See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/274080086

Challenges for Network Aspects of Cognitive Radio

Conference Paper · May 2015

DOI: 10.1109/ICMCIS.2015.7158680

CITATIONS READS

2 189

4 authors, including:

Jaroslaw Krygier Vincent Le Nir

Military University of Technology Royal Military Academy
26 PUBLICATIONS   44 CITATIONS    110 PUBLICATIONS   539 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

IST-140 Cognitive Radio Networking View project

All content following this page was uploaded by Vincent Le Nir on 27 March 2015.

The user has requested enhancement of the downloaded file.

 Challenges for Network Aspects of Cognitive Radio
Stefan Couturier Jaroslaw Krygier Ole Ingar Bentstuen Vincent Le Nir
Fraunhofer FKIE Military University of Norwegian Defence Research Royal Military Academy
Wachtberg, Germany Technology Establishment (FFI) Brussels, Belgium
stefan.couturier@ Warsaw, Poland P.O. Box 25 [email protected]
fkie.fraunhofer.de [email protected] 2027 Kjeller, Norway
[email protected]

Abstract—Cognitive Radio is often promoted to be a promising spectrum decision, spectrum mobility, and spectrum sharing.
solution for the spectrum scarcity problem. Hence, a lot of
research activities in both the civilian and the military world CR has so far mostly been identified with exploiting unused
focus on introducing cognitive radio technology into modern frequency bands licensed for TV. The idea is that an unlicensed
networks. Besides the obvious changes in the radio architecture radio, called Secondary User (SU), can use a licensed
(sensing capability, cognitive engine), this combination of frequency band when the license holder, the Primary User (PU)
technologies also poses new challenges to existing protocols, like is not using the band. The idea was first promoted by Mitola
the support for Dynamic Spectrum Access. Moreover, the [31] in 1999.
additionally gathered information about the spectrum
environment can also be used to improve the quality of
DSA can be realized both in infrastructure-based networks
communication. In our paper, we focus on network aspects like and in ad hoc networks. An infrastructure-based network is
Quality of Service or routing and identify challenges regarding usually organized in a centralized way, featuring a main entity
their application in a cognitive radio system. We consider such as a base station or an access point. In contrast to that, ad
architectural features of infrastructure and ad hoc networks, and hoc networks do not have any infrastructure backbone and are
we analyse their influence on the control channel. The analysis is organized in a distributed manner. Both types of networks can
carried out in terms of specific requirements for military be enhanced using CR technology to allow for real-time
networks. adjustment of spectrum utilization in response to changing
environments and objectives. Nevertheless, there will be
Keywords-cognitive radio; network; infrastructure; ad hoc; differences between the enhancements regarding multi-hop
QoS; routing; topology; control channel architecture, the dynamic network topology, and the time and
location varying spectrum availability.
I. INTRODUCTION
The spectrum is a limited resource. Military tactical Networking in a radio is responsible for making sure that
networks are expected to support a greater number of services, messages are transmitted on the optimal route with the highest
and the bandwidth requirements of many of the new services possible quality. In a dynamically changing environment, such
are also rapidly increasing. This means that we are gradually as a military operation, this can be a challenging task.
getting closer to a situation where there will be insufficient Especially when using CR networks, the link quality might
bandwidth to support future military operations. Additionally, vary with every frequency change. Nevertheless, a CR network
in the battlefield the spectrum availability can vary quite possesses advanced situation awareness, e.g. the knowledge
drastically depending on the geographical location and with about the spectrum environment at different places. This
transitory and long-term communications patterns. information can be used to optimize network properties, which
Furthermore, many future military operations will likely be requires that networking parameters are adapted based on
undertaken by several participating nations cooperating in a spectrum conditions and exchanged between network nodes.
coalition force. With the increasing demand for radio Our paper will identify such parameters and analyse the
communication in such operations, it makes sense to challenges regarding their optimization by CR technology.
dynamically coordinate and share spectrum between different The next sections will be organized as follows. Section II
participating nations. will look at existing infrastructure-based CR networks and
Dynamic Spectrum Access (DSA) is a concept for non- analyse their use for military purposes. In Section III the
static spectrum utilization. A promising approach for characteristics of ad hoc networks, like clustering or routing
implementing DSA is Cognitive Radio (CR). A CR senses its aspects, will be studied regarding their application to CR
environment and takes the decision to dynamically adapt its networks, with a focus on military networking. Design options
transmission parameters, e.g. the used frequency. That requires for a common control channel in military CR networks will be
the capability to change frequencies without interrupting presented in Section IV. Section V will conclude our findings.
communication and without interfering with other radios. In II. INFRASTRUCTURE-BASED NETWORKS
[1], these functions are denominated spectrum sensing,
Infrastructure-based solutions are the field with most
This research was carried out as part of NATO STO IST-ET-074 progress so far within Cognitive Radio, with the availability of
“Network Aspects of Cognitive Radio”. IEEE 802.11af [32] and IEEE 802.22 [33]. There is also work
within national regulation bodies and standardisation base station is mandated to update the central database once per
organizations within the field of databases for cognitive radios. 24 hours.
A. Existing standards B. Other work
IEEE 802.11af and 802.22 are the two current standards for US DARPA issued its DARPA XG program around 2003.
infrastructure-based CR. 802.11af allows WLANs to exist The aim for the XG-program was not to build a new radio, but
within frequency bands allocated to TV transmission. IEEE to enable technologies for dynamic spectrum access. The work
