Routing in Distributed Cognitive Radio Networks: A Survey

Abstract

Cognitive Radio Networks (CRNs) have been receiving significant research attention recently due to their ability to solve issues associated with spectrum congestion and underutilization. In a CRN, unlicensed users (or Secondary Users, SUs) are able to exploit and use underutilized licensed channels, but they must evacuate the channels if any interference is caused to the licensed users (or Primary Users, PUs) who own the channels. Due to the dynamicity of spectrum availability in CRNs, design of protocols and schemes at different layers of the SU’s network stack has been challenging. In this article, we focus on routing and discuss the challenges and characteristics associated with it. Subsequently, we provide an extensive survey on existing routing schemes in CRNs. Generally speaking, there are three categories of challenges, namely channel-based, host-based, and network-based. The channel-based challenges are associated with the operating environment, the host-based with the SUs, and the network-based with the network-wide SUs. Furthermore, the existing routing schemes in the literature are segregated into three broad categories based on the relationship between PUs and SUs in their investigation, namely intra-system, inter-system, and hybrid-system; and within each category, they are further categorized based on their types, namely Proactive, Reactive, Hybrid, and Adaptive Per-hop. Additionally, we present performance enhancements achieved by the existing routing schemes in CRNs. Finally, we discuss various open issues related to routing in CRNs in order to establish a foundation and to spark new interests in this research area.

Full text

1
Routing in Distributed Cognitive Radio
Networks: A Survey
Hasan A. A. Al-Rawi1, *, and Kok-Lim Alvin Yau2
Department of Computer Science and Networked System, Sunway University
No. 5 Jalan Universiti, Bandar Sunway, 46150 Petaling Jaya, Selangor, Malaysia
Tel: +6 03 7491 8622 Ext. 3216
Fax: +6 03 5635 8633
Email: 08011843@imail.sunway.edu.my1, *; koklimy@sunway.edu.my2
*Correspondence Author
Abstract Cognitive Radio Network (CRN) has been receiving significant research attention
recently due to its ability to solve issues associated with spectrum congestion and underutilization.
In CRN, unlicensed users (or Secondary Users, SUs) are able to exploit and use underutilized
licensed channels, but they must evacuate the channels if any interference is caused to the licensed
users (or Primary Users, PUs) who own the channels. Due to the dynamicity of spectrum
availability in CRNs, design of protocols and schemes at different layers of the SU’s network stack
has been challenging. In this article, we focus on routing and discuss the challenges and
characteristics associated with it. Subsequently, we provide an extensive survey on existing routing
schemes in CRNs. Generally speaking, there are three categories of challenges, namely channel-
based, host-based, and network-based. The channel-based challenges are associated with the
operating environment, the host-based with the SUs, and the network-based with the network-wide
SUs. Furthermore, the existing routing schemes in the literature are segregated into three broad
categories based on the relationship between PUs and SUs in their investigation, namely intra-
system, inter-system, and hybrid-system; and within each category, they are further categorized
based on their types, namely Proactive, Reactive, Hybrid, and Adaptive Per-hop. Additionally, we
present performance enhancements achieved by the existing routing schemes in CRNs. Finally, we
discuss various open issues related to routing in CRNs in order to establish a foundation and to
spark new interests in this research area.
Keywords Cognitive radio . Routing . Route discovery . Route selection . Route
maintenance
1 Introduction
With the considerable growth in wireless applications, the need for more radio
spectrum has been increasing explosively due to the increasing demand of
bandwidth [1]. Generally speaking, the traditional static spectrum allocation

2
policy has been partitioning the radio spectrum into unlicensed or licensed bands.
The unlicensed bands, such as Industrial, Scientific and Medical (ISM) band, have
been widely utilized, and so congestion has been a concern. Popular wireless
communication systems, including IEEE802.11 [2] and IEEE802.15.4 [3], have
been operating over unlicensed bands. On the other hand, the licensed bands have
been exclusively allocated to licensed users in a static manner, and it has been
shown to be underutilized by Federal Communications Commission (FCC) [4, 5].
Generally speaking, FCC has shown that, at any time and location, the utilization
of the licensed spectrum varies between 15% and 85% [5, 6].
Cognitive Radio (CR) has been proposed as the next generation wireless access
technology that addresses the problem of underutilization in licensed spectrum. In
CR Networks (CRNs), there are licensed users (or Primary Users, PUs) and
unlicensed users (or Secondary Users, SUs). PUs are the rightful radio spectrum
licensees, while SUs, which are equipped with CR capabilities, exploit the
underutilized portions in licensed spectrum (or white spaces) while minimizing
interference to PUs [6-8]. Due to the dynamic nature of PUs’ activities, SUs must
change their operating channels when the PUs’ activities reappear, or the
interference to PUs becomes pronounced [9-12].
Similar to the traditional wireless networks, CRNs are established with or
without fixed network infrastructure [13]. In the former one, which is also called
centralized networks, a SU Base Station (BS) provides single-hop connections to
SU hosts. On the other hand, in the latter one, which is also called distributed
networks, SUs communicate with each other in an ad-hoc manner over multi-hop
connections without the need of fixed network infrastructure. Fig. 1 shows a
distributed CRN, which is the focus of this article. The SUs communicate with
each other in a distributed manner as well as access a SU BS in a multi-hop
manner. Examples of the application of distributed CRNs are disaster recovery
and rescue operation scenarios, as well as the extension of broadband service to
rural communities.
In distributed CRNs, routing algorithms are important to find a route between a
source SU and a destination SU. Since SUs must be adaptive to the dynamic
changes in spectrum utilization by PUs, routing in CRNs has been challenging and
it differs from the routing in traditional wireless networks, particularly multi-
channel networks in which static sets of channels are available for the nodes [14,

3
15]. In CRNs, SUs may have different sets of channels that impose more routing
challenges especially in multi-hop communication [11, 16]. For this reason,
routing must be spectrum aware in CRNs, and thus routing schemes for traditional
wireless networks cannot be readily applied to CRNs [12-14, 17-19].
We provide a scenario to show the necessity of a spectrum-aware routing. Fig.
1 shows a scenario of CRN co-located with three PU networks. Suppose, SU (A)
wants to establish a route to SU BS. Using a traditional routing algorithm may
provide a route with minimum number of hops or intermediate SUs (A-C-E-G) to
the SU BS. However, this may provide poor network performance because the
route passes through the three PU BSs and their hosts (B, D, F, H) resulting in
high interference to the PUs. On the other hand, CR-based spectrum-aware
routing may provide a route with higher number of hops (A-C-I-K-L) that
generates less interference to the PUs and their hosts (B, D, J), and this enhances
the SU end-to-end performance [20].
This article provides an extensive survey on existing routing schemes in CRNs.
The purposes are to establish a foundation and to spark new interests in this
research area. Our contributions are as follows. Section 2 presents challenges of
routing in CRNs. Section 3 presents the characteristics of routing schemes in
traditional and CR-based networks. Section 4 presents existing routing schemes in
CRNs, as well as to relate them to the challenges (see section 2) and routing
characteristics (see section 3). Section 5 presents performance enhancement
Fig. 1 CRN scenario

4
achieved by the routing schemes. Section 6 presents open issues. Section 7
presents conclusions.
2 Challenges of Routing in Cognitive Radio
Networks
This section discusses the challenges of routing brought about by CR. Fig. 2
shows the taxonomy of the challenges. Generally speaking, there are three
categories of challenges, namely, channel-based, host-based, and network-based.
The channel-based challenges are associated with the operating environment, the
host-based with the SUs, and the network-based with the network-wide SUs. We
assign each challenge a label, for instance, the challenge Dynamicity of Channel
Availability is assigned a label C(1.1).
2.1 Channel-based Challenges
This section presents four types of channel-based challenges.
C1.1 Dynamicity of Channel Availability. The availability of the channel
for data transmission is dynamic, and it is dependent on the SU’s physical
location, and channel utilization by PUs. A low PU Utilization Level
Fig. 2 Taxonomy of the routing challenges in CRNs

5
(PUL) provides higher level of channel availability to SUs, and vice-versa.
This also means that the number of channels available at each SU is
dynamic [17]. Channel switches may be necessary. As a consequence, the
need to perform routing may increase due to higher occurrence of link
failure, resulting in higher amount of routing overhead.
C1.2 Diversity of Operating Channels. A link between a SU node pair is
established only if there exists at least one channel in common between the
two SUs [21]. Channels with different center frequencies provide different
levels of data rates and transmission ranges, which in turn affect route
selection [22]. As a consequence, QoS provisioning is challenging because
intermediate SUs may switch to a new operating channel with different
level of data rate and transmission range due to dynamicity of channel
availability (see challenge C(1.1)).
C1.3 Lack of a Fixed Common Control Channel. There are two types of
channels, namely Common Control Channel (CCC) and data channel. The
SUs exchange routing control packets, such as Route Request (RREQ) and
Route Reply (RREP), in CCC; and send data packets in data channels. Due
to the Dynamicity of Channel Availability (see challenge C(1.1)) and the
requirement to vacate a channel once PU becomes active, the assumption
of the availability of a static CCC, which has been widely adopted in the
literature, for routing control packet exchange may be impractical [23-26].
Furthermore, the use of CCC to select routes may not reflect the
characteristics of the actual route [27]. As a consequence, designing
routing schemes in CRNs without relying on a fixed CCC may be
necessary to provide feasibility of implementation.
C1.4 Integration of Route Discovery with Channel Decision. Due to the
Dynamicity of Channel Availability (see challenge C(1.1)), selecting a
route without channel information may not be optimum. For example, a
stable end-to-end performance may need a route that provides stable links
over the intermediate SUs in order to minimize interruptions from link
breakage. As a consequence, route selection should be based on channel
information obtained prior to route selection [19, 23, 24, 28, 29].

6
2.2 Host-based Challenges
This section presents three types of network-based challenges.
C2.1 Minimizing Channel Switch Delay and Backoff Delay. Upon
detection of PU activities, a SU incurs a channel switch delay if it chooses
to switch its channel, or it incurs a backoff delay while waiting for the PU
to relinquish its channel if it chooses to remain in its current channel.
Additionally, load balancing among channels, and hence channel switches,
may help to minimize channel contention among SUs [21, 28], and so
channel switch delays are inevitable. As a consequence, routing schemes
should cater for the minimizing both channel switch and backoff delays as
these delays affect the routing performance.
C2.2 Multiple Transmissions for Each Broadcast/Multicast.
Broadcasting/multicasting is commonplace in routing for control message
exchange. Due to the Diversity of Operating Channels (see challenge
C(1.2)), each SU may use different available channels; hence a single
broadcast in a particular channel may not reach all neighbor SUs. Multiple
transmissions in different channels for each broadcast/multicast may be
necessary [30]. As a consequence, the number of transmissions and
bandwidth requirement increase with the number of available channels for
each broadcast. Therefore, higher number of available channels may have
no contribution towards throughput performance [28]. Alternatively, a
synchronization window, which is a fixed time duration when all SUs are
tuned to a particular channel for broadcasting, may be applied; however,
time synchronization or centralized clocking is necessary [31]. As a
consequence, routing schemes should reduce the number of transmission
for each broadcast/multicast in order to reduce the number of channel
switches and bandwidth consumption [30].
C2.3 Heterogeneity of SUs. A CRN may consist of SUs with different
capabilities such as transmission power and processing speed [32]. For
instance, an intermediate SU with limited capability may become a
bottleneck and this degrades the end-to-end performance. As a
consequence, routing schemes should be aware of the heterogeneity of
SUs.

