrum-aware routing pro-
tocol, where all spectrum availability information is gathered
in a centralized entity. Xin et al. [17] is an example of a fully
spectrum-aware routing protocol, where data about links
and spectrum availability is collected in a centralized entity
then conceptualized as a layered graph, one layer encodes
neighboring nodes, and other layers to capture spectrum avail-
ability between nodes.
Routing protocols can also be locally spectrum-aware,
where spectrum availability information is collected locally in
nodes via exchanging control messages. The routing protocol
presented in this paper takes this approach. The source node
broadcasts a route request (RREQ) message which floods the
network looking for the destination node and looking for the
best path to it. The discovered route is sent back to the source
node in a route reply (RREP) message. This technique for find-
ing candidate routes is called source routing [8]. Control mas-
sages are usually broadcasted in a common control channel
(CCC) [10] -as assumed in this paper-, or over all available
data channels as in [9].
As pointed in [12], routing protocols can also be classified
based on their optimization objective, minimizing power con-
sumption [15], minimizing delay [11] and maximizing through-
put [13]. The primary objective of the routing protocol that
this paper is based on [4] is to avoid PU’s interruption as pos-
sible which is going to increase throughput. The channel
assignment algorithm presented in this paper targets minimiz-
ing channel switching which minimizes the overall delay.
This paper is an extension for [4]. Beltagy et al. [4] proposed
a multipath routing protocol for cognitive networks using
‘‘routes closeness’’ as the routing metric. It discovers multiple
paths between source and destination then select the most non-
close routes. According to the routes closeness metric, close-
ness(R1, R2) < closeness(R3, R4) if R1, R2 are more likely to
be interrupted by a single PU that R3, R4. We say that R1,
R2 are ‘‘closer’’ to each others than R3, R3. Closeness of R1,
R2 is calculated as the ‘‘area’’ where a single PU can interrupt
both R1, R2. This area is approximated as polygon to facilitate
calculating its area. Once closeness of all pairs of candidate
routes is calculated, the routing protocol uses a greedy approx-
imate algorithm to select the most non-close set of routes.
3. Channel assignment
This section starts by describing the system model followed by
problem description. Then, it proposes the algorithm of the
channel assignment technique.
3.1. System model
We assume an environment of stationary - with known loca-
tions - SUs and mobile PUs. Each SU is equipped with two
interfaces, one dedicated to the common control channel that
is not subject to interruption by PUs. The other interface is
used for the data channels. Data channels are shared with
Channel Assignment with Closeness Multipath Routing in Cognitive Networks
PUs and subject to interruption because of PUs activities. This
interface can switch between multiple channels. Channel
switching encounter a channel switching delay resulting from
the hardware switching time and synchronization protocol be-
tween the communicating nodes.
3.2. Problem description
The channel assignment problem is: Given a route -as a set of
consecutive links-, and the channels availability information
on each link, assign a channel for each link.
Consider that the routes are selected using the routes close-
ness metric as in [4] despite the channel availability informa-
tion. The source node has R selected routes, but the channel
assignment is not determined.
The channel assignment technique needs to have the follow-
ing features:
1. Fast adaptation to the rapidly changing channels availabil-
ity. Channels availability changes rabidly because of the
speedy mobile PUs.
2. Assign different channels to different routes. If routes are
on different channels, they could not be interrupted by
the same PU even if they are close to each others. The ques-
tion is why to select non-close routes if they are not going to
be interrupted by a single PU even if they are close. Select-
ing non-close routes is beneficial in case of large number of
PUs. In the worst case, the large number of PUs may
reduce the channel’s availability to a single channel as in
the single channel assignment presented in [4]. It may seem
that this case has low probability and it does not need to be
handled, although that is not correct. If number of PUs is
small, the channel assignment can be trivially solved by
selecting a channel with no PUs. Therefore, the large num-
ber of PUs and the small number of spectrum opportunities
are typical. The other reason to use non-close routes is to
allow time for the transmission of the control messages.
3. Notify routes before being interrupted by a PU. Once the
route is notified, it can switch to another channel that is
not going to be interrupted. This way, nodes are able to
select channel that is less likely to be interrupted.
4. Minimize number of channel switching. It is a general
objective to any channel assignment technique. Arguably,
a more important objective is to minimize interference
between different simultaneous communication flows; how-
ever, we do not consider it here because it is already han-
dled by the routing protocol. The routing protocol selects
non-close routes which are less subject to inter-connections
interference.
A channel assignment technique that satisfies these require-
ments is suitable to an environment of mobile PUs and high
variability in the spectrum opportunities.
3.3. Algorithm
This section suggests a channel assignment technique that sat-
isfies the requirements discussed before. It consists of two
phases, PUs detection and notification then channel assign-
ment along routes.
667
Phase1: PUs’ detection and notification
PUs’ detection step is the role of SUs along routes. When-
ever a PU interrupts a route, the interrupted SU sends a noti-
fication to the source node. Now, source node has a list of
interrupted channels and interrupted routes.
The notification step is the role of the source node. The
source node sends a control message over each route contain-
ing the updated information about interrupted channels.
Source node sends the following information to each route.
The information sent changes from route to another
1. RouteId: routes are numbered from 0 to R 1. RouteId is
the index of the route.
2. INTERRUPTED: A set of interrupted channels. Each
interrupted channel is sent as a pair (Channel Number,
Closeness Value), where Closeness Value is calculated
between this route and the last route interrupted by a PU
active on this channel.
The above information is sent over all nodes in the routes.
This information is enough for each node to select the most
appropriate channel for accessing.
Phase2: Channel Assignment at nodes
In this phase, each node selects the most suitable channel to
be used. Nodes use information received from source in addi-
tion to the local sensing information to make the channel
assignment decision. Each node has the following information
1. RouteId: Received from source node.
2. INTERRUPTED: Received from source node.
3. AVAILABLE: Set of available channels to this node. The
channel is available if it has no PU in range. The list of avail-
able channels is determined locally using channel sensing.
Each node uses Fig. 1 to make the channel assignment deci-
sion. Fig. 1 runs locally it each node.
3.4. Algorithm description
Fig. 1 selects the most appealing channel. The channel is pre-
ferred if it is available and not interrupted in any other route.
If no such channel, the algorithm selects a channel from the
available but interrupted channels. Otherwise, no channel is
available.
Channels in each group are accessed according to a spe-
cific order. In case of available not interrupted channels,
channels are accessed according to the channel access se-
quence. Channel access sequence is a predefined order as
a function of RouteId. Each route accesses the channels in
an order different from other routes. The access sequence
reduces probability of two routes accessing the same chan-
nel. The channel access sequence is to access channels start-
ing from channel number RouteId \ ChN/R circularly, where
ChN is the total number of channels, and R is the total
number of routes. E.g. if number of channels = 6 and num-
ber of routes = 2, Route0 has the channel access sequence
0, 1, 2, 3, 4, 5 while Route1 has the sequence 3, 4, 5, 0,
1, 2. This assures that each route has a designated set of
floor(ChN/R) routes to try, before trying channels desig-
nated to other routes.
668 I. Beltagy et al.