802.22 is a long range (30 to 100 km) point-multipoint solution has defined several modules that can be used for cognitive
aimed for rural areas with limited broadband services available. radios, ranging from signalling policies affecting the CR-
For both IEEE 802.11af and 802.22, it is the base station that function, setting radio parameters to signalling between
will coordinate the frequency used by all the clients (CPE – cooperating cognitive radios. Successful testing has been
Customer Premise Equipment) connected to that base station. conducted by modifying existing radios in use by the US DoD.
Both standards cover usage of TV white space in Within the EU research program, the project CORASMA
frequencies from around 50 MHz to around 900 MHz, but will [6] has explored cognitive radio systems.
also allow operations in the 3 GHz band. Both standards
coordinate frequency usage through some form of centralized There exists a proposal [38] for peer-to-peer technology for
database. How this database is implemented, and how CRs coordination between cognitive radio systems. One practical
should use such databases will depend on the regulatory body use-case used of the technology is coordination between Wi-Fi
access points in a building or a campus. Proposals like this are
in each nation.
identified to also cover use-cases where radio networks can
Co-existence of IEEE 802.11af and 802.22 is covered by coordinate their dynamic spectrum access through a common
the standardization group within IEEE 802.19 [35]. IEEE backbone, to which each radio network has a gateway.
802.19.1 regulates co-existence within the TV white space.
C. Centralized databases
1) IEEE 802.11af Both the standards mentioned earlier require central
The IEEE 802.11af is an extension that allows Wi-Fi to database for coordinating with both primary and secondary
coexist as a secondary user in the TV white space. The users. There will likely be a lot of different approaches within
standard only covers the radio interface part of the system. this field due to different regulatory restrictions in different
Communication with neither a centralized database, nor nations. In US, the databases themselves are implemented by
implementation of the database itself is part of the standard. commercial vendors, under oversight of FCC. In other
Different bodies are standardizing the database; ETSI BRAN countries, like in EU, it is expected that government agencies
[37] in EU-countries while IETF PAWS [36] makes a generic will run these databases.
standard.
All the proposed solutions support a nested or hierarchical
The US Federal Communication Commission (FCC) and approach to implementing the databases. Each database might
its counterparts in EU have different approaches and support a geographical area, a license band or following
regulations for 802.11af. FCC allows for 48 hours between organizational structures.
updates of the base station, with frequency bands of 6MHz and
power emission at either 100 mW or 40 mW depending on A very important part of the database implementation is the
proximity to any primary user. In EU, the "lease" time is 2 message sets defined for communication between a cognitive
hours for 8 MHz bands, and power emission is dynamic based radio and the database, or for coordination between databases.
on proximity to both primary users and other secondary users. There are several efforts for standardizing these messages, with
Any mobility mandates a new request to the central database. IETF PAWS and ETSI BRAN as two major efforts.
The architecture for IEEE 802.11af allows for nested base D. Issues for military use
stations. This allows a base station to be a customer to another None of the proposed standards support mobile users. ETSI
base station and use that connection to negotiate with the BRAN do support and require a position update to the database
central database. Maximum range for 802.11af is 5 km. each time a CPE moves more than 50 m. There is however a
possibility to extend the proposed IETF PAWS
2) IEEE 802.22 recommendation since that protocol allows a geographical
IEEE 802.22 is made for rural areas where broadband position to be described by a polygon. If that polygon is
connections are not typically available. Maximum distance
allowed to be a larger size, a cognitive radio can specify in
between a base station and a customer is 30 km, but can be which geographical area it wants to operate.
extended up to 100 km.
All solutions discussed earlier require exact geographical
Due to the extended coverage compared to IEEE 802.11af,
positions of the base station and the mobile users. This has both
802.22 have more functionality for spectrum sensing and a practical side and a security side to it. Implementing GPS in
coordination. Each base station will handle the frequency
all radio systems will lead to a high cost, but we already see
negotiation for all its connected CPEs (Customer Premise that some radio systems include a GPS-receiver as standard.
Equipment). All CPEs can participate in the spectrum sensing The security issue might be a larger problem. A centralized
process, and different base stations can coordinate directly in
register of the location of all CR-enabled radio systems might
case of overlapping coverage between the base stations. Each get too high classification to be of practical usage. In addition,
these positions might be in different security domains, thereby routing layer can take this into consideration and move traffic
severely hindering the exchange of such information between away from the affected areas.
different databases.
2) Spectrum-aware routing
One possible use-case for infrastructure-based CR- The routing protocol can be informed about the availability
technology is to coordinate different Combat Net Radio of spectrum in different parts of the networks. This might be
systems that are connected to a common backbone through a used by the routing protocols to route majority of the traffic
set of gateways. None of the standards supports this use-case. through parts of the network with good availability of
IEEE 802.11af might partly support this by allowing nested frequencies. The routing protocols might also take link quality
base stations. This use-case might need extensions of the and similar parameters into account if those are available in the
protocols, but is likely to be useful for the military community. database.
The current proposed message sets for database 3) Traffic-aware DSA
communication are tuned to intended frequency bands of 50 to With cross-layer communication between the network layer