7
C2.4 Mobility of SUs. Due to the Dynamicity of Channel Availability (see
challenge C(1.1)), each physical location may have different levels of
white spaces for a particular channel. Therefore, higher mobility of SUs
may reduce channel access time, and hence, it increases the number of
channel switches. Due to the increased amount of overhead as a result of
rerouting, this may increase energy consumption [33]. Fast and
unpredictable movements of SUs make it more challenging to guarantee
QoS requirements as well as to minimize interference to PUs. As shown in
[34] when the PUL is static and the movement speed of SUs is increasing,
more channel switches are performed compared to the opposite scenario.
This shows that mobility of SUs may have greater effects on the CRN
performance compared to the PU activities. As a consequence, routing
schemes should be aware of the mobility of SUs.
2.3 Network-based Challenges
This section presents three types of network-based challenges.
C3.1 Tradeoff between Number of Hops and Network-wide
Performance. A route with lower number of hops (or with longer
transmission range among the SUs) has three effects. Firstly, long
propagation distance among the SUs may increase interference to PUs.
Secondly, link failures may be more frequent, and this incurs higher route
maintenance cost [35]. Thirdly, it incurs higher energy consumption due
to longer transmission range [36]. Suitable number of hops is necessary to
improve end-to-end route performance (e.g. lower delay, higher
throughput and lower energy consumption) [37]. As a consequence, route
selection schemes should take into account the number of hops in order to
achieve the desired network performance.
C3.2 Network-wide Energy Consumption. The intrinsic characteristics of
CRNs to address various challenges (i.e. channel switches in challenges
C(1.1) and C(2.1), as well as distinctive transmission power in various
channels in challenge C(1.2)) add more challenges to the design of energy-
aware routing schemes in CRNs. Hou et al. [38] discuss some of these
challenges in each routing phase:

8
o Route discovery: Due to the Multiple Transmissions for Each
Broadcast/Multicast (see challenge C(2.2)), flooding of RREQ
control messages incurs energy consumption. For instance,
flooding of RREQ to all SUs in reactive routing requires
intermediate SUs to rebroadcast the packet even though these SUs
may not be chosen as part of the route, and this increases energy
consumption, which is pronounced as a result of challenge C(2.2).
o Route selection: Routing control packets have been widely
transferred over the common control channel, which is accessed by
all SUs. The transfer of these control packets may not provide an
accurate estimation on energy consumption for actual data packet
transmission in data channels, and hence the route selection may
not provide energy load balancing among the SUs. In addition,
route selection in energy-aware CRNs should be based on the
specific network requirements. Challenge C(3.1) also provides
insight on how to reduce energy consumption through the
reduction of transmission range.
o Route maintenance: To maintain routes and to recover broken
routes, SUs need to periodically check their respective link’s status.
This period is inversely proportional to energy consumption of the
SU nodes and proportional to the downtime of the links. In
addition, due to the Dynamicity of Channel Availability (see
C(1.1)) in CRNs, route maintenance is much required and thus
more energy is required to recover broken routes.
C3.3 Fast and Spectrum-Adaptive Route Recovery. In addition to node
mobility in the traditional wireless networks, fast route recovery in CRNs
is essential due to the frequent channel switches performed by SUs when
the PUs activities are detected as a result of the Dynamicity of Channel
Availability (see challenge C(1.1)) [39]. Additionally, routing schemes
should be able to tackle different routing failure events (e.g. channel
switch vs. SU failure), and to reduce route maintenance cost [6, 35].

9
3 Characteristics of Routing in Cognitive Radio
Networks
This section discusses the characteristics of routing found in traditional and CR-
based networks. Fig. 3 shows the taxonomy of the characteristics of routing
schemes in CRNs. Generally speaking, there are two categories, namely CR-
specific and traditional-based. The former category describes the features brought
about by CR, while the latter one describes the features found in the conventional
routing schemes in traditional wireless networks. Further investigation is
necessary to address the new open issues associated with CR-specific
characteristics.
3.1 CR-Specific Characteristics
This section presents three types of CR-specific characteristics.
3.1.1 SU-PU System Investigations
Based on the assumptions adopted by various investigations in the literature,
research into routing schemes for CRNs can be segregated into three categories
based on the performance achieved by the schemes as follows [40]:
Fig. 3 Taxonomy of the characteristics of routing in cognitive radio networks

10
R1.1 Intra-system. From the perspective of CRN itself, the routing scheme
chooses a route that provides the best-known end-to-end network
performance for SUs.
R1.2 Inter-system. From the perspective of system co-existence with PUs, the
routing scheme chooses a route that minimizes interference to the PUs.
Hence, investigation into inter-system provides results on the effects of
SU interference to the PUs.
R1.3 Hybrid-system. The routing scheme combines both intra-system and
inter-system approaches.
3.1.2 Levels of Channel Utilization by PUs
There are three types of channels categorized by their level of channel utilization
by PUs as follows [31]:
R2.1 Static channel. The licensed channels have relatively low level of
dynamicity of channel availability (see challenge C(1.1)), and once it is
available, it remains usable to SU for hours to days. Specifically, a
channel is available for data transmission for duration longer than the
communication time. Hence, from the SU’s perspective, a channel can be
exploited for an unlimited period of time. Traditional routing schemes for
multi-hop multichannel mesh networks can be used [31].
R2.2 Dynamic channel. The licensed channels have moderate level of
dynamicity of channel availability (see challenge C(1.1)). Thus, SUs may
have to switch their operating channels from time to time. Per-hop
routing (see routing characteristics R(4.4)) may be preferable since the
route may not be available for the entire communication time, and
establishing new routes from time to time increases routing overhead and
end-to-end delay.
R2.3 Opportunistic channel. The licensed channels have very high level of
dynamicity of channel availability (see challenge C(1.1)). Specifically,
the channel is available for data transmission for a duration that is highly
likely to be less than the communication time. Thus, the SUs may have to
actively switch their respective operating channels for most of the times.

11
3.2 Traditional-based Characteristics
This section presents four types of CR-specific characteristics.
3.2.1 Modes of Communication
Two modes of communication are unicast and multicast, and they are described as
follows:
R3.1 Unicast. A message sent by a SU source node is only received by its
intended SU destination node.
R3.2 Multicast. A message sent by a SU source node is received by all SU
destination nodes that belong to the same group (or multicast tree).
Multicasting in CRNs may be more complex than the traditional
networks due to the Diversity of Operating Channels (see challenge
C(1.2)).
3.2.2 Routing Types
The characteristics of routing types are described as follows:
R4.1 Proactive. In proactive (or table-driven) routing, each SU source node
exchanges routing packets with neighbor nodes, and keeps track of each
route in a routing table. An example of a traditional proactive routing
scheme is Optimized Link State Routing (OLSR) [41]. An advantage of
proactive routing is that, routes are updated periodically, and so it
provides up-to-date routing information, which helps to reduce end-to-
end delay. A disadvantage is that, it increases bandwidth consumption
and network overhead. The performance of a proactive routing scheme
mainly depends on the network scenario, such as network size and SU
mobility. Larger network size and higher SU mobility may increase
network overhead.
R4.2 Reactive. In reactive (or on-demand) routing, a SU source node maintains
routing information of destination SUs to which it has packets to send.
Basically, A SU floods the network with RREQ control packets. Upon
receiving the RREQ packet, the SU destination node responds with a
Route Reply (RREP) control packet. An example of a traditional reactive
routing scheme is Ad hoc On-demand Distance Vector (AODV) [42]. An
advantage of reactive routing is that, routes to a particular destination SU

12
are discovered only when the SU source node has packets to be sent to the
destination SU, and so it reduces bandwidth consumption and network
overhead. A disadvantage is that, reactive routing incurs higher delay in
route discovery.
R4.3 Hybrid. The hybrid approach combines the characteristics of both
proactive and reactive routing schemes. It achieves a balanced
performance tradeoff between proactive and reactive routing schemes in
various network scenarios with different requirements. For instance, a
clustering scheme (see routing characteristic R(6.3)) may use proactive
routing for intra-cluster communication, and reactive routing for inter-
cluster communication. An example of a traditional hybrid routing
scheme that adopts this clustering concept is Zone Routing Protocol
(ZRP) [43].
R4.4 Adaptive Per-hop. In adaptive per-hop routing, a SU chooses a next-hop
node to a destination SU node based on its local information. The SU may
adapt its next-hop selection according to the local characteristics of
network and channels, such as dynamicity levels of channel availability
(see challenge C(1.1)). Machine learning algorithms, such as
Reinforcement Learning (RL) [44, 45] enable routing schemes to select
next-hop nodes based on local information. Unlike the proactive and
reactive approaches, exchanges of routing packets involve a SU and its
neighbor SUs. Hence, the main advantage of this type of routing is that, it
is a localized approach, which can further reduce bandwidth consumption
and network overhead introduced by flooding.
3.2.3 Routing Models
A routing model in distributed CRNs can be either centralized or distributed as
follows:
R5.1 Centralized. In the centralized approach, when a SU joins or leaves its
neighbor SUs, the neighbors must update routing information with a
central entity that performs and distributes routing decision to the entire
CRN.
R5.2 Distributed. In the distributed approach, when a SU joins or leaves its
neighbors, only one-hop or two-hop neighbors of the SU update their

13
routing information. The SUs exchange routing packets with their
respective one-hop or two-hop neighbors. Therefore, the amount of
routing overhead is lower compared to the centralized approach.
3.2.4 Other Routing Features
This section presents four features that can be adopted by the routing schemes.
R6.1 Multipath. The SU source node discovers more than one (joint/disjoint)
route to the SU destination node in order to enhance route reliability and
achieve load balancing among the selected routes [12].
R6.2 Geographical. A route is selected based on the SU physical location.
Positioning data is usually obtained using Global Positioning System
(GPS) [46] embedded at each SU. This type of routing may reduce
network-wide bandwidth consumption by controlling flooding of the
routing control packets; however, it incurs additional hardware cost and
energy consumption, as well as has limited operation of GPS in indoor
environment.
R6.3 Hierarchical/Clustering. Clustering schemes organize SUs into groups.
Each group is comprised of a clusterhead, and the rest are member nodes.
The member nodes forward data packets to their respective clusterhead,
which in turn forwards these packets until they reach a gateway. The
objective of a clustering scheme is to reduce routing control packets
although it may incur clustering control packets for cluster formation and
maintenance.
R6.4 Flow level approach. Two types of flow interferences along a route are
intra-flow and inter-flow interferences. The former one occurs on links
that belong to the same data flow (or neighboring SUs of the same route),
whereas the latter one occurs on links that belong to different data flows
(or neighboring SUs of different routes) [22, 47]. The interferences may
degrade the end-to-end performance (e.g. higher delay and lower
throughput) due to collisions found in the shared wireless medium. A
joint routing and channel selection scheme has been applied to reduce
these interferences.

14
4 Routing Schemes in Cognitive Radio Networks
This section presents existing work on routing schemes in CRNs. Summaries of
intra-system-based routing schemes, as well as inter-system-based and hybrid-
system-based routing schemes are presented in Tables 1 and 2, respectively. The
rest of this section is organized as follows. Section 4.1, 4.2 and 4.3 present intra-
system, inter-system, and hybrid-system routing schemes. Other routing schemes
are also presented. Within Section 4.1, it further organizes the section according to
routing types as defined in Section 3.2.2, namely proactive R(4.1), reactive R(4.2),
hybrid R(4.3), and adaptive per-hop R(4.4); while Sections 4.2 and 4.3 follow the
similar organization.
4.1 Intra-system Routing Schemes
This section discusses various kinds of intra-system routing schemes (see routing
characteristic R(1.1)).
4.1.1 Proactive Routing Schemes
The proactive routing schemes (see routing characteristic R(4.1)) have been
shown to achieve performance enhancements, particularly lower end-to-end delay
and higher throughput performance (see Table 3).
Xie and Xi [30] propose a multicast (see routing characteristic R(3.2)) routing
scheme called Core Based Bottom Up (CBBU) combined with a channel selection
in order to minimize bandwidth consumption and the number of transmissions for
various channels for each broadcast. This routing scheme addresses challenges
C(2.2), and C(2.4). Assuming static or low mobility SUs, the CBBU constructs a
multicast tree for each SU in a group of multicast SUs; while the channel selection
mechanism chooses a channel that reduces transmission time.
Chen-li et al. [25] propose a tree-based algorithm with channel selection
mechanism. This algorithm addresses challenges C(1.3) and C(1.4). It uses
statistic-based metric and it eliminates the need for a common control channel.
The routing tree is formed using statistic-based metric, which assigns priorities to
channels based on the number of SUs already allocated on these channels.
Subsequently, channel selection is performed based on the channel priority in
order to minimize interference to other SUs. Higher priority indicates lower

15
number of SUs using a particular channel. The algorithm increases throughput by
achieving load balancing among the available channels at different SUs in order to
increase number of simultaneous transmissions in multiple channels.
Tuggle [48] propose an architecture for multipath routing (see routing
characteristic R(6.1)). This routing scheme addresses challenge C(1.1). It
establishes multiple disjoint routes to support mission critical applications, which
are sensitive to end-to-end delay and packet loss. Each packet is duplicated at the
SU source node and forwarded to each disjoint route in order to maximize packet
delivery rate. At the SU destination node, only a single packet is forwarded to
upper layers.
Yun et al. [49] propose a routing metric to minimize end-to-end delay. This
routing scheme addresses challenge C(1.1). Firstly, it assigns weight in terms of
estimated transmission delay at each link, which takes into account channel
bandwidth, probability of channel availability, and the number of available
channels at each link. Secondly, it estimates minimum end-to-end delay along the
route using Dijkstra algorithm [50].
Xin et al. [26] propose a joint routing and channel selection scheme, which
maximizes network throughput. This routing scheme addresses challenges C(1.4),
and C(2.1), and it assumes static channels (see routing characteristic R(2.1)). Each
SU constructs a multi-layer graph to model the entire CRN in order to facilitate
route discovery and channel selection along the route. Each layer represents
connections using a single channel; while each node in the graph represents a
single SU node. Each horizontal edge represents a connection within a layer using
a particular channel. Each vertical edge represents a channel switch, and so the
number of vertical edges represents the number of operating channels. Next, a
traditional routing algorithm, such as Dijkstra’s algorithm [50], is run over the
graph in order to find the least cost route, which is computed using the horizontal
and vertical edges. As part of the cost computation, a horizontal edge represents
the traffic load of a link and SU-SU interference, while a vertical edge represents
switching delay between channels. Lastly, channel selection is invoked on the
selected route to achieve a balanced tradeoff between switching delay and traffic
loads. Specifically, when the traffic load is high (or low), it is necessary to reduce
channel contention (or channel switching delay), and so an intersecting SU node

16
may switch and use more channels (or choose to stay on the same channel) to
maximize throughput.
Wang and Aceves [51] propose a joint multipath routing (see routing
characteristic R6.1), scheduling and channel selection scheme to maximize
network throughput and packet delivery rate, as well as to reduce end-to-end
delay. This routing scheme addresses challenges C(1.2), and C(1.4). Scheduling
and channel selection, which takes into account two-hop SU neighbors, use a
heuristic algorithm. This enables a SU to choose its channel based on traffic
demand in order to provide load-balancing among different channels and links.
Since SUs may operate on different channels, packets are scheduled to each
neighbor SU individually, and the scheduling scheme assigns different time slots
for different links in order to maximize channel reuse. Furthermore, a metric is
used to calculate the efficiency of scheduling and channel selection, and then it is
used to calculate routing metric. Specifically, using a proactive routing protocol,
the efficiency metric is disseminated among neighbor SUs. Subsequently, the SU
source node uses the efficiency metric to compute multiple routes, and it chooses
multiple routes with high throughput to the SU destination node.