Figure 1 Channel assignment algorithm running on nodes.

In the other case of accessing available interrupted

channels, they are ordered and accessed using closeness va-
lue. Source node calculates routes closeness values using
the technique in [4]. As long as all available channel are
already interrupted by PUs in other routes, use a channel
that is interrupted by a far PU because this channel is
the one with the lowest probability to be interrupted soon.
Knowing which PU is far and which PU is not is done in
the PU detection and notification phase. We do not have
access to PU’s location, but when a PU interrupts a route,
we have a rough idea how close this PU is. The closeness
between the current route and the interrupted route is used
as an estimate of how far the PU is. E.g. Fig. 2 suggests
assigning Channel1 to Route3 and Route4 because it is far

Figure 3 Channel availability forces channel switching.

from Route1 that is interrupted by a PU active on Chan-

nel1, also assigning Channel2 to Route1 and Route2.

3.5. Channel assignment analysis

One of the objectives of any channel assignment technique is to

minimize number of channel switchings. The channel switching
characteristics of this technique are

1. Channel switching in the source node. As long as differ-

ent routes are assigned to different channels, the source
node has to channel switch to send packets over differ-
Figure 2 Assign channels interrupted by far PUs. ent routes.
Channel Assignment with Closeness Multipath Routing in Cognitive Networks 669

1500 1.4
Proposed Proposed
Segment Segment
1400
1.2
1300
1
1200
Throughput

Delay (sec)
1100 0.8

1000 0.6
900
0.4
800
0.2
700

600 0
0 5 10 15 20 0 5 10 15 20
Number of PUs Number of PUs

Figure 4 Effect of varying number of PUs on throughput (kbyte/s) and delay (s).