900 MHz and 3 GHz. Military cognitive radios will most likely and the spectrum databases, information from the network
operate outside of these frequencies. These protocols need to layer can influence the spectrum allocations. The network layer
be studied further to check if other frequency bands will be can inform the database about traffic volumes and possibly the
supported. location of the important or heavy users of network services.
E. Ideas for use in military networks The database can then allocate more frequencies to the parts of
the network with large amount of traffic.
The two proposed standards for infrastructure-based CR
have in themselves dubious usage in military networks, but III. AD HOC NETWORKS
they might be highly valuable starting points for new Cognitive Radio Ad Hoc Networks (CRAHN) are
technology that might enter military networks in the future.
characterised by distributed multi-hop architecture, dynamic
Both the standards are for point-to-multipoint static network topology, and time and location varying spectrum
networks. Today, there are few deployments of such networks availability. Compared to classical ad hoc networks, CRAHNs
in military systems. It is however likely that armed forces will can operate in wide frequency range varying in time and space,
start to use that kind of networks more frequently in the future. while in a classical approach, whole networks operate mainly
The advantages of LTE and similar technologies will lead to an on one or more predefined channels all the time. The CRAHN's
increasing number of base station technologies. IEEE 802.22 users can use the spectrum allocated to the PUs if it is not used,
could be an alternative for relatively static operations like some and they must vacate it in other cases to protect the PUs
of the operations NATO has been involved in lately in nations transmissions. In CRAHN, each node can detect different
like Afghanistan. Likewise, technologies like 802.11af might spectrum availability according to the PUs activity, which
have its use-cases within deployable field Headquarters. limits global network management and optimal resource
allocation and utilisation. Additionally, classical ad hoc
A more interesting aspect is what a common database and networks generally use periodic beacon transmission in
protocols for exchanging radio parameters can have for predefined channels to ensure topology control.
military networks. There are two approaches to this; 1) using
the exchange protocols for better and faster spectrum Since CRAHNs can operate in wide range of channels
management, and 2) utilizing the central database for network depending on PU activity, controlling the network topology is
aspects. very difficult. It leads to the problem of unstable network
operation. A further CRAHN feature is a multi-hop end-to-end
Policy language and database approach might utilize much transmission using many different channels. It influences the
faster and simplified (in man hours) spectrum coordination and routing algorithms that need to find and build the paths
management in a military operation. Each nation, unit or composed of the links with different characteristics, depending
similar in an operation can coordinate their frequency on allocated bandwidth in any part of the network. Moreover,
requirements and usage much more dynamically thanks to a the PUs activity allows frequent channels handoff, which
formal language understandable by computers. This dynamic influences the end-to-end transmission regarding routing and
frequency management can even be performed with both quality of service (QoS) support. It is also required to
legacy and cognitive radio systems. distinguish typical user mobility from bandwidth handoff. Both
Network layer access to the frequency databases might events influence routing algorithms in the CRAHN. User's
allow for a range of new exciting functionalities: mobility is taken into account in classical ad hoc networks
routing protocols, but frequent bandwidth handoff may lead to
1) Seal-healing from electronic warfare the slow route recovery.
If an adversary performs electronic warfare by jamming our
network, there are several ideas that can be explored. On the In order to adapt to the dynamic spectrum environment the
CR-level, frequencies might be "moved" between geographical CRs perform following functions [1]:
areas so that the radios closest to the jammer are allocated  spectrum sensing,
frequencies outside the jamming range.
 spectrum decision,
If the databases receive information about jamming through
the spectrum sensing functions in the cognitive radios, the  spectrum sharing,
  spectrum mobility. CRAHNs the network can be dynamically organised in clusters
to support spectrum allocation, interference minimization,
Each function influences different layers of the CR and faster cognitive control messages flow and better cooperation
then the operation of the whole CRAHN. Spectrum sensing between the nodes (i.e. cooperative sensing).
process operates mainly in PHY and MAC layers and it can
provide the information about the spectrum availability to the In military CRAHNs, there exist no mature solutions for
other functional processes and upper layer protocols. Spectrum clustering that effectively support cognitive functions. Further
decision process is responsible for selecting appropriate studies of the existing solutions are required, to either update
channels based on the sensing results and spectrum sharing the existing solutions or propose new ones that meet military
procedures. Spectrum decisions should also be made based on CRAHNs requirements.
the user (applications) requirements resulting from the QoS
One of the challenges is to propose the clustering methods
assumptions. In order to allocate appropriate channels to the
supporting a cooperative spectrum sensing. Sensing
network, the CRs have to communicate using network layer.
performance degradation can be observed due to the fading and
The spectrum decisions should also cooperate with the routing
shadowing that happens in the reporting channel, which
protocol to adjust the routes according to the routing metrics
forwards the sensing observations to a common receiver.
calculated based on the links characteristics. Spectrum sharing
Cluster-based cooperative spectrum sensing method may be
process is responsible for resource allocation to the CRs in
applied to improve the sensing performance. Organising the
order to avoid spectrum overlapping in the network and
SU network as a set of clusters with some cluster heads allows
thereby lowering the possibility of interference and collisions.