17
Table 1 A summary of intra-system-based routing schemes for addressing routing challenges
Intra-system
References
Channel-based
Host-based
Network-based
C1.1 Dynamicity of channel
availability
C1.2 Diversity of operating channels
C1.3 Lack of a fixed CCC
C1.4 Integration of route discovery
with channel decision
C2.1 Minimizing channel switch delay
or backoff delay
C2.2 Multiple transmissions for each
broadcast/multicast
C2.3 Heterogeneity of SUs
C2.4 Mobility of SUs
C3.1 Tradeoff between number of
hops and network-wide performance
C3.2 Network-wide energy
consumption
C3.3 Fast and spectrum-adaptive
route recovery
Proactive
Chen-li et al. (2009) [25]
×
×
Tuggle (2010) [48]; Yun et al. (2010)
[49]
×
Wang and Aceves [51]
×
×
Xie and Xi (2011) [30]
×
×
Xin et al. (2008) [26]
×
×
Reactive
Almasaeid et al (2010) [20]
×
×
Cheng et al. (2007) [29]
×
×
×
Ding and Xiao (2010) [61]; Shih and
Liao (2010) [9]
×
×
Jia et al. (2009) [59]; Li et al. (2009) [14]
×
Hincapie et al. (2008) [22]; How et al.
(2010) [60]
×
×
Huang et al. (2011) [23]
×
×
×
×
×
×
Kamruzzaman et al. (2011) [54]
×
×
Ma et al. (2008) [28]; Song et al. (2009)
[52]
×
×
×
Qin et al. (2009) [62]
×
×
×
Song and Lin (2009) [57]; Zeeshan et al.
(2010) [39]
×
×
×
Wang and Huang (2010) [47]
×
×
×
Zheng et al. (2011) [56]
×
Hybrid
Han and Huang (2010) [64]
×
×
Khalife et al.(2008) [65]
×
×
Zhu et al. (2008) [24]
×
×
×
×
×
Adaptive Per-hop
Badoi et al. (2010) [69]; Kim et al.
(2011) [77]
×
×
Pan et al. (2008) [78]
×
×
Ding et al. (2010) [17]; Jashni et al.
(2010) [71]; Lin and Chen (2010) [74];
Pefkianakis et al. (2008) [73]; Talay and
Altilar (2009) [68]
×
Soltani and Mutka (2011) [76]; Xia et al.
(2009) [66]
×
×
Other Routing Schemes
Bütün et al. (2010) [81]
×
×
Chen et al. (2011) [83]; Huang et al.
(2011) [33]
×
Filippini et al. (2009) [35]
×
×
Gao et al. (2011) [80]
×
×
Hu et al. (2007) [79]
×
×
Wen and Liao (2010) [84]
×
×
×

18
4.1.2 Reactive Routing Schemes
The reactive routing schemes (see routing characteristic R(4.2)) have been shown
to achieve performance enhancements, particularly lower end-to-end delay, lower
amount of control overhead, and higher throughput performance (see Table 3).
Cheng et al. [21, 29] incorporate a flow-level channel selection scheme into
reactive routing (see routing characteristic R(4.2)). This routing scheme addresses
challenges C(1.1), C(1.2) and C(2.1). It minimizes end-to-end delay of each flow
through the reduction of two kinds of delays, namely channel switching delay for
each flow along its route, and backoff delay for intersecting flows. The channel
selection scheme provides an appropriate tradeoff for the two kinds of delays.
Specifically, assigning an unused available channel to a new flow incurs channel
switching delay when a SU switches between flows; while assigning an assigned
available channel to a new flow incurs backoff delay. For intersecting SU nodes, a
Table 2 A summary of inter-system-based and hybrid-system-based routing schemes for
addressing routing challenges
References
Channel-based
Host-based
Network-based
C1.1 Dynamicity of channel
availability
C1.2 Diversity of operating channels
C1.3 Lack of a fixed CCC
C1.4 Integration of route discovery
with channel decision
C2.1 Minimizing channel switch delay
or backoff delay
C2.2 Multiple transmissions for each
broadcast/multicast
C2.3 Heterogeneity of SUs
C2.4 Mobility of SUs
C3.1 Tradeoff between Number of
Hops and Network-wide Performance
C3.2 Network-wide energy
consumption
C3.3 Fast and spectrum-adaptive
route recovery
Inter-system
Other
Routing
Metrics
Yuan et al. (2010) [85]
×
Hybrid-system
Proactive
Guan et al. (2010) [86]
×
×
×
Reactive
Chowdhury and Akyildiz
(2011) [37]
×
×
×
×
×
Chowdhury and Di Felice
(2009) [27]
×
×
×
×
Abbagnale et al. (2011) [11, 87]
×
×
×
Adaptive
Per-hop
Zhu et al. (2010) [88]
×
Other
Routing
Schemes
Xie et al. (2010) [36]
×
×

19
flow scheduling scheme is proposed to hierarchically group flows according to
their operating channels, and it serves these flows in a round robin fashion. SU
nodes exchange delay cost (i.e. switching, backoff, and queuing delays)
information among neighbor nodes in order to discover and choose better routes,
as well as to negotiate re-routing decisions with a chosen neighbor SU node. An
intersecting SU reduces the total delay cost while establishing route for an
incoming flow; and subsequently it decides whether to accept the flow or to re-
route it through a neighbor SU. For instance, since an intersecting SU node may
increase its channel switches in order to reduce channel contention during
congestion, the proposed routing scheme re-routes some of its flows to other
neighbor SUs in order to achieve load balancing and reduce channel switching
delay for different flows.
Ma et al. [28] incorporate a flow-level channel selection scheme into a
spectrum-aware reactive routing scheme. This routing scheme addresses
challenges C(1.1), C(1.3) and C(2.1). It maximizes the throughput of each flow
through achieving a balanced tradeoff between two kinds of delays, namely
channel switching delay and backoff delay. Based on a delay analysis, channel
selection to enhance the channel’s utilization is achieved. After a route to the
destination has been discovered, channel selection is performed. An intermediate
SU node is aware of the channel selection made by SUs between itself and the
destination, thus it selects a channel for its link towards the source SU that
increases the available time for data transmission in order to reduce switching
delay. Additionally, the channel selection also limits the opportunity of channel
switching to a single SU among its SU neighbors in order to cater for the deafness
problem in which there is lack of coordination among two neighbor SUs such that
they operate in different channels at most of the time.
Song et al. [52] incorporate Swarm Intelligence (SI) [53] into a traditional
reactive routing scheme, namely AODV [42] in order to minimize end-to-end
delay. This routing scheme addresses challenges C(1.1), C(1.3) and C(2.1).
Swarm Intelligence (SI) is [53] a distributed approach in which SUs communicate
locally in order to enhance network-wide performance. For a particular route
request, a certain number of RREQ control packets (or ants) are sent by the SU
source node, and each SU intermediate node tunes to each available channel
repeatedly so that it can receive the RREQ packet on the channel in which it wants

20
to receive data packets. The routing scheme computes a routing metric using
channel switching delay, transmission delay and backoff delay for a particular
route.
Kamruzzaman et al. [54] propose a joint routing and timeslot-based channel
selection scheme, which assigns traffic over timeslots of different channels along
a route in order to maximize network throughput. This routing scheme addresses
challenges C(3.1) and C(3.2). It improves load balancing and bandwidth
availability, as well as ameliorates collisions among different flows using
synchronized channel access. Furthermore, route selection is based on SUs’
residential energy and minimum number of hops in order to reduce energy
consumption and increases network-wide lifetime.
Li et al. [14] propose a routing mechanism inspired by Swarm Intelligence (SI)
[53] and RL [44, 45] in order to minimize end-to-end delay in highly dynamic or
opportunistic channels (see routing characteristic R(2.3)). This routing scheme
addresses challenge C(1.1). There are two phases, Firstly, ants, which are called
F-ants, are used to discover white spaces along a route using a common control
channel. Secondly, another kind of ants, which are called B-ants, are sent along
the reverse route to the SU source node in data channel in order to collect
information about the network environment, as well as to update the routing tables
of the intermediate SU nodes. The RL approach assigns rewards to routes based
on their channel quality in order to increase convergence rate of selecting an
optimal route.
Shih and Liao [9] propose a joint routing and channel selection scheme in order
to enhance route robustness and maximize end-to-end throughput. The robustness
of a route is defined by interference from the PU, so a route with high robustness
experiences low PU interference. This routing scheme addresses challenges C(1.1)
and C(1.4). Using integer linear programming [55] approach, it takes into
consideration the heterogeneity and dynamicity of channel availability. Firstly, it
selects the most robust channels on each link along a route such that the route
guarantees at least a minimum threshold of robustness. Secondly, it allocates a
channel to each link along the route in order to maximize end-to-end throughput.
Zheng et al. [56] propose a channel selection scheme in order to reduce number
of channel switches along a route. This routing scheme addresses challenge
C(2.1). The routing metric selects a set of common channels that can cover the

21
maximum number of SUs along a route, thus reducing the number of channel
switches required to transmit a packet along the route.
Song and Lin [57] propose a multipath-based routing (see routing characteristic
R(6.1)) scheme. It selects multiple routes at the discovery phase in order to
enhance route robustness and route recovery. This routing scheme addresses
challenges C(1.1), C(1.4) and C(3.3). Firstly, all possible routes to the SU
destination node are discovered. Secondly, the most robust route, in terms of
channel stability time and channel switching delay, is selected as the main route.
Channel stability time, which is measured based on PUL, is defined as the time
period in which the channel is available to the SU. Thirdly, a backup route is
selected, which has high degree of divergence with the main route. The degree of
divergence between the main and backup routes is based on the number of hops
and route stability. This ensures that the two routes are disjoint and share no or
minimum SU intermediate nodes.
Almasaeid et al. [20] propose a joint multicast and channel assignment
algorithm, which takes into account the effects of channel heterogeneity and
switching delay in multicasting (see routing characteristic R(3.2)), to improve
end-to-end delay. This routing scheme addresses challenges C(2.1) and C(2.2). A
distributed channel selection scheme for multicast trees is performed using
dynamic programming [58] in two phases, namely forward and reverse phases.
Firstly, the forward phase finds the optimal route from the SU destination node to
the SU source node. Secondly, the reverse phase allocates channels along the
route. Furthermore, to minimize channel switching delay in the presence of
multiple multicast flows, it allocates channels that are closer to the already
assigned ones to existing flows.
Jia et al. [59] propose a routing scheme that improves spectrum utilization
using SU relay nodes. This routing scheme addresses challenge C(1.1). In addition
to the main link between two SUs, a SU relay node bridges the communication
between these SUs using a different set of channels, thus improving the overall
throughput. An Orthogonal Frequency-Division Multiplexing (OFDM) based
technique is applied to allow concurrent transmissions in multiple channels. A
routing metric, which takes into account channel availability, as well as its
conditions in terms of packet loss, utilization and candidate relay SUs, is applied
in route selection. Furthermore, proactive adjustments on channel and relay SU