2. If possible, no channel switching along the same route.
Channel switching along routes occurs in uncommon cases
like the one shown in Fig. 3. Channel switching in such
cases can be eliminated by equipping nodes with an extra
interface.
3. Nodes sharing multiple flows may suffer from multiple
channel switching because, depending on the list of
INTERRUPTED and RouteId of each flow, the node
may assign different channels to different flows.
experiments. PUs are assigned equally to the different chan-
nels. The results are the average of 10 simulation runs.
4.2. Simulation parameters and metrics
The protocol is simulated to study the effect of changing num-
ber of PUs, speed of PUs and total number of channels on the
system’s performance. The metrics measured are end-to-end
throughput and end-to-end delay.

4. Performance evaluation
4.1. Simulation setup
The protocol is evaluated using the network simulator (NS2).
Multichannels support is implemented as described in [16]. For
comparison, the segment-based channel assignment algorithm
from [5] is implemented. For each segment, any channel that is
not occupied by a PU is used. Both the proposed and the seg-
ment-based channel assignment algorithms are implemented
on top of the routing protocol of [4].
The topology used is a 100 m · 100 m terrain with a uni-
form random deployment of stationary SUs. PU and SU
transmission range is set to 25 m. Each node is equipped with
two interfaces, one is used for the control and the other is used
for the data. Each interface runs the IEEE 802.11a MAC layer.
The traffic generated is saturated UDP traffic. PUs are mobile
using the random way-point mobility model [6] with different
averages of speeds. Number of PUs is changed in different
4.3. Result
The following is the result of comparing the proposed channel
assignment protocol with the segment-based algorithm [5] for
changing number of PUs, number of channels and PU’s speed.
Experiments are done for all combinations of the three param-
eters. Results of changing one parameter is an average of
experiments of changing the other two parameters.
4.3.1. Varying number of PUs
Fig. 4 shows that increasing number of PUs increases routes
interruption which decreases end-to-end throughput. The de-
creased end-to-end throughput results in less contention in
MAC layer, consequently, end-to-end delay decreases for a
constant number of channels. Our channel assignment tech-
nique achieves higher throughput and lower delay than the
segment-based algorithm. However, as number of PUs in-
creases, the two techniques approaches each others because
of the significant decrease in spectrum opportunities with the
increase of number of PUs.

1600 0.7

1400 0.6

1200 0.5
Throughput

Delay (sec)

1000 0.4

800 0.3

600 0.2

400 0.1
Proposed Proposed
Segment Segment
200 0
2 3 4 5 6 7 8 2 3 4 5 6 7 8
Number of Channels Number of Channels