reporting sensing results to the common receiver with a spatial
Most CRAHN solutions in this area are focused on the
diversity.
clustering algorithms implemented in data link layers of the
CRs. Such solutions support controlled and hierarchical When clustering is applied in the CRAHNs, the network
resource allocation (i.e. channels are allocated to the clusters also needs to support clustering technology. Routing
organised dynamically or depending on military units boundaries and topologies should adhere to the clustering of
structure). High dynamics in clustering processes implicates the radios.
high dynamics in resource allocation, and then network
management. Spectrum sensing activity allows detecting free B. Cognitive Routing
bandwidth that can be used by CRs. It leads to the spectrum Classical wireless ad hoc networks use wide range of
handoff in some part of the network and in consequence to routing protocols which construct typical routing tables
appropriate reactions of the topology control mechanisms and keeping only the next hop and metric information. Most of
routing, transport and application protocols. Based on this them are based on proactive Optimized Link State Routing
introduction we can distinguish the following main CRAHN (OLSR) [22] or reactive Ad-hoc On-demand Distance Vector
issues that should be raised: (AODV) [23] protocols with some modifications to adjust them
to a specific wireless environment. For example, to select links
 cognitive clustering, with the highest transmission quality instead of the shortest
 cognitive routing, route, OLSR uses quality extension by calculating the ETX
metric [24]. Some of the routing protocols use geographical
 cognitive topology control, positions to find and select the best routes [25]. Others are
equipped with mechanisms borrowed from nature (e.g. ant
 cognitive data transport, routing [26]). All of them are not designed for CR networks.
 cognitive QoS management. CRAHN routing protocols should not only find routes with
The next sections are focused on identification and the shortest path (or the path composed of the strongest links),
description of challenges of the above networking issues, but must take into account allocated channels in the network,
especially in terms of military requirements. PUs activity and potential channels that can be used, user's
QoS requirements and other relevant channel characteristics.
A. Cognitive clustering techniques Moreover CRAHN routing mechanisms should cooperate both
A lot of research has been conducted in the field of ad hoc with spectrum sensing and spectrum decision processes during
network clustering. Generally, two types of ad hoc network channel allocation. For example, next hop nodes have to be
topology structures have been proposed [18], flat topology selected not only according to possible shortest path to the
structure and hierarchical clustering structure. The flat destination, but also based on the information about possible
topology structure used in wireless network composed of many channels (and their characteristics) that can be used in the link
nodes results in unstable topology, thus it is inefficient for self- between the nodes. The CR nodes have to collect information
organizing networks. Therefore, hierarchal clustering structures not only about possible channels that can be used (i.e. released
are often proposed. It can also be assumed that CRs will be by PUs), but also statistics about PU activity, to predict
operated in networks composed of many nodes. Some of the channels occupation time. This information can support CR
CRAHN solutions [6][12] propose hierarchical structures using nodes to perform both the routing and channel allocation
clustering techniques. It gives possibility not only to better decisions, taking into account end-to-end path and bandwidth
utilize the resources but also to organize the network according stability.
to spectrum availability, node's membership to the groups of
The next problem with CRAHN routing is that the fixed
interests, strongest links, or addressing. Additionally, in
common control channel (CCC) used in traditional routing
protocols is infeasible. To find an appropriate path over the are some drawbacks of such a solution. Firstly, allocated
network that can use multiple channels, all the possible links channels can be changed during network operation (due to the
(with different channels used) should be checked. When spectrum handoff resulting from PU activity). The standard
sending the Route Request (RREQ) message by some kind of proactive routing protocol will react in the same manner as
reactive protocol in the CCC, just one common channel during SU node (or even whole cluster) movement using
characteristic is taken into account. But, by expanding the route inbuilt mechanisms that are typically slow. Additionally,
tables to cover the full channel usage along the entire path from typical reactions provide sizable routing signalling traffic.
the current node to the destination, the choice of channels may Secondly, the OLSR uses its own solutions for signalling and
be so chosen to minimize the number of channel switches route optimizations that are based on the multipoint relays
along the path [27]. (MPRs) selection. MPRs are selected using the criterion based
on the maximal two-hop neighbours, which cannot be a good
Some proposals exist that are tailored to specific types of criterion in CRAHNs. Such approach allows selecting the MPR
cognitive radio networks (for example [8], [28], [29]). They nodes that can perform frequent channels handoff, which can
often do not meet military requirements in the area of reliable
lead to network instability. Thirdly, clustering algorithms
path (or multiple paths) selection and effective reaction on located in data link layers employ their own neighbour
information from cognitive entities (route reconfiguration discovery mechanisms. Typical routing protocols located at the
because of dynamically changing spectrum access or spectrum network layer will then superpose the neighbour discovery
assignment policy). Especially TDMA-based radios are not
signalling on the clustering neighbour discovery traffic. These
taken into account. The set of metrics should be proposed that problems are also challenges for effective routing protocols for
are relevant from military CRAHN’s point of view and which CRAHNs.
can be effectively measured and used.
The routing mechanisms for military CRAHNs should:
PU network
 cooperate with cognitive entities located at each layer, Channels 1, 2, 3, 4, 5
 cooperate with cognitive clustering mechanisms, PU1