22
selections are made with two purposes. Firstly, it improves link throughput by
reducing the effects of PU activities. Secondly, it does not assign channels, which
have been used by other neighboring SUs, to new nodes.
How et al. [60] perform route selection using multiple metrics in order to
improve route robustness and minimize end-to-end delay for various traffic
classes with different QoS requirements. This routing scheme addresses
challenges C(1.1) and C(3.1). It ensures that communication time of a flow is less
than the route lifetime. The route lifetime, which can be interrupted by the PU
activities, is based on the channels availability, as well as channel switching and
queuing delays. In addition, in order to support QoS, it controls the transmission
power and selects the nearest forwarding SUs to the SU destination node.
Zeeshan et al. [39] propose a joint channel selection and routing scheme in
order to enhance route robustness and network capacity. This routing scheme
addresses challenges C(1.1), C(1.4) and C(3.3). Each SU makes use of channel
diversity to select multiple channels locally for transmission and backup. A
cooperative scheme among neighbor SUs is proposed whereby the SUs switch to a
common channel periodically to share spectrum backup information.
Consequently, when a channel fails due to external conditions such as when PU
activities reappear, neighboring SUs switch to the same backup channel in a
coordinated manner, which allows them to re-establish their communication and
minimize global re-routing due to the channel failure.
Hincapie et al. [22] propose a routing scheme comprised of route selection,
channel allocation and scheduling in order to satisfy end-to-end bandwidth
requirement. This routing scheme addresses challenges C(1.1) and C(3.1). The
scheme is formulated as an optimization problem, and it is solved using integer
linear programming [55]. During the discovery phase, in order to achieve higher
possibility of fulfilling a bandwidth request, the scheme establishes multiple
routes with minimum hop-count consideration. An intermediate SU node
rebroadcasts received RREQ from its neighbor SUs only if its channel selection
algorithm indicates that the link to its next hop has sufficient bandwidth. There are
two heuristic algorithms for channel selection. The first algorithm selects channel
that has approximately similar and sufficient amount of residual bandwidth to the
requested bandwidth requirement. The second algorithm selects channel that has
high amount of residual bandwidth, and is less likely to be chosen by the neighbor

23
SUs in order to minimize intra-flow and inter-flow interferences on neighbor SUs
(see routing characteristic R(6.4)).
Ding and Xiao [61] propose a routing and channel selection scheme to build
multiple disjoint routes (see routing characteristic R(6.1)) in order to enhance
route robustness and network utilization. This routing scheme addresses
challenges C(1.1), and C(1.4). Distributed and centralized heuristic algorithms
have been proposed (see routing characteristics R(5.1) and R(5.2), respectively).
In the distributed algorithm, the goal is to reduce total bandwidth consumption
using an efficient channel selection scheme. To minimize intra- and inter-flow
interference (see routing characteristic R(6.4)), SUs inform their two-hop
neighbor SUs about their operating channels and links. However, due to the static
channel assignment along a route and fixed timers for route discovery, this
algorithm may not provide optimal solutions. Subsequently, the centralized
algorithm is proposed and its goal is to find links with higher number of available
channels for increased flexibility in channel selection. It attempts to discover the
disjoint routes and then allocate channels on them in a centralized manner.
Qin et al. [62] propose a routing scheme with channel selection to reduce the
impact of PU interruptions in order to improve route robustness. This routing
scheme addresses challenges C(1.1), C(2.1) and C(3.1). The scheme chooses
channels with low PUL and number of hops to destination in order to reduce
channel switches and the possibility of PU reappearance. Furthermore, a route
maintenance mechanism using backup channels is proposed. After establishing a
route, each SU monitors a set of backup channels continuously, and shares this
information with its respective SU neighbors. Hence, when the PU reappears in
the current operating channel, a group of SUs switch to a backup channel and
continue with their communications.
Huang et al. [23] propose a routing scheme that takes into account route
reliability and SU node capacity for high mobility CRNs. This routing scheme
addresses challenges C(1.1), C(1.3), C(1.4), C(2.3), C(2.4) and C(3.3). Link
reliability is estimated based on the mobility patterns of airborne SUs, such as
unmanned aerial vehicles, using a bird-flocking mobility model [63]. For relay
selection, SUs with higher processing capability are selected as SU intermediate
nodes. Furthermore, a topology management mechanism based on clustering (see
routing characteristic R(6.3)) is proposed in which the formation of clusters is

24
based on various metrics, namely spectrum heterogeneity, number of hops,
transmission delay between clusters, and nodal mobility. The channel availability
for SUs is based on the respective SU’s physical location and PUL. Also, in order
to enhance throughput and delay performances between clusters, Common
Control Channel (CCC) selection mechanism, which is comprised of two main
steps, is proposed to expedite the selection of CCC. Firstly, within a group of
clusterheads, SU node contraction chooses clusterheads with higher bandwidth
availability as a representative in order to simplify the network. Secondly, the
representatives of the clusterheads choose a channel that provides low
transmission delay and high throughput as the CCC.
Wang and Huang [47] propose a routing metric that takes into account
cumulative intra-flow interference (see routing characteristic R(6.4)) and
switching delay along a route in order to improve performances on end-to-end
throughput, delay and routing overhead. This routing scheme addresses challenges
C(1.4), C(2.1), and C(2.4). The metric assigns weights to routes so that a route
with minimum interference and number of channel switches is chosen. It selects
next hop and its operating channel using a cross-layer design, which incorporates
the network, data link and physical layers. Basically, the physical and data link
layers control the transmission power and collect channel information, particularly
the intra/inter flow interferences. Subsequently, the network layer performs route
selection using this information.
4.1.3 Hybrid Routing Schemes
The hybrid routing schemes (see routing characteristic R(4.3)) have been shown
to achieve performance enhancements, particularly lower end-to-end delay, and
lower amount of control overhead (see Table 3).
Han and Huang [64] compute a link reliability metric to reduce flooding
control overhead triggered by route breakage. This routing scheme addresses
challenge C(1.1). The metric uses PUL and the number of available common
channels at a SU node pair. In this hybrid approach, proactive routing is applied to
construct a two-hop spanning tree from each SU intermediate node, which has
relatively high link reliability metric compared to its neighbor SUs; while reactive

25
routing is applied to connect the intermediate SU nodes to form a route to the
destination.
Zhu et al. [24] propose to build a routing tree for each available channel in
order to simplify channel and route selection. This routing scheme addresses
challenges C(1.1), C(1.3), C(1.4), C(2.4) and C(3.3). This hybrid approach is
comprised of proactive and reactive routings. Proactive routing is applied to
maintain an intra-channel routing tree, which is established using a single channel.
Since a SU may be an overlapping SU, which is a SU that can access multiple
channels, it may be a member of different routing trees. Proactive routing is also
applied to establish routes across different routing trees, which is also called inter-
channel routing trees if there is only a single overlapping SU that can serve as an
intermediate SU for different routing trees. Otherwise, if there are several
overlapping SUs, reactive routing is applied by the overlapping SU to establish a
route between the SU source and destination using a routing metric. Additionally,
a SU may use reactive routing to establish routes within a routing tree using a
single channel. The routing metric is based on end-to-end delay and PUL. As for
route recovery, two heuristic mechanisms are proposed to maintain the intra-
channel and inter-channel routing trees in response to link failure and SU
mobility.
Khalife et al. [65] propose a joint routing and channel selection scheme to
fulfill bandwidth requirements of flows. This routing scheme addresses challenges
C(1.1) and C(1.4). Channel bandwidth of each link is estimated using probability
distribution of PU-SU interference on that link, while taking into account channel
contention caused by other SUs. Firstly, a SU selects a route based on the
probability of meeting the bandwidth requirements of its flows. Secondly, the SU
verifies the selected route. Thirdly, if the bottleneck link of the route does not
meet the bandwidth requirements, the SU bottleneck node can add more channels
to reduce the bottleneck links.
4.1.4 Adaptive Per-hop Routing Schemes
The adaptive per-hop routing schemes (see routing characteristic R(4.4)) have
been shown to achieve performance enhancements, particularly higher throughput
performance, higher packet delivery rate, and lower end-to-end delay (see Table
3).

26
Xia et al. [66] propose a routing scheme that minimizes end-to-end delay. This
routing scheme addresses challenges C(1.1) and C(1.4). Reinforcement Learning
(RL) [44, 45] is applied to enable each SU to learn the number of available
channels along the route locally, so that a route with higher number of available
channels is chosen. This reduces the time incurred in seeking for an available
common channel for a SU node pair. In addition, backward exploration technique
[67] is proposed to update SU intermediate nodes with channel information and to
establish a route in the reverse direction, and this may reduce additional overheads
incurred by route discovery.
Talay and Altilar [68] propose two routing metrics to improve throughput and
end-to-end delay. This routing scheme addresses challenge C(1.1). The first cost
metric uses statistical information on channel availability, such as the historical
PUL, to evaluate a link. The second metric takes into account the traffic load in
the available channels by monitoring the traffic traversing through each SU node
within a time window. These metrics are used to predict the availability of the
next hop for data transmission. So, when the next hop is not available, but if it has
been predicted to be available soon, the packets will be buffered first at the
intermediate SU nodes until the next hop is available again.
Badoi et al. [69] propose a geographical-based routing scheme (see routing
characteristic R(6.2)) that uses information on channel availability, as well as
local and global topological information, including physical location of
destination SU. This routing scheme addresses challenges C(1.1) and C(2.4). The
routing metric takes into account information along a route including bandwidth
and channel availability at each SU node pair, bit error rate, and physical location
of the SU destination node. This scheme incorporates routing signaling
information into data packets in order to minimize the need for bi-directional
links, as well as route recovery that occurs due to high level of dynamicity in
channel availability and nodal mobility. This work assumes that the SU source
node is always aware of the physical location of the SU destination node, and so it
may not be feasible in large scale networks. For this reason, another work has
been proposed by the same authors [70] to address this issue by partitioning the
network into clusters (see routing characteristic R(6.3)). The SU clusterheads and
SU gateway nodes, which are the member nodes situated at the boundary of a

27
cluster, have information about their SU member nodes, and so they help establish
routes at inter- and intra-cluster levels.
Jashni et al. [71] propose a routing scheme with channel selection to support
multimedia applications. This routing scheme addresses challenge C(1.1). Each
SU uses local information received from its neighbor SUs, and selects a route with
the minimum end-to-end delay, which includes queuing and transmission delays.
Also, it takes into account channel selections of neighbor SUs in order to reduce
channel contention. The probability of channel selection (or SU behavior) of
neighbor SUs is learnt using Fictitious Play, which is a game theory
technique[72]. The queuing delay is estimated based on a packet’s traffic class
according to the QoS requirement; while the transmission delay is estimated based
on the underlying MAC/PHY layers.
Pefkianakis et al. [73] propose a routing scheme that takes into account channel
availability and quality in order to enhance route robustness, which in turn
improves throughput performance. This routing scheme addresses challenges
C(1.1) and C(3.1). Each SU measures channel availability and quality locally.
Channel availability is based on the bandwidth a SU can offer and the SUs
contention rate; while channel quality is based on the packet loss rate caused by
interference from both PUs and SUs. To enhance route robustness, the routing
metric optimizes a tradeoff between number of hop counts in a route and channel
availability, which are updated dynamically in a forwarding table. The scheme
establishes routes using periodically collected global information (see routing
characteristic R(4.1)), and subsequently route packets opportunistically based on
the forwarding table.
Ding et al. [17] combines a routing scheme with cross-layer mechanisms,
which covers channel sensing and selection, scheduling, as well as power control
schemes, in order to achieve higher throughput performance. This routing scheme
addresses challenge C(1.1). Collaborative sensing provides SUs with accurate
spectrum information. Channel selection allocates channels efficiently in a
distributed manner by reducing channel contention with other SUs. The scheduler
assigns higher priority to links with longer queues, which may indicate
congestion, so that routing through these links is minimized. The queue size is
estimated using Reinforcement Learning (RL) [44, 45]. The power control

28
mechanism selects channel that can maximize the Shannon capacity using
gradient decent algorithm.
Lin and Chen [74] propose a multipath routing (see routing characteristic
R(6.1)) that mainly considers the fading characteristics of highly dynamic (see
routing characteristic R(2.3)) wireless channels. This routing scheme addresses
challenge C(1.1). The routing metric takes into account transmission, queuing and
link-access delay for a given packet size in order to provide guarantee on end-to-
end throughput requirement. The metric is computed using a spectrum map [75],
which is generated using local sensing information in order to adapt to the
dynamic nature of the channels. Furthermore, the routing scheme selects as many
intermediate SUs as possible based on a SU’s requirement on end-to-end delay.
Naturally, this scheme establishes multiple routes and improves end-to-end delay.
Soltani and Mutka [76] propose a distributed (see routing characteristic R(5.2))
probabilistic routing scheme to improve end-to-end throughput in high density
networks with tree topology. This routing scheme addresses challenges C(1.1) and
C(1.4). Each SU establishes a route to the root of a tree based on a probability
distribution metric. The routing metric enables each SU to select the best next-
relay SU node adaptively in terms of bandwidth availability at the upper layer
based on transitional probability distribution of Markov chain.
Kim et al. [77] propose a geographical-based routing scheme (see routing
characteristic R(6.2)) to improve network throughput and route stability in highly
dynamic vehicular mobile networks. This routing scheme addresses challenges
C(1.1) and C(2.4). The SUs estimate PUL in several channels and share this
information among themselves periodically in order to select an operating channel
that provides higher network throughput and less interference from PUs. The
quality of links and routes are estimated using two metrics, namely Expected
Transmission Time (ETT) and Expected Transmission Count (ETX). ETX
estimates the mean number of transmissions required to deliver a data packet (or
data loss) over a link using PUL and distance information between the sender and
receiver; whereas ETT estimates the transmission delay over a link. Firstly, the
routing scheme constructs a list of best two-hop SU intermediate nodes towards
the destination node based on the estimated quality of links to the respective two-
hop neighbor. Subsequently, when packets reach the final destination SU via the