Figure 5 Effect of varying number of channels on throughput (kbyte/s) and delay (s).
670
890
Proposed
Segment
880
870
Throughput
860
850
840
830
0 20 40 60 80 100
PUs speed (m/s)
Figure 6 Effect of varying PUs’ speed on throughput (kbyte/s) and delay (s).
4.3.2. Varying number of channels
Fig. 5 shows that increasing number of channels increases
spectrum opportunities which increases throughput. As num-
ber of channels increases, MAC layer contention decreases,
which decreases the delay. Also, the increase in spectrum
opportunities decreases number of channel switchings, hence
end-to-end delay decreases. Our channel assignment technique
achieves higher throughput and lower delay than the segment-
based algorithm. Furthermore, the difference between our
technique and the segment-based increases with the increase
of number of channels because our technique efficiently utilizes
the extra resources.
4.3.3. Varying PUs’ speed
Fig. 6 shows the effect of changing PUs speed on the perfor-
mance. Increasing PUs speed increases links interruption rate
therefore, end-to-end throughput decreases. Also, as in case
of varying number of PUs, end-to-end delay decreases.
5. Conclusion
This work proposes a multichannel assignment technique for
multipath routing using routes closeness metric in cognitive
networks. It extended the routing protocol proposed by Bel-
tagy et al. [4] to support multiple channels. It relies on the early
detection of PUs to avoid channels that are going to be inter-
rupted. Once a SUs detects a PU, it notifies other SUs. SUs use
the channels availability information and routes closeness val-
ues to select the channel that is the least likely to be inter-
rupted. Simulation results showed the effectiveness of the
channel assignment technique in increasing end-to-end
throughput and decreasing delay. Currently, we are extending
the work in multiple directions like supporting mobile SUs and
supporting multiple interfaces.
References
[1] I. Akyildiz, W. Lee, M. Vuran, S. Mohanty, NeXt generation/
dynamic spectrum access/cognitive radio wireless networks: a
survey, Computer Networks 50 (13) (2006) 2127–2159.
[2] I.F. Akyildiz, B.F. Lo, R. Balakrishnan, Cooperative spectrum
sensing in cognitive radio networks: a survey, Physical
Communication 4 (1) (2011) 40–62.
[3] P. Bahl, R. Chandra, J. Dunagan, SSCH: slotted seeded channel
hopping for capacity improvement in IEEE 802.11 ad-hoc
wireless networks, in: Proceedings of the 10th Annual
International Conference on Mobile Computing and
Networking, ACM, 2004, pp. 216–230.
I. Beltagy et al.
0.45
Proposed
Segment
0.4
0.35
0.3
Delay (sec)
0.25
0.2
0.15
0.1
0.05
0
120 0 20 40 60 80 100 120
PUs speed (m/s)
[4] I. Beltagy, M. Youssef, M. El-Derini, A new routing metric and
protocol for multipath routing in cognitive networks, in:
Wireless Communications and Networking Conference
(WCNC), 2011 IEEE, 2011, pp. 974–979.
[5] K. Bian, J. Park, Segment-based channel assignment in cognitive
radio ad hoc networks, in: 2nd International Conference on
Cognitive Radio Oriented Wireless Networks and
Communications, 2007. CrownCom 2007, IEEE, 2007, pp. 327–335.
[6] T. Camp, J. Boleng, V. Davies, A survey of mobility models for
ad hoc network research, Wireless Communications and Mobile
Computing, WCMC 2 (5) (2002) 483–502.
[7] A. De Domenico, E.C. Strinati, M.G. Di Benedetto, A survey
on MAC strategies for cognitive radio networks,
Communications Surveys & Tutorials IEEE 14 (1) (2012) 21–44.
[8] D.B. Johnson, D.A. Maltz, J. Broch, DSR: The dynamic source
routing protocol for multi-hop wireless ad hoc networks, Ad
Hoc Networking (2001) 139–172.
[9] H. Liu, Z. Ren, C. An, Backup-route-based routing for cognitive
ad hoc networks, in: 2012 IEEE International Conference on
Oxide Materials for Electronic Engineering (OMEE), IEEE,
2012, pp. 618–622.
[10] B.F. Lo, A survey of common control channel design in cognitive
radio networks, Physical Communication 4 (1) (2011) 26–39.
[11] H. Ma, L. Zheng, X. Ma, Spectrum aware routing for multi-hop
cognitive radio networks with a single transceiver, in: 3rd
International Conference on Cognitive Radio Oriented Wireless
Networks and Communications, 2008 CrownCom 2008, IEEE,
2008, pp. 1–6
[12] N. Meghanathan, A survey on the communication protocols
and security in cognitive radio networks, International Journal
of Communication Networks and Information Security
(IJCNIS) 5 (1) (2013).
[13] I. Pefkianakis, S.H. Wong, S. Lu, SAMER: spectrum aware
mesh routing in cognitive radio networks, in: 3rd IEEE
Symposium on New Frontiers in Dynamic Spectrum Access
Networks, 2008, DySPAN, 2008, IEEE, 2008, pp. 1–5.
[14] C. Perkins, E. Royer, Ad-hoc on-demand distance vector
routing, in: Proceedings of the Second IEEE Workshop on
Mobile Computing Systems and Applications, 1999, pp.
90–100.
[15] C.W. Pyo, M. Hasegawa, Minimum weight routing based on a
common link control radio for cognitive wireless ad hoc networks,
in: Proceedings of the 2007 International Conference on Wireless
Communications and Mobile Computing, ACM, 2007, pp. 399–404.
[16] B. Wang, Multi-Channel Multi-Interface Simulation in ns2, 2006.
<http://www.cse.msu.edu/wangbo1/ns2/nshowto8.html>.
[17] C. Xin, B. Xie, C.C. Shen, A novel layered graph model for
topology formation and routing in dynamic spectrum access
networks, in: 2005 First IEEE International Symposium on New
Frontiers in Dynamic Spectrum Access Networks, 2005,
DySPAN 2005, IEEE, 2005, pp. 308–317.