 support fast reaction on dynamic spectrum allocation SU cluster b

and access (spectrum awareness routing, joint path- Channel 1 SU2
SU6
channel optimization, reaction on spectrum handoff), a
SU cluster
Gateway c Channel 2
 support dynamic network topology control, SU3
SU1
 support high network survivability (route maintenance SU4
and reparation), Gateway
SU7 Gateway
 support QoS requirements, d
PU2 SU cluster
 support node mobility, Channel 3
PU3
SU5
 ensure reliable connection with infrastructure-based
networks and other CRAHNs (via gateways),
 optimise broadcast and multicast traffic and minimize
signalling overhead. Figure 1. Network composed of PUs and CR SUs
Let us consider for example the TDMA-based network
Typical reactions on network or node mobility (caused by
shown in Figure 1, where a CRAHN composed of the CR
secondary users (SUs) operates in the area of a primary user the spectrum handoff) can be changed over to solutions that
cooperate with spectrum sensing and spectrum decision
(PUs) network. The CRAHN, in addition to using channels for
its sole usage, can also use some channels allocated to PU processes and also with cognitive clustering mechanisms in
order to inform the routing protocol about the events connected
network (channel 1, 2, and 3). The CRAHN uses hierarchical
network organization, so it dynamically creates the clusters with channel switching and cluster modifications. Also
(using cognitive clustering) and allocates free channels for each selection of MPRs can be adjusted based on the cluster
cluster. To perform communication between SU1 and SU5, the organization. It seems to be reasonable that one of the MPR
routing protocol has to find an appropriate path (i.e. composed selection criterions should be the role of SU node in the
of a, b, c, and d links). For example, a standard OLSR-based network. For example, to minimize signalling overhead, the
routing protocol will recognize network topology using its best MPRs could be cluster heads and gateways (nodes that
inbuilt neighbour discovery and topology control mechanisms. operate in many channels – see Figure 1), but ones that are not
The routing mechanisms at the network layer start operating overloaded and with predictable stable channels used in the
after clustering and channel allocation procedures finish their cluster. Routing protocols can also use some work done by the
work. They can rely on channels that have already been clustering mechanisms and MAC to find the neighbours or to
allocated and they recognize the neighbours depending only on forward the frames (in some solutions, frames can be
the links characteristics (allocated channel dependent). There forwarded at data link layer without network routing). Finally,
the actions can be inverted: before selecting the new channels
(i.e. during spectrum handoff), current and predicted routes and The CRAHN topology control (TC), both at network
route requirements should be taken into account. For sure, creation phase and during network topology maintenance,
these proposals require cooperation of many cognitive entities needs ND mechanisms [9] [10] [12]. The effective ND
such as sensing, spectrum decision and spectrum mobility solutions in military TDMA-based networks require
processes, application requirements and MAC and clustering cooperation between CR MAC, cognitive clustering, and
mechanisms. Moreover, the CR nodes have to keep and cognitive routing. Typically, the routing protocols are equipped
synchronize the database with all information coming from with topology control mechanisms (i.e. using Hello protocols
cooperation with cognitive entities in the given node and and traffic control messages). In CRAHNs, an effective
information received from other nodes. Collecting of such a cognitive TC must be spread out between network and data
comprehensive database is not an easy problem. Many link layer.
effective mechanisms have to be enabled in the CRAHN to
optimize the signalling overhead. The challenges presented D. Cognitive data transport
above are not only valid for proactive routing protocols (i.e. The CRAHNs require efficient end-to-end data transport
OLSR-based). Nevertheless, such protocols are often used in control algorithms. Standard TCP or UDP protocols are not
military networks since they constantly build up the topology designed for wireless networks. Some modifications proposed
database, which can be easily extended and used for in literature can be used in typical ad hoc networks [16], but
complicated metric calculation. Similarly, in CRAHNs it can they are not efficient enough for CRAHNs. Transport layer
be used to gather all information relevant for cognitive routing protocols have limited knowledge of the network conditions in
mechanisms. between the end nodes. Standard TCP is responsible for
congestion control and data transmission rate adaptation of the
C. Cognitive topology control source nodes to the receiving possibility of the destination
Network topology control consists of two processes: nodes. It was designed for typical wide area fixed networks,
topology creation, and topology maintenance. Topology where congestion comes mainly from intermediate nodes
creation includes a process of neighbour discovery (ND). The overloading. The modifications for wireless ad hoc networks
CRAHNs that dynamically react on spectrum availability (and provide the mechanisms that are able to identify a source of
often spectrum handoff) are especially susceptible to effective data segments loss: nodes overloading or wireless links
ND. Most of the current ND mechanisms do not take into characteristics (data segments loss because of signal fading and
account radio node cognition features (i.e. sensing, spectrum interferences or nodes mobility).
management, sharing and mobility). The main challenge of
cognitive topology mechanisms is to cooperate with cognitive The CRAHNs provide new challenges in data transmission
control. The data segments can also be lost or delayed because
spectrum access and routing algorithms. Topology
maintenance (TM) processes should keep overall network of spectrum mobility (handoff) and spectrum sensing. The TCP
will react on such situation by decreasing the transmission
connectivity and reliability during military cognitive network