29
selected intermediate SU nodes, the route quality values are propagated back on
the CCC in order to update intermediate SU nodes, and select the best route.
Pan et al. [78] propose a cross-layer routing metric to enhance end-to-end
throughput. This routing scheme addresses challenges C(1.1) and C(3.1). In this
scheme, the network layer selects multiple next-hop SUs and the link layer
chooses one of them to be the actual next hop. The candidate next hops are
prioritized based on their respective links’ packet delivery rate, which in turn is
affected by the PU activities. The metric takes into account two types of links,
namely traditional link (or unlicensed link) and CR link (or licensed link). Thus,
the probability of selecting a next hop that maintains two types of links may be
higher than the ones that maintain a CR link only, which may be unavailable when
the PU reappears, resulting in throughput degradation. In addition, the routing
metric favors the next-hop SU with lower hop count to the destination, as well as
lower retransmissions rate in order to ensure a route has satisfactory throughput.
4.1.5 Other Routing Schemes
This section presents routing schemes that use other routing approaches, rather
than the traditional routing approaches (i.e. proactive R(4.1), reactive R(4.2),
hybrid R(4.3) and adaptive per-hop R(4.4)). The routing schemes have been
shown to achieve performance enhancements, particularly higher throughput
performance, higher route robustness, and lower end-to-end delay (see Table 3).
Hu et al. [79] propose a routing scheme to enhance end-to-end throughput.
This routing scheme addresses challenges C(1.1) and C(1.2). This scheme
estimates end-to-end throughput of links along a route using link rates, channels
heterogeneity, PUL and channel contention among the SU links. Accurate
estimation is achieved by observing the correlation of channel availability among
neighbor SUs. Lower value of correlations among SUs leads to higher throughput
due to the less channel contention among the SUs.
Gao et al. [80] propose a centralized (see routing characteristic R(5.1))
polynomial-time algorithm that supports multicast (see routing characteristic
R(3.2)) sessions at a given bit rate. This routing scheme addresses challenges
C(1.1) and C(2.2). It is a cross-layer approach that incorporates scheduling and
lower-layer functionalities into routing in order to minimize the required network

30
resources in terms of spectral and spatial utilization. The problem is formulated as
a mixed-integer linear program with several variables [55]. A joint scheduling and
channel selection scheme is proposed to minimize the set of channels used by a
SU sender to transmit packets to its SU receiver in order to reduce number of
broadcasts at each hop. Consequently, an iterative process using Linear
Programming (LP) [55] is applied to determine these scheduling variables with
consideration of network topology. When all variables are fixed, a multicast-tree
is constructed for each session from a SU destination node to a SU source node.
Filippini et al. [35] propose a routing metric that takes into account the cost of
route recovery in centralized and distributed (see routing characteristics R(5.1)
and R(5.2), respectively) routing models. The metric aims to improve route
robustness and to reduce repair cost incurred by channel switches when PU
activities reappear. This routing scheme addresses challenges C(3.1) and C(3.3). It
favors routes with higher hop counts to the destination, but with lower
maintenance cost. In centralized networks, the SUs have complete information
about the PU activities; while in distributed networks, they have partial is
information only. So, the repair cost optimization problem is formulated as mixed
integer linear programming [55]. In centralized networks, a polynomial time
algorithm is applied to compute routes with minimum repair cost, while in
distributed networks, a heuristic algorithm is applied.
Bütün et al. [81] predict the mobility pattern of SUs so that route recovery or
rerouting can be performed prior to route breakage in order to improve route
reliability and efficiency. This routing scheme addresses challenges C(2.4) and
C(3.3). This scheme applies Markov family predicator [82] and cumulative
distribution function to predict the next location of a mobile SU, and its time,
respectively. In other words, it uses historical data to provide the probability of a
SU moving to a given location within a future time interval. This scheme has been
shown to provide more accurate prediction compared to other predication models,
such as moving-average and static neighbor graph predictors.
Chen et al. [83] propose multi-path routing schemes (see routing characteristic
R(6.1)) to maximize throughput and minimize end-to-end delay. This routing
scheme addresses challenge C(1.1). The routing scheme minimizes both of
number of hops and SU transmission power in order to optimize MAC layer
access, hop count and retransmission delays. Two multi-path routing schemes are

31
proposed, namely, duplicated-based multipath routing scheme and coding-aided
multipath routing scheme. The multi-path routing schemes are mainly applied to
improve route robustness. In the duplicated-based approach, packets are
duplicated and forwarded over different disjoint routes. In the coding-aided
approach, instead of sending duplicated packets, it encodes packets and delivers
them over different disjoint routes. The coding-aided approach has been shown to
further improve throughput performance compared to the duplicated approach.
Huang et al. [33] propose spectrum-aware routing metrics that consider
channel availability to enhance route robustness. This scheme addresses challenge
C(1.1). The availability of a channel is defined by spectrum mobility (e.g. channel
switch or SU mobility), and route stability is defined with regards to number of
available channels at each link and PUL at each SU node. An optimal route
selection mainly takes into account PUL of links in order to minimize possibility
of channel unavailability along the route. The metric favors a route with lower
cumulative PUL and peak PUL at any of its links, so that the route has higher
route robustness, or not likely experiences a breakage.
Wen and Liao [84] propose a routing metric to reduce end-to-end delay. This
routing scheme addresses challenges C(1.1), C(2.3) and C(3.1). During route
selection, the metric takes into account the effect of PUs and two types of delays,
namely nodal delay and link delay. The nodal delay includes channel sensing time
and negotiating time; while the link delay is dependent on the number of available
channels, transmission range and PUL on the respective channel. In route
selection, the routing metric achieves an optimal trade-off between the
aforementioned parameters, such as lesser hop counts and longer transmission
range, in order to reduce end-to-end delay.
4.2 Inter-system Routing Schemes
This section discusses an inter-system routing scheme (see routing characteristic
R(1.2)).
Yuan et al. [85] propose a routing scheme that minimizes the accumulated
interference of SUs on PUs. This routing scheme addresses challenge C(3.1).
There are two types of algorithms, specifically, centralized and distributed
approaches (see routing characteristics R(5.1) and R(5.2), respectively). In the
centralized approach, a combination of flow control, channel selection, time

32
sharing and routing are proposed. On the other hand, in the distributed approach, it
utilizes interference costs and flow assignment of each link in its operation.
Subsequently, with the aid of a conventional shortest-path routing algorithm, the
distributed approach finds a route with minimum interference cost to PUs by
adjusting the transmission power in order to avoid the PU area.
4.3 Hybrid system Routing Schemes
This section discusses various kinds of hybrid-system routing schemes (see
routing characteristic R(1.3)).
4.3.1 Proactive Routing Schemes
Guan et al [86] a proactive routing scheme R(4.1) with topology control that
predicts the network topology dynamics in order to improve route robustness, as
well as end-to-end throughput and delay performances. This routing scheme
addresses challenges C(1.1), C(1.4), and C(2.4). The reliability metric predicts the
probability of the duration of link availability with consideration of interference to
PUs and mobility of SUs. Specifically, it monitors the distance between SUs and
PUs (see routing characteristic R(6.2)), and assigns weight, which is based on the
movements of SUs towards PU interference ranges, to links as part of the
prediction of link availability. Subsequently, the topology control algorithm uses
the link weights to build a reliable network topology. Hence, conventional routing
protocols can be applied on top of the topology without any modifications for the
CR environment.
4.3.2 Reactive Routing Schemes
The reactive routing schemes (see routing characteristic R(4.2)) have been shown
to reduce end-to-end delay (see Table 3).
Chowdhury and Di Felice [27] propose a geographical routing (see routing
characteristic R(6.2)) scheme for mobile SUs in order to minimize interference to
PUs and SU’s end-to-end route delay. This routing scheme addresses challenges
C(1.1), C(1.3), C(2.4) and C(3.1). There are two operation modes, namely greedy
forwarding and PU avoidance mode. Greedy forwarding mode is used during
normal operation, in which a SU forwards a RREQ to its next hop, which is

33
located in its focus region towards the destination SU node. PU avoidance mode is
triggered when PU area emerges or no SU next hop is found, and hence a RREQ
is routed far from the PU area in order to mitigate the interference to PUs. After
the RREQ is forwarded on each channel, the destination SU node selects the final
route and assigns channels to the route based on minimum number of hops in
order minimize end-to-end delay. To further optimize a route, a route maintenance
scheme is also proposed. After the initial route is selected, the route maintenance
scheme explores more potential channels that may provide lesser number of hops
to the destination than the current route.
Chowdhury and Akyildiz [37] propose a geographical routing scheme (see
routing characteristic R(6.2)) that has three main goals, namely protection of PU
receivers from SUs’ interference, provisioning of multiple routing modes, as well
as joint routing and channel selection. This routing scheme addresses challenges
C(1.1), C(1.4), C(2.4), C(3.1) and C(3.3). A SU minimizes its interference to PU
receivers by calculating its overlapping transmission range with the PU
transmitters’ coverage in order to minimize the presence of PU receivers in that
area. The routing scheme defines two operation modes based on the
implementation requirements. The first routing mode aims at achieving lower SU
end-to-end delay, while the second routing mode aims at achieving minimal
interference to PU receivers at the expense of higher end-to-end delay using
higher number of hops. The routing discovery phase is composed of local channel
selection stage at each SU, next-hop selection stage, and final route selection by
destination node. Firstly, a SU selects its operating channel to satisfy the
requirements of a chosen routing mode, and this channel is also mapped to a delay
function in the common control channel in order to reflect the estimated delay
conditions of the used data channel. Secondly, next-hop neighbors are selected
based on their initiative of forwarding packets as well as their forwarding delay of
a RREQ. Lastly, destination node selects the final route that satisfies the chosen
routing mode requirements. Furthermore, to cater for SU mobility, a route
maintenance scheme, which is run regularly, is also proposed because a SU may
overlap with PU areas, whereas reactive maintenance is used to recover from
conventional route breakage.
Abbagnale et al. [11, 87] propose a multipath-based and geographical routing
scheme (see routing characteristics R(6.1) and R(6.2)) in order to improve packet

34
delivery rate and end-to-end delay. This routing scheme addresses challenges
C(1.1), C(2.4) and C(3.1). A graph model is applied to compute a routing metric
using link connectivity. The routing scheme selects routes in consideration of the
number of hops to the destination, PU physical location of PUs, PUL, interference
caused by the other SUs, as well as the channel switching delay caused by channel
switches along the selected route.
4.3.2 Adaptive Per-hop Routing Schemes
Zhu et al. [88] propose an adaptive per-hop routing scheme R(4.4) to route
RREQs and RREPs between SUs and Cognitive Pilot Channel (CPC) base
stations, which provide sensing outcome on white spaces to SUs in order to
improve efficiency, as well as to minimize the interference to PUs and end-to-end
delay. This routing scheme addresses challenge C(1.1). Firstly, the scheme builds
a hierarchical topology from the SU source node to the SU destination node using
a reference medial axis [89], which is dependent on the geographical location of
PUs and their activities. Secondly, communications between SUs and interference
from PUs are formulated using game theory with a non-cooperative time and
location. At each level of the hierarchy, fictitious play [72] is applied so that the
SUs perform multi-stage learning, and choose the best next hop, which minimizes
delay, and interference to the PUs. The learning continues until the game
converges to Nash equilibrium.
4.3.4 Other Routing Schemes
Xie et al. [36] use geometrical approach to improve spectrum utilization. This
routing scheme addresses challenges C(3.1) and C(3.2). The geometric approach
takes into account three factors, namely SUs’ interference to PUs, SU’s network
reliability, and QoS of both SU and PU networks. To minimize SU’s interference
to PUs, the routing scheme computes the maximum allowable transmission range
based on transmission power and the physical location of both PUs and SUs (see
routing characteristic R(6.2)). To improve SU route reliability and QoS of both
SU and PU networks, a PU and SU may transmit to their destinations using
single-hop route or relays in multi-hop route in order to minimize SU-SU and SU-
PU interference. Hence, the routing scheme achieves desirable tradeoff between
lower hop count, which provides higher overall channel utilization and lower end-

35
to-end delay; and higher hop count, which provides lower energy consumption
and SU-PU interference at the expense of higher end-to-end delay.
5 Performance Enhancement of Routing Schemes
Tables 3 and 4 present the performance enhancement achieved by the routing
schemes compared to conventional and traditional approaches in CRNs. Generally
speaking, the performance enhancement metrics for a routing scheme can be
segregated into two categories, namely Application-based and Network-based.
Application-based metrics indicate the performance in a global sense, particularly
the end-to-end QoS performance. Network-based metrics indicate the
performance in a local sense, such as the bandwidth consumption and number of
operating channels at each SU intermediate node.
5.1 Application-based Performance Metrics
There are five types of application-based performance metrics as follows:
P1.1 Lower End-to-End Delay.
P1.2 Higher Throughput.
P1.3 Higher Packet Delivery Rate. Higher packet delivery rate indicates less
packet loss and higher throughput. For example, Han and Huang [64]
achieve higher packet delivery rate by selecting SUs with more reliable
links as SU intermediate nodes in order to reduce link breakage and
subsequently, to reduce flooding of RREQ control packets in search of
new routes.
P1.4 Higher Success Call Rate. Higher success call rate indicates higher
number of successful application-related requests over the total number of
requests. Consequently, this also indicates higher packet delivery rate. For
example, Xie and Xi [30] achieve higher success rate of multicast call
requests using lesser number of SU intermediate nodes.
P1.5 Higher Route Robustness. High route robustness indicates that the
selected route encounters lesser interruption by PUs, and also it reduces
overhead incurred by rerouting [9, 33, 65, 73]. For example, Filippini et
al. [35] enhance route robustness using a routing metric, which selects
routes with minimum repair cost during their entire lifetime.