operation, taking also into account efficient reactions on window. Nodes could inform the source that it is a transitory
state caused by the cognitive entities. Moreover, intermediate
military unit organization (i.e. group of interests), resource
unavailability (i.e. because of jamming), network load, and nodes that are particularly engaged in cognitive procedures can
QoS requirements. also work as something like proxy nodes for TCP
transmissions. In CRAHNs, large bandwidth variations can
As already mentioned, PU activity may enforce spectrum appear in some segments of the networks according to PUs
switching for many SUs, which is known as ripple effect. After activity. Thus, the network can radically increase or decrease
this effect the network can be divided into new clusters, and a the throughput. TCP cannot adapt its transmission rate to such
set of topology control mechanisms are initiated mainly on data events in an effective way, especially in case of high data
link and network layers, resulting in packet dropping or segment Round Trip Time (RTT). Thus, new bandwidth
significant packet delay for the affected users. The cognitive estimation methods have to be elaborated, that can use CRs
network topology must react on these events by controlling the characteristics. Some example proposals for bandwidth
effective traffic flow over the unstable part of the network estimation can be found in [13] and [14].
(caused by spectrum handoff). One of the solutions can be a re-
routing mechanism that re-routes packets along an unaffected The cognitive data transport solutions for military
path to avoid packet dropping and delay [21]. Other solutions CRAHNs should propose updated or completely new
[20] are based on maintaining a backup channel for SUs that connection management and congestion control mechanisms.
can be used during spectrum handoff to send the packets, but They should take into account at least information on spectrum
these solutions are based mainly on predictable primary user sensing, spectrum change, route failure and mobility
activity that is difficult to achieve in a military environment. prediction. One of the example transport protocol for CRAHNs
Some more solutions are proposed in [19], where Centralized is presented in [15] - the authors named it TCP CRAHN. It
defines following stages: connection establishment, normal,
Robust Topology Control Algorithm (CRTCA) and Distributed
Robust Topology Control Algorithm (DRTCA) were spectrum sensing, spectrum change, mobility predicted, and
route failure. Each of these states reflects the cognitive network
elaborated. These algorithms are a mix of the algorithms
proposed in [21] and [20] with some elements that do not need behaviour and its influence on the node that has to control the
knowledge of PU activity in advance. Unfortunately, most of data transmission. Similar assumptions are taken in TFRC-CR
these proposals need beacons detections, which is not possible [17], where following states are proposed: normal, PU
detected, PU exit, slow start, resume and paused. Both TCP
in military TDMA-based CRs.
CRAHN and TFRC-CR can be taken into account during IV. INFLUENCE ON THE CONTROL CHANNEL
cognitive data transport mechanism elaboration, but there are The exchange of control information is one of the most
still many challenges in CRAHNs that must be addressed. First important characteristics of modern communication systems.
of all, the solutions should be compatible with standard TCP For example, ad hoc networks have to transmit hello messages
and UDP protocols. The data transport connections can be for joining networks and to check whether an apparently silent
initiated in fixed networks, where TCP or UDP is a standard node is still alive. Optimal routes in such a network are
protocol. Secondly, control messages should be reliably automatically determined and adapted via control messages.
delivered (TFRC-CR and TCP CRAHN provide many control Cognitive radios use control messages to negotiate frequency
messages, but they can be discarded even if sensing or changes. These examples show that such control information is
spectrum mobility is performed in some intermediate nodes). used to set up, preserve and optimize data connections.
E. Cognitive QoS management Therefore, a reliable exchange is essential for communication
Dynamic spectrum management gives a chance to increase systems.
a CRAHN network capacity, but also poses many problems in Such control information is usually transmitted in a so
QoS management [11]. Users can utilize some new available called common control channel (CCC), which is a logical
spectrum won by a cognitive radio but appropriate information channel between all or a subset of the devices in a network.
should be passed to the application to adjust traffic According to [2], a control channel can be classified as shown
characteristics or user requirements. Once the available in Figure 2.
spectrum bands are characterized, the most appropriate
spectrum band should be selected by considering the QoS
requirements and the spectrum characteristics. Unfortunately,
the spectrum characteristic can change dynamically because of
both the network parameter modifications and PU activity.
Thus, there is a need to perform both spectrum decisions and
application interactions.
The cognitive QoS management also refers to the network
management allowing appropriate network reconfiguration (i.e.
routing, bandwidth reservation) in order to control traffic
stream flows with different QoS requirements over different
parts of the network. For example, in case some applications
require so called hard QoS (i.e. VoIP), the packet flow should
avoid network parts, which have allocated channels used by a
PU network, especially if spectrum handoff is activated very Figure 2. Common control channel classification [2]
often. And on the contrary, if some application streams are not
delay sensitive, they can be sent over the nodes that won a new Basically two approaches can be differentiated: a control
spectrum from PUs. Thus, the cognitive entities in CRs should signal can either be an overlay or an underlay signal, where the
deliver required information to the applications and should also latter one is usually a UWB signal. In case it is an overlay