36
5.2 Network-based Performance Metrics
There are six types of network-based performance metrics as follows:
P2.1 Lower Bandwidth Consumption. Lower bandwidth consumption,
particularly the routing control packets, may improve the network-wide
throughput and channel utilization. For instance, Xie and Xi [30] achieve
lower bandwidth consumption using lesser number of SU intermediate
nodes.
P2.2 Lower Level of Interference to PU Receivers. Most of the routing
schemes attempt to minimize the interference to PUs by detecting their
transmission signals; however, it is challenging to minimize the
interference to the PU receivers, such as a TV receiver that listens
passively without transmitting any signal [37]. For instance, Chowdhury
and Akyildiz [37] achieve lower interference to PU receivers. Each SU
calculates its fractional transmission area that overlaps with the PU
transmitter coverage in order to avoid PU receivers located in that area.
P2.3 Lower Energy Consumption. As an example, Kamruzzaman et al. [54]
achieve longer network lifetime using load balancing approach. It chooses
intermediate SU nodes with lesser traffic load as part of its route.
P2.4 Lower Amount of Control Overhead. Due to the dynamicity of channel
availability (see challenge C(1.1)) and PUL, routing overhead can be
significant as a result of frequent route discovery and maintenance.
Furthermore, these overhead messages can affect network performance by
interfering with PU messages, as well as reducing the residual bandwidth
at the SUs, thus they need to be minimized as much as possible. In [23,
86], clustering (see routing characteristic R(6.3)) has been proposed to
reduce the network overhead taking into account spectrum heterogeneity
in CRNs.
P2.5 Lower Computational Complexity. Routing schemes with lower
computational complexity require lesser computing resources such as
processing delay and energy consumption [18]. For example, Huang et al.
[23] propose a fast common control channel selection scheme that reduces
the computational complexity while selecting a common control channel
in a highly dynamic CRN.

37
Table 3 Performance enhancements achieved by the intra-system-based routing schemes
Intra-system
References
Application-based
Network-based
P1.1 Lower end-to-end delay
P1.2 Higher throughput
P1.3 Higher packet delivery
rate
P1.4 Higher success call rate
P1.5 Higher route robustness
P2.1 Lower bandwidth
consumption
P2.2 Lower level of
interference to PU receivers
P2.3 Lower energy
consumption
P2.4 Lower amount of
control overhead
P2.5 Lower computational
complexity
Proactive
Chen-li et al. (2009) [25]
×
×
Tuggle (2010) [48]
×
×
Wang and Aceves [51]
×
×
×
Xie and Xi (2011) [30]
×
×
Xin et al. (2008) [26]
×
Yun et al. (2010) [49]
×
Reactive
Almasaeid et al (2010) [20]
×
×
Cheng et al. (2007) [29]; How et al.
(2010) [60]; Song et al. (2009) [52]
×
Ding and Xiao (2010) [61]
×
×
×
×
Hincapie et al. (2008) [22]
×
Huang et al. (2011) [23]
×
×
×
×
Jia et al. (2009) [59]
×
×
Kamruzzaman et al. (2011) [54]
×
×
×
×
Li et al. (2009) [14]
×
×
Ma et al. (2008) [28]
×
Qin et al. (2009) [62]
×
Shih and Liao (2010) [9]
×
×
×
Song and Lin (2009) [57]
×
×
×
Wang and Huang (2010) [47]
×
×
×
×
Zheng et al. (2011) [56]
×
×
×
Zeeshan et al. (2010) [39]
×
Hybrid
Han and Huang (2010) [64]
×
×
×
Khalife et al.(2008) [65]
×
×
Zhu et al. (2008) [24]
×
×
Adaptive Per-hop
Badoi et al. (2010)[69]
×
×
Ding et al. (2010) [17]; Soltani and
Mutka (2011) [76]; Xia et al. (2009) [66]
×
×
Jashni et al. (2010) [71]
×
×
Kim et al. (2011) [77]
×
Lin and Chen (2010) [74]
×
Pan et al. (2008) [78]
×
×
Pefkianakis et al. (2008) [73]
×
×
Talay and Altilar (2009) [68]
×
×
×
Other Routing Schemes
Bütün et al. (2010) [81]
×
Chen et al. (2011) [83]
×
×
×
Filippini et al. (2009) [35]
×
Gao et al. (2011) [80]
×
Hu et al. (2007) [79]
×
Huang et al. (2011) [33]
×
×
Wen and Liao (2010) [84]
×

38
6 Open Issues
This section discusses some of the important open issues that can be pursued in
this research area. Most open issues are associated with the important routing
challenges in CRNs. Hence, the benefit of investigating the following open issues
can further enhance the performance of CRNs.
6.1 Broadcasting using a Single Transceiver
Broadcasting is an essential part of routing to negotiate for a common channel for
data transmission, to disseminate routing packets such as RREQs and RREPs, and
to perform neighbor discovery. Equipped with a single transceiver only, multiple
transmissions are necessary for a SU to deliver a broadcast packet to its neighbor
SUs, which may be listening to different channels [30, 66]. In [66], a SU sends a
Request-to-Send (RTS) in each available channel until it has successfully found
an available common channel for data transmission with its neighbor SU. Further
investigation is necessary to reduce the number of multiple transmissions for each
broadcast in order to reduce the channel switching delay incurred for each
broadcast as well as bandwidth consumption. For instance, this may be achieved
by reducing the set of available common channels shared by the SUs. This open
issue is mainly associated with challenges C(1.2) and C(2.2).
Table 4 Performance enhancements achieved by the hybrid-system-based routing schemes
Hybrid-system
References
Application-based
Network-based
P1.1 Lower end-to-end delay
P1.2 Higher throughput
P1.3 Higher packet delivery
rate
P1.4 Higher success call rate
P1.5 Higher route robustness
P2.1 Lower bandwidth
consumption
P2.2 Lower level of
interference to PU receivers
P2.3 Lower energy
consumption
P2.4 Lower amount of control
overhead
P2.5 Lower computational
complexity
Proactive
Guan et al. (2010) [86]
×
×
×
Reactive
Chowdhury and Akyildiz (2011)
[37]
×
×
Chowdhury and Di Felice
(2009) [27]; Abbagnale et al.
(2011) [11]
×
×
Adaptive
Per-hop
Zhu et al. (2010) [88]
×
Other
Routing
Schemes
Xie et al. (2010) [36]
×
×
×

39
6.2 Lack of Bi-directional links for Routing Message Exchange
The dynamic nature of the PU activities may reduce the similar set of channels at
both SU sender and receiver [90]. As a consequence, traditional routing schemes
such as AODV [42] which requires bidirectional links (e.g a RREQ requires a
RREP), must be enhanced in order to use unidirectional links. For instance,
adaptive per-hop routing scheme (see routing characteristic R(4.4)) may be
applied so that end-to-end bi-directional route is not necessary [69]. This open
issue is mainly associated with challenges C(1.1) and C(1.3).
6.3 Clustering Mechanism with PU Interference Awareness
Traditional clustering mechanisms have been shown to reduce routing overhead in
ad hoc networks (see performance enhancement P(2.4)). Further investigation is
necessary to enhance the clustering mechanisms in order to minimize interference
to PUs in cluster formation and maintenance [10]. For instance, in mobile
networks, a SU member node may re-associate with another SU clusterhead,
which possesses information or estimation about the PUs, so that interference to
the PUs is minimized. This open issue is mainly associated with the challenges
C(1.1), C(1.4) and C(2.4).
6.4 Routing Mechanisms with Route Recovery Avoidance
Due to the dynamicity of channel availability and PUL levels, SU must evacuate
their operating channels, and so route recovery is required to resume current route.
However, route recovery requires resources and it introduces new overhead [39].
Furthermore, slow or inefficient route recovery reduces average throughput [9].
Further investigation is desired to maintain route connectivity and to reduce the
number of route recoveries. For instance, selecting a robust route that has less
exposure to PUs interruptions reduces the need of route recovery, and thus
improving throughput and end-to-end delay performances [9, 65]. This open issue
is associated with challenge C(1.1).
6.5 Lack of Regular Route Maintenance
The dynamic nature of channels in CRNs caused by PU activities imposes re-
routing on SUs. A new discovered route may not be optimal compared to the
original route. Furthermore, conventional routing schemes such as AODV [42] do

40
not perform continues route maintenance in order to monitor and select better
routes, instead they keep on transmitting on a new route until it is broken while
ignoring better routes when they become available again (e.g. PUL becomes lower
on original route). As a consequence, further investigation is required to address
the issues of triggering continuous (or proactive) route maintenance in order to
utilize the best routes whenever possible (e.g. routes with higher throughput,
lesser delay, or lesser interference to PUs). For instance, Li et al. [14] applies
Swarm Intelligence (SI) [53] and Reinforcement Learning (RL) [44, 45]
techniques to keep SUs up-to-date with the spectrum information of their
neighboring SU nodes in order to trigger route maintenance. Chowdhury and
Akyildiz [37] incorporate proactive maintenance scheme that is triggered by SU
mobility in order to minimize the interference to PU receivers. This issue
addresses the challenges associated with C(1.1).
6.6 Lack of Implementation of Routing Schemes in CR Platform
To the best of our knowledge, most of the existing routing schemes have been
evaluated using simulations; and there is a limited research in the literature
regarding the implementation of routing schemes in CR platforms and testbeds.
These implementations are important to validate the correctness and performance
of the proposed algorithms [91]. By gathering real and detailed measurements on
the implementation performance, this may allow further refinements on these
algorithms. In the literature, Nagaraju et al. (2010) [92] implement a throughput
enhancement scheme that exploits a cross-layer scheme of routing and dynamic
channel selection with interference avoidance. The prototype has been
implemented using Python programming language on a software defined radio
that is based on GNU Radio and Universal Software Radio Peripheral V2
(USRP2) platform. This implementation uses Ethernet cables to simulate the CR
links which may be further investigated by implementing it on wireless medium in
order to provide more realistic and accurate evaluation results.
6.7 Further Investigation into Inter-system and Hybrid-system-based
Routing Schemes
Tables 1 and 2 show that there is limited research in the literature for inter-system
and hybrid-system compared to the intra-system based routing schemes (see

41
routing characteristics R(1.2), R(1.3) and R(1.1)). For intra-system-based routing
schemes, further investigation into their interference to the PUs may help to
improve their coexistence with PUs.
6.8 Lack of Routing Schemes for High Mobility Nodes
Routing with high mobility support in multi-hop CRNs has received limited
research attention [23, 81]. To the best of our knowledge, Huang et al. [23] and
Kim et al. [77] address routing with high mobility SUs; while a few other works
[11, 24, 27, 30, 37, 47, 69, 86] address routing with medium mobility SUs, and
most literature investigates into static networks. The fast and unpredictable
mobility pattern of SUs adds more dynamicity in addition to the dynamic,
opportunistic and heterogeneous channels (see routing characteristics R(2.2), and
R(2.3)) in CRNs. Additionally, it may introduce more interference to PUs. Even
though most schemes minimize interference to PUs at the physical and data link
layers, highly mobility of SUs may require a cross-layer design with network
layer to optimize network performance [49, 65, 73]. Further investigation is
necessary to extend the existing routing schemes, which have been designed with
static networks in mind, to minimize the interference to PUs in mobile CRNs.
This issue is mainly associated with challenge C(2.4).
6.9 IPv6 Performance in CRNs
Internet Protocol (IP) [93] is a widely deployed network-layer protocol in the
internet, which is mainly responsible for packet addressing and routing. The
current popular version, namely IPv4, supports 32-bit addressing [94]. However,
with the increasing number of IP devices in the internet, there has been increasing
demand for addresses. To this end, IPv6 [95] has been proposed, and it offers 128-
bit addressing, as well as some other improvements, such as efficient routing
capabilities and integrated security features, over the existing IPv4. The migration
from IPv4 to IPv6 has become inevitable, and research attention has been focusing
on implementing new applications over IPv6 [94].
IPv6 was not designed to operate over CRNs in mind. The characteristics of
CRNs, such as the dynamicity of operating environment, have posed challenges to
the upper layers. To enhance performance of IPv6 over CRNs, further
investigation is needed to address the following open issues:

42
• The effects of the characteristics of CR (see Section 3) on the performance
achieved by routing protocols in IPv6, such as Routing Information
Protocol next generation (RIPng) [96] and Open Shortest Path First
Version 6 (OSPFv3) [97], and Multiprotocol Border Gateway Protocol
(MBGP) [98]. Further research can also be carried to investigate new
challenges introduced by the characteristics of CR on IPv6.
• The performance of mobile IPv6 [93, 99] in CRNs.
• IPv6 approaches to QoS provisioning in CRNs.
• The performance of IPv6 multicasting services [93], such as Router
Solicitations and Advertisements, in CRN.
• The effects of the characteristics of CR on the performance achieved by
two IPv6 address configurations [93], namely stateless and stateful.
6.10 QoS Provisioning in Distributed CRNs by Telecom Providers
QoS provisioning in distributed CRNs is a daunting challenge considering that a
SU’s packet may traverse over multiple hops through more than one PU regions.
A telecom provider based on distributed CRNs may need to negotiate with
multiple PUs while establishing a route, and so achieving the QoS requirements as
stipulated in the service level agreement may be difficult. The relative lack of
research interest to address the challenges posed by the telecom providers in
distributed CRNs reflects the need to further investigate in this area.
7 Conclusions
Compared to the traditional wireless networks, Cognitive Radio Network (CRN),
which is the next generation wireless communication system, has brought new
challenges, such as dynamicity of channel availability and lack of a fixed common
control channel for control packet exchange. Firstly, this survey presents the
routing challenges in CRNs. Secondly, it presents the characteristics of both CR-
based and traditional-based routing schemes. Thirdly, it provides an extensive
survey on routing schemes, and associates each scheme to the challenges and
characteristics. Fourthly, it presents performance enhancements achieved by the
routing schemes. Fifthly, it presents open issues associated with the routing
schemes. Discussions in this article has established a foundation and sparked new
interests in this research area.

43
References
1.
Wang, C., Jiang, C., Li, X. Y., & Liu, Y. (2010). Multicast throughput for large scale
cognitive networks. Wireless Networks, 16(7), 1945-1960.
2.
Crow, B. P., Widjaja, I., Kim, L. G., & Sakai, P. T. (1997). IEEE 802.11 wireless local area
networks. IEEE Communications Magazine, 35(9), 116-126.
3.
Callaway, E., Gorday, P., Hester, L., Gutierrez, J. A., Naeve, M., Heile, B., & Bahl, V. (2002).
Home networking with IEEE 802.15.4: A developing standard for low-rate wireless personal
area networks. IEEE Communications Magazine, 40(8), 70-77.
4.
Marinho, J., & Monteiro, E. (2012). Cognitive radio: survey on communication protocols,
spectrum decision issues, and future research directions. Wireless Networks, 18(2), 147-164.
5.
FCC Spectrum Policy Task Force (2002). Report of the spectrum efficiency working group.
Federal Communications Commission, Technical Report 02-155.
6.
Akyildiz, I. F., Lee, W. Y., Vuran, M. C., & Mohanty, S. (2006). Next generation/dynamic
spectrum access/cognitive radio wireless networks: A survey. Computer Networks: The
International Journal of Computer and Telecommunications Networking, 50(13), 2127–2159.
7.
Zaidi, S. A. R., McLernon, D. C., & Ghogho, M. (2012). Quantifying the primary's guard zone
under cognitive user's routing and medium access. IEEE Communications Letters, 16(3), 288-
291.
8.
Zhou, X., Lin, L., Wang, J., & Zhang, X. (2009). Cross-layer routing design in cognitive radio
networks by colored multigraph model. Wireless Personal Communications: An International
Journal, 49(1), 123-131.
9.
Shih, C. F., & Liao, W. (2010). Exploiting route robustness in joint routing and spectrum
allocation in multi-hop cognitive radio networks. In Proceedings of IEEE Wireless
Communications and Networking Conference (WCNC) (pp. 1-5). Sydney, NSW.
10.
Li, D., & Gross, J. (2011). Robust clustering of ad-hoc cognitive radio networks under
opportunistic spectrum access. In Proceedings of the International Conference on
Communications (ICC) (pp. 1-6). Kyoto, Japan.
11.
Abbagnale, A., & Cuomo, F. (2011). Leveraging the algebraic connectivity of a cognitive
network for routing design. IEEE Transactions on Mobile Computing, In Press.
12.
Lin, L., Wang, A. P., Zhou, X. W., & Miao, X. N. (2012). Noncooperative differential game
based efficiency-aware traffic assignment for multipath routing in CRAHN. Wireless Personal
Communications: An International Journal, 62(2), 443-454.
13.
Liang, Y. C., Chen, K. C., Li, G.Y., & Mahonen, P. (2011). Cognitive radio networking and
communications: An overview. IEEE Transactions on Vehicular Technology, 60(7), 3386-
3407.
14.
Li, B., Li, D., Wu, Q. H., & Li, H. (2009). ASAR: Ant-based spectrum aware routing for
cognitive radio networks. In Proceedings of International Conference on Wireless

44
Communications & Signal Processing (WCSP) (pp. 1-5). Nanjing, China.
15.
Shi, Y., Hou, Y. T., Kompella, S., & Sherali, H. D. (2011). Maximizing capacity in multihop
cognitive radio networks under the SINR model. IEEE Transactions on Mobile Computing,
10(7), 954-967.
16.
Anifantis, E., Karyotis, V., & Papavassiliou, S. (2011). Time-based cross-layer adaptation in
wireless cognitive radio ad hoc networks. In Proceedings of IEEE Symposium on Computers
and Communications (ISCC) (pp. 1044-1049). Kerkyra, Greece.
17.
Ding, L., Melodia, T., Batalama, S. N., Matyjas, J. D., & Medley, M. J. (2010). Cross-layer
routing and dynamic spectrum allocation in cognitive radio ad hoc networks. IEEE
Transactions on Vehicular Technology, 59(4), 1969-1979.
18.
Zhang, L., Zhou, X., & Wu, H. (2009). A rough set comprehensive performance evaluation
approach for routing protocols in cognitive radio networks. In Proceedings of Global Mobile
Congress (GMC) (pp. 1-5). Shanghai, China.
19.
Akyildiz, I. F., Lee, W. Y., & Chowdhury, K. R. (2009). CRAHNs: cognitive radio ad hoc
networks. Ad Hoc Networks, 7(5), 810-836.
20.
Almasaeid, H. M., Jawadwala, T. H., & Kamal, A. E. (2010). On demand multicast routing in
cognitive radio mesh networks. In Proceedings of IEEE Global Telecommunications
Conference (GLOBECOM) (pp. 1-5). Miami, FL.
21.
Yang, Z., Cheng, G., Liu, W., Yuan, W., & Cheng, W. (2008). Local coordination based
routing and spectrum assignment in multi-hop cognitive radio networks. Mobile Networks and
Applications, 13(1-2), 67-81.
22.
Hincapie, R., Tang, J., Xue, G., & Bustamante, R. (2008). QoS routing in wireless mesh
networks with cognitive radios. In Proceedings of IEEE Global Telecommunications
Conference (GLOBECOM) (pp. 1-5). New Orleans, LA.
23.
Huang, X. L., Wang, G., Hu, F., & Kumar, S. (2011). Stability-capacity-adaptive routing for
high-mobility multihop cognitive radio networks. IEEE Transactions on Vehicular
Technology, 60(6), 2714-2729.
24.
Zhu, G. M., Akyildiz, I. F., & Kuo, G. S. (2008). STOD-RP: A spectrum-tree based on-
demand routing protocol for multi-hop cognitive radio networks. In Proceedings of Global
Telecommunications Conference (GLOBECOM) (pp. 1-5). New Orleans, LA.
25.
Chen-li, D., Guo-an, Z., Jin-yuan, G., & Zhi-hua, B. (2009). A route tree-based channel
assignment algorithm in cognitive wireless mesh networks. In Proceedings of International
Conference on Wireless Communications & Signal Processing (WCSP) (pp. 1-5). Nanjing,
China.
26.
Xin, C., Ma, L., & Shen, C. C. (2008). A path-centric channel assignment framework for
cognitive radio wireless networks. Mobile Networks and Applications, 13(5), 463-476.
27.
Chowdhury, K. R., & Di Felice, M. (2009). SEARCH: A routing protocol for mobile
cognitive radio ad-hoc networks. In Proceedings of Sarnoff Symposium (SARNOFF) (pp. 1-6).

45
Princeton, NJ.
28.
Ma, H., Zheng, L., Ma, X., & Luo, Y. (2008). Spectrum aware routing for multi-hop cognitive
radio networks with a single transceiver. In Proceedings of 3rd International Conference on
Cognitive Radio Oriented Wireless Networks and Communications (CROWNCOM) (pp. 1-6).
Singapore.
29.
Cheng, G., Liu, W., Li, Y., & Cheng, W. (2007). Spectrum aware on demand routing in
cognitive radio networks. In Proceedings of 2nd IEEE International Symposium on New
Frontiers in Dynamic Spectrum Access Networks (DySPAN) (pp. 571-574). Dublin, Ireland.
30.
Xie, L., & Xi, J. (2011). A QoS routing algorithm for group communications in cognitive
radio ad hoc networks. In Proceedings of International Conference on Mechatronic Science,
Electrical Engineering and Computer (MEC) (pp. 1953-1956). Jilin, China.
31.
Khalife, H., Malouch, N., & Fdida, S. (2009). Multihop cognitive radio networks: To route or
not to route. IEEE Network, 23(4), 20-25.
32.
Zhong, Z., & Wei, T. (2010). Cognitive routing metric with improving capacity (CRM-IC) for
heterogeneous ad hoc network. In Proceedings of International Conference on Information
Networking and Automation (ICINA) (pp. 271-274). Kunming, China.
33.
Huang, X., Lu, D., Li, P., & Fang, Y. (2011). Coolest path: Spectrum mobility aware routing
metrics in cognitive ad hoc networks. In Proceedings of 31st International Conference on
Distributed Computing Systems (ICDCS) (pp. 182-191). Minneapolis, MN.
34.
Ning, G., Duan, J., Su, J., & Qiu, D. (2011). Spectrum sharing based on spectrum
heterogeneity and multi-hop handoff in centralized cognitive radio networks. In Proceedings
of 20th Conference on Wireless and Optical Communications Conference (WOCC) (pp. 1-6).
Newark, NJ.
35.
Filippini, I., Ekici, E., & Cesana, M. (2009). Minimum maintenance cost routing in cognitive
radio networks. In Proceedings of IEEE 6th International Conference on Mobile Adhoc and
Sensor Systems (MASS) (pp. 284-293). Macau, China.
36.
Xie, M., Zhang, W., & Wong, K. K. (2010). A geometric approach to improve spectrum
efficiency for cognitive relay networks. IEEE Transactions on Wireless Communications,
9(1), 268-281.
37.
Chowdhury, K. R., & Akyildiz, I. F. (2011). CRP: A routing protocol for cognitive radio ad
hoc networks. IEEE Journal on Selected Areas in Communications, 29(4), 794-804.
38.
Hou, L., Yeung, K. H., & Wong, K. Y. (2011). A vision of energy-efficient routing for
cognitive radio ad hoc networks. In Proceedings of 6th International Symposium on Wireless
and Pervasive Computing (ISWPC) (pp. 1-4). Hong Kong, China.
39.
Zeeshan, M., Manzoor, M. F., & Qadir, J. (2010). Backup channel and cooperative channel
switching on-demand routing protocol for multi-hop cognitive radio ad hoc networks
(BCCCS). In Proceedings of 6th International Conference on Emerging Technologies (ICET)
(pp. 394-399). Islamabad, Pakistan.