exploit application requirements and specifications. The signal, it can be either in-band or out-of-band, in relation to the
proposed cognitive QoS management solutions should take data communication. While an out-of-band signal is always
into account the military requirements on QoS. transmitted in a dedicated control channel, an in-band signal
can either be designed in a sequence-based manner or in a
Besides, in case of military networks, the ability to group based manner. A sequence-based CCC allocates
autonomously monitor, re-plan, and reconfigure the network channels according to a hopping sequence, while a group-based
requires knowledge and actions across different protocol CCC connects neighbouring devices with similar spectrum
layers. Network management mechanisms are therefore conditions. Sequence-based and group-based designs can also
required. One solution for tactical networks is proposed in [30]. be applied to out-of-band approaches, as e.g. conducted in
It is named Proactive and Adaptive Cross layer CORASMA [6].
Reconfiguration (PACR) framework for network monitoring,
analysis and reconfiguration in dynamic tactical networks. The In addition to that, in [2] four general design challenges for
following mechanisms were proposed: proactive network such control channels are reconsidered. The first one is control
monitoring and prediction to achieve early knowledge of channel saturation. A CCC must be designed to transmit all
node/network problem, integrated cross-layer information requested control information in due time, therefore its capacity
sharing to reduce latency caused by layered architecture, needs to be scaled sufficiently. A second challenge is
advanced graphic model to assist fault diagnosis based on the robustness to primary user activity. In the same way as a
real-time network topology and event information, and cognitive radio must not hamper the primary user’s
adaptive and dynamic network (re-)configuration. All these communication, it must be guaranteed that an appearing
mechanisms influenced the QoS, improving the network primary user does not interrupt the communication of a
mostly based on the COTS devices. Therefore, they should be cognitive radio network. The third challenge addresses the
further investigated. coverage of a CCC. As already mentioned above, one control
channel may be used only for a subset of devices, e.g. in a
clustered network. Nevertheless, it must be guaranteed that
messages can be forwarded to all nodes in the network. Finally, stream. This can either be done in a regular alternating manner
security is the fourth challenge identified. or on request. The latter case means that control data is
interlaced into the user data according to queueing theory. In
In the following, the challenges for overlay in-band, out-of- order to minimize impact of PU activity, channel hopping can
band, and underlay designs will be described, with a focus on be applied (sequence-based CCC). In case a network is
military requirements. In addition to that, the influence of
widespread, so that nodes at different ends of the network face
multi-hop connections will be addressed, and the impact on the different spectrum environment conditions, the network can be
OSI layers will be analysed. Finally, challenges for coexistence split into groups of nodes close to each other (and therefore
between networks will be investigated. observing similar spectrum availability). Such clustered
A. Out-of-band control channel networks have a group-based CCC.
An out-of-band control channel means that control As the control channel is not dedicated to a fixed
information is transmitted in a different spectrum band than the frequency, the network cannot be a primary user on that
user data. A reason for this design is that a change of frequency frequency, but will always be secondary user. Nevertheless, as
exhibits the risk of losing connection; therefore it is the used frequency is not known to devices outside the
comprehensible to only change the frequency of the user data network, from a military point of view it becomes more
channel and to have a constant control channel. Furthermore, difficult to jam the network. A further advantage is that the
there is usually much more user data traffic than control traffic, radio does not need to be retuned when changing between
which implies that the control channel can have much less control and user data. But, as control and user data share the
bandwidth. Thus, it may be possible to assign the control traffic same logical channel, an increase of control information is also
to a fixed dedicated channel, even though there is not enough a decrease of data transmission capacity. In case the used
constant free bandwidth for the user data. That makes the channel is interfered, it might not be possible to negotiate a
network primary user on the control channel while being new channel or even to initiate the common change to an
secondary user on the data channel. This either requires a already negotiated new channel. A possible solution for this
second RF front-end, which is an unusual requirement for the can be found in [4], where it is proposed to have a second
used hardware, or fixed time slots for control information receiver already tuned to a previously determined new
exchange, which presupposes a protocol supporting this and a frequency, so that the initiation of a frequency change can be
front-end which allows fast retuning and fast bandwidth conducted there. According to [5], it is recommended to use
changes. sequences with good correlation characteristics for such
A dedicated out-of-band control channel, for which the initiation messages, so they can be detected with high
network is primary user, requires a license. Another issue is probability.
how much bandwidth this control channel requires. Some C. UWB control channel
examples of control messages, like hello or routing messages,
UWB systems usually transmit their content spread over a
have already been mentioned above. The negotiation regarding large bandwidth. For short-range they have a very low energy
frequency changes requires exchanging the information about
level and allow for high data rates. The larger the bandwidth