46
40.
Lei, G., Wang, W., Peng, T., & Wang, W. (2008). Routing metrics in cognitive radio
networks. In Proceedings of 4th IEEE International Conference on Circuits and Systems for
Communications (ICCSC) (pp. 265-269). Shanghai, China.
41.
Clausen, T., & Jacquet, P. (2003). Optimized link state routing protocol (OLSR). IETF RFC
3626.
42.
Perkins, C.E., & Royer, E.M. (1999). Ad-hoc on-demand distance vector routing. In
Proceedings of Mobile Computing Systems and Applications (WMCSA) (pp. 90-100). New
Orleans, LA.
43.
Haas, Z. H., & Pearlman, M. R. (2001). ZRP: A hybrid framework for routing in ad Hoc
networks. In Perkins, C. E., Ad Hoc Networking. Boston, MA: Addison-Wesley Longman
Publishing Co., Inc.
44.
Sutton, R. S., & Barto, A. G. (1998). Reinforcement learning: An introduction. Cambridge,
Massachusetts: MIT Press.
45.
Yau, K.-L., Komisarczuk, P., & Teal, P. D. (2012). Reinforcement learning for context
awareness and intelligence in wireless networks: Review, new features and open issues.
Journal of Network and Computer Applications, 35(1), 253-267.
46.
Imielinski, T., & Navas, J. (1996). GPS-based addressing and routing. IETF RFC 2009.
47.
Wang, J., & Huang, Y. (2010). A cross-layer design of channel assignment and routing in
Cognitive Radio Networks. In Proceedings of 3rd IEEE International Conference on
Computer Science and Information Technology (ICCSIT) (pp. 542-547). Chengdu, Sichuan.
48.
Tuggle, R. (2010). Cognitive multipath routing for mission critical multi-hop wireless
networks. In Proceedings of 42nd Southeastern Symposium on System Theory (SSST) (pp. 66-
71). Tyler, TX.
49.
Yun, L., Fengxie, Q., Zhanjun, L., & Hongcheng, Z. (2010). Cognitive radio routing algorithm
based on the smallest transmission delay. In Proceedings of 2nd International Conference on
Future Computer and Communication (ICFCC). Wuhan, Hubei.
50.
Dijkstra, E. W. (1959). A note on two problems in connexion with graphs. Numerische
Mathematik, 1(1), 269-271.
51.
Wang, X., & Aceves, J. J. G. L. (2011). Collaborative routing, scheduling and frequency
assignment for wireless ad hoc networks using spectrum-agile radios. Wireless Networks,
17(1), 167-181.
52.
Song, Z., Shen, B., Zhou, Z., Kwak, , & S., K. (2009). Improved ant routing algorithm in
cognitive radio networks. In Proceedings of 9th International Symposium on Communications
and Information Technology (ISCIT) (pp. 110-114). Incheon, South Korea.
53.
Kennedy, J., & Eberhart, R. (1995). Particle swarm optimization. In Proceedings of IEEE
International Conference on Neural Networks (pp. 1942-1948). Perth.
54.
Kamruzzaman, S. M., Kim, E., & Jeong, D. G. (2011). Spectrum and energy aware routing
protocol for cognitive radio ad hoc networks. In Proceedings of IEEE International

47
Conference on Communications (ICC) (pp. 1-5). Kyoto, Japan.
55.
Bazaraa, M. S., Jarvis, J. J., & Sherali, H. D. (2010). Linear programming and network flows.
4th ed. New Jersey, USA: John Wiley & Sons.
56.
Zheng, C., Liu, R. P., Yang, X., Collings, I. B., Zhou, Z., & Dutkiewicz, E. (2011). Maximum
flow-segment based channel assignment and routing in cognitive radio networks. In
Proceedings of IEEE 73rd Vehicular Technology Conference (VTC Spring) (pp. 1-6).
Budapest, Hungary.
57.
Song, H., & Lin, X. (2009). Spectrum aware highly reliable routing in multihop cognitive
radio networks. In Proceedings of International Conference on Wireless Communications &
Signal Processing (WCSP) (pp. 1-5). Nanjing, Jiangsu.
58.
Sniedovich, M. (2010). Dynamic programming: Foundations and principles. 2nd ed. Boca
Raton, FL: CRC Press.
59.
Jia, J., Zhang, J., & Zhang, Q. (2009). Relay-assisted routing in cognitive radio networks. In
Proceedings of IEEE International Conference on Communications (ICC) (pp. 1-5). Dresden,
Germany.
60.
How, K. C., Ma, M., & Qin, Y. (2010). An opportunistic service differentiation routing
protocol for cognitive radio networks. In Proceedings of IEEE Global Telecommunications
Conference (GLOBECOM) (pp. 1-5). Miami, FL.
61.
Ding, Y., & Xiao, L. (2010). Routing and spectrum allocation for video on-demand streaming
in cognitive wireless mesh networks. In Proceedings of IEEE 7th International Conference on
Mobile Adhoc and Sensor Systems (MASS) (pp. 242-251). San Francisco, CA.
62.
Qin, L., Wang, J., & Li, S. (2009). Stability-driven routing and spectrum selection protocol in
cognitive radio networks. In Proceedings of International Conference on Communications
Technology and Applications (ICCTA) (pp. 269-273). Beijing, China.
63.
Zhao, Z. J., Xu, S. Y., Zheng, S. L., & Niu, Y. X. (2009). Cognitive radio decision engine
based on binary particle swarm optimization. Acta Physica Sinica, 58(7), 5118-5125.
64.
Han, R., & Huang, X. (2010). Reliable link routing in cognitive radio networks. In
Proceedings of 2nd International Asia Conference on Informatics in Control, Automation and
Robotics (CAR) (pp. 55-58). Wuhan, Hubei.
65.
Khalife, H., Ahuja, S., Malouch, N., & Krunz, M. (2008). Probabilistic path selection in
opportunistic cognitive radio networks. In Proceedings of Global Telecommunications
Conference (GLOBECOM) (pp. 1-5). New Orleans, LA.
66.
Xia, B., Wahab, M. H., Yang, Y., Fan, Z., & Sooriyabandara, M. (2009). Reinforcement
learning based spectrum-aware routing in multi-hop cognitive radio networks. In Proceedings
of 4th International Conference on Cognitive Radio Oriented Wireless Networks and
Communications (CROWNCOM) (pp. 1-5). Hannover, Germany.
67.
Kumar, S., & Miikkulainen, R. (1997). Dual reinforcement Q-routing: An on-line adaptive
routing algorithm. In Proceedings of the Artificial Neural Networks in Engineering

48
Conference (ANNIE). Louis, Missouri.
68.
Talay, A. C., & Altilar, D. T. (2009). ROPCORN: Routing protocol for cognitive radio ad hoc
networks. In Proceedings of International Conference on Ultra Modern Telecommunications
& Workshops (ICUMT) (pp. 1-6). Petersburg, Russia.
69.
Badoi, C. I., Croitoru, V., & Prasad, R. (2010). IPSAG: An IP spectrum aware geographic
routing algorithm proposal for multi-hop cognitive radio networks. In Proceedings of 8th
International Conference on Communications (COMM) (pp. 491-496). Bucharest, Romania.
70.
Badoi, C. I, Croitoru, V., Prasad, N., & Prasad, R. (2010). HC-IPSAG and GC-IPSAG
algorithm proposals: Cluster based IPSAG algorithm variations for large cognitive radio
networks. In Proceedings of the 52nd International Symposium ELMAR (pp. 239-242). Zadar,
Croatia.
71.
Jashni, B., Tadaion, A. A., & Ashtiani, F. (2010). Dynamic link/frequency selection in multi-
hop cognitive radio networks for delay sensitive applications. In Proceedings of 17th
International Conference on Telecommunications (ICT) (pp. 128-132). Doha, Qatar.
72.
Fudenberg, D., & Levine, D. (1998). The theory of learning in games. Cambridge, MA: MIT
Press.
73.
Pefkianakis, I., Wong, S. H. Y., & Lu, S. (2008). SAMER: Spectrum aware mesh routing in
cognitive radio networks. In Proceedings of 3rd IEEE Symposium on New Frontiers in
Dynamic Spectrum Access Networks (DySPAN) (pp. 1-5). Chicago, IL.
74.
Lin, S.-C., & Chen, K.-C. (2010). Spectrum aware opportunistic routing in cognitive radio
networks. In Proceedings of IEEE Global Telecommunications Conference (GLOBECOM)
(pp. 1-6). Miami, FL.
75.
Wilson, J., & Patwari, N. (2010). Radio tomographic imaging with wireless networks. IEEE
Transactions on Mobile Computing, 9(5), 621-632.
76.
Soltani, S., & Mutka, M. (2011). On transitional probabilistic routing in cognitive radio mesh
networks. In Proceedings of IEEE International Symposium on a World of Wireless, Mobile
and Multimedia Networks (WoWMoM) (pp. 1-9). Lucca, Italy.
77.
Kim, W., Oh, S. Y., Gerla, M., & Lee, K. C. (2011). CoRoute: A new cognitive anypath
vehicular routing protocol. In Proceedings of 7th International on Wireless Communications
and Mobile Computing Conference (IWCMC) (pp. 766-771). Istanbul, Turkey.
78.
Pan, M., Huang, R., & Fang, Y. (2008). Cost design for opportunistic multi-hop routing in
Cognitive Radio networks. In Proceedings of Military Communications Conference
(MILCOM) (pp. 1-7). San Diego, CA.
79.
Hu, C., Lei, G., & Qian, R. (2007). Observing correlation aware (OCA) routing metric in
cognitive radio networks. In Proceedings of 4th International Conference on Communications
and Networking in China (ChinaCOM) (pp. 1-5). Xian, Shaanxi.
80.
Gao, C., Shi, Y., Hou, Y. T., Sherali, H. D., & Zhou, H. (2011). Multicast communications in
multi-hop cognitive radio network. IEEE Journal on Selected Areas in Communications,

49
29(4), 784-793.
81.
Bütün, I., Talay, A. C., Altilar, D. T., Khalid, M., & Sankar, R. (2010). Impact of mobility
prediction on the performance of cognitive radio networks. In Proceedings of Wireless
Telecommunications Symposium (WTS) (pp. 1-5). Tampa, FL.
82.
Cheng, C., Jain, R., & Berg, E. V. D. (2003). Location prediction algorithms for mobile
wireless systems. In Furht, B., & Ilyas, M., Wireless Internet Handbook: Technologies,
Standards, and Applications (pp. 245-263). Boca Raton, FL: CRC Press, Inc.
83.
Chen, P. Y., Cheng, S. M., Ao, W. C., & Chen, K. C. (2011). Multi-path routing with end-to-
end statistical QoS provisioning in underlay cognitive radio networks. In Proceedings of IEEE
Conference on Computer Communications Workshops (INFOCOM WKSHPS) (pp. 7–12).
Shanghai, China.
84.
Wen, Y. F., & Liao, W. (2010). On QoS routing in wireless ad-hoc cognitive radio networks.
In Proceedings of 71st IEEE International Conference on Vehicular Technology Conference
(VTC 2010-Spring) (pp. 1-5). Taipei, Taiwan.
85.
Yuan, Z., Song, J. B., & Han, Z. (2010). Interference minimization routing and scheduling in
cognitive radio wireless mesh networks. In Proceedings of IEEE Wireless Communications
and Networking Conference (WCNC) (pp. 1-6). Sydney, NSW.
86.
Guan, Q., Yu, F. R., Jiang, S., & Wei, G. (2010). Prediction-based topology control and
routing in cognitive radio mobile ad hoc networks. IEEE Transactions on Vehicular
Technology, 59(9), 4443-4452.
87.
Abbagnale, A., Cuomo, F., & Salvo, P. (2011). Comparison of utility functions for routing in
cognitive wireless ad-hoc networks. In Proceedings of the 10th IFIP Annual Mediterranean
on Ad Hoc Networking Workshop (Med-Hoc-Net) (pp. 127-130). Favignana Island, Sicily.
88.
Zhu, Q., Yuan, Z., Song, J. B., Han, Z., & Basar, T. (2010). Dynamic interference
minimization routing game for on-demand cognitive pilot channel. In Proceedings of IEEE
Global Telecommunications Conference (GLOBECOM) (pp. 1-6). Miami, FL.
89.
Bruck, J., Gao, J, & Jiang, A. A. (2006). MAP: Medial Axis Based Geometric Routing in
Sensor Networks. Wireless Networks, 13(6), 835-853.
90.
Badoi, C. I., Croitoru, V., & Popescu, A. (2011). Comparative performance evaluation of
IPSAG and HC-IPSAG cognitive radio routing protocols. In Proceedings of the 10th
International Symposium on Signals, Circuits and Systems (ISSCS) (pp. 1-4). Lasi, Pakistan.
91.
Cesana, M., Cuomo, F., & Ekici, E. (2010). Routing in cognitive radio networks: Challenges
and solutions. Ad Hoc Networks, 9(3), 228–248.
92.
Nagaraju, P. B., Ding, L., Melodia, T., Batalama, S. N., Pados, D. A., & Matyjas, J. D. (2010).
Implementation of a distributed joint routing and dynamic spectrum allocation algorithm on
USRP2 radio. In Proceedings of the 7th Annual IEEE Communications Society Conference on
Sensor Mesh and Ad Hoc Communications and Networks (SECON) (pp. 1-2). Boston, MA.
93.
Gai, S. (1998). Internetworking IPv6 with cisco routers. TX: Mcgraw-Hill.

50
94.
Dunlop, M., Groat, S., Marchany, R., & Tront, J. (2011). The good, the bad, the IPv6. In
Proceedings of 9th Annual Conference on Communication Networks and Services Research
(CNSR) (pp. 77-84). Blacksburg, VA.
95.
Stallings, W. (1996). IPv6: the new internet protocol. IEEE Communications Magazine, 34(7),
96-108.
96.
Malkin, G., & Minnear, R. (1997). RIPng for IPv6. IETF RFC 2080.
97.
Coltun, R., Ferguson, D., Moy, J., & Lindem, A. (2008). OSPF for IPv6. IETF RFC 5340.
98.
Marques, P., & Dupont, F. (1999). Use of BGP-4 multiprotocol extensions for IPv6 inter-
domain routing. IETF RFC 2545.
.99.
Johnson, D., Perkins, C., & Arkko, J. (2004). Mobility support in IPv6. IETF RFC 3775.