which spectrum bands are free. As this information may and the shorter the distance, the higher data rates can be
change very often and encompass a high amount of data, an achieved. In addition to this, UWB systems are able to share
exchange protocol needs to be developed that abstracts the
the spectrum with other systems. Information is transmitted by
spectrum information without loss of precision. In [3], a generating radio energy at specific time intervals, so that UWB
method called “hard combining” is proposed for this purpose. signals consist of modulated pulses.
In this method the information about whether a channel is free
or not is reduced to binary information, which can be An UWB signal is per se less vulnerable to interference
transmitted in one bit, but that requires a unique numbering of than non-spread signals, which makes it attractive for the use in
the available channels. control channels, but this reliability is only valid for small
ranges. Nevertheless, most military communication requires
From a military point of view, such fixed and licensed medium or long-range user data transmission. However, if the
control channel exhibits vulnerability, as it can be easily density of a network is high enough, it can be possible to use
attacked, which leads to a complete breakdown of the system. UWB for multi-hop control information exchange (see also
Therefore, an out-of-band control channel can be a single- next section) while using medium or long-range waveforms for
point-of-failure (SPoF). It needs to be analysed whether a CCC direct connections to other nodes of the network, which are not
can have transmission security (TRANSEC) features like low in direct vicinity. In order not to limit the distance of
probability of intercept (LPI) and low probability of detection neighbouring nodes too much, a trade-off between range and
(LPD), e.g. by using UWB (see also Section 2.2.3). control data rate must be found.
Furthermore, it is important to transmit as little control
information as possible. D. Multi-hop control information exchange
B. Exchange of control information inside the user data Not only for networks using UWB, but also for networks
stream using other waveforms, it might be required that information is
transmitted via multiple hops. E.g. if not all nodes of a network
Instead of having a separate frequency or even a separate are in range of each other, it might be necessary to forward a
RF front-end for transmitting control information, it is also message over several hops instead of using a direct connection.
possible to send control information inside the user data As a drawback, it must be regarded that each hop introduces
latency to the transmission time of a message. Furthermore, in identical protocols as well as the capability to synchronise with
group-based networks some nodes do not even use the same each other become compulsory. Therefore, a common standard
transmission frequency. Therefore, gateway nodes are required for such CCC is necessary. In the civilian world there are
to connect the groups. These gateway nodes temporarily use already task groups dealing with this topic (e.g. IEEE 802.19
the frequency of one group and at all other time use the and 1900.2). Their work includes existing waveform standards
frequency of another group. A message passing such a node like IEEE 802.11, 802.15, 802.16h, and 802.22. As mentioned
will have to wait until the gateway changes to the other before, 802.11af and 802.22 are already aimed at using CR
frequency, which again is a source for latency. techniques.
From a military point of view the end-to-end transmission Coexistence between networks implies that those networks
time can be critical in operations. Depending on the automatically manage to share the available spectrum and to
application, e.g. frequency change messages or sensor-to- avoid interferences concerning both user and control data apart
shooter connections, there might be soft or hard real-time from the coexistence control information. Possible spectrum
requirements. Therefore, the amount of hops and the delay sharing solutions include using different frequencies, using
introduced in each hop have an impact on the usability of the TDMA on the same frequency, or achieving orthogonality on
network for those applications. It needs to be analysed how to the PHY layer.
optimally configure the network dependent on the application.
V. CONCLUSIONS
E. Impact on other OSI layers
In this paper, we have identified challenges for network
The impact on other OSI layers than the NET layer is aspects of cognitive radio while taking into account specific
mainly dependent on the selected CCC design. E.g. using requirements for military networks. We have considered
underlay requires pulse modulation on the PHY layer. But architectural features of infrastructure and ad hoc networks,
there are also further requirements which are less obvious. and we have analysed the influence on the control channel.
When using separate front-ends for control and user data, co- Some issues of civilian infrastructure-based solutions have
site aspects need to be regarded. The prioritized handling of been identified and some ideas have been proposed for use in
control messages compared to user data must be addressed military networks. Several challenges have been identified for
when only one front-end is available. Furthermore, it must be military ad hoc networks, namely cognitive routing, cognitive
analysed whether the control messages require unicast, topology control, cognitive data transport, and cognitive
multicast, or broadcast transmission. clustering. A comparison between different control channel
In addition to this, the demands of other layers on the approaches has been given, namely out-of-band control
control channel need to be considered. E.g. the amount of channel and in-band control channel. Some of the challenges
sensing data exchanged for cooperative spectrum sensing can discussed in this paper have been addressed by rethinking the
vary heavily. Therefore, the control channel needs to be sized solutions developed for classical wireless networks and by
accordingly. Especially in military networks priority is not only taking into account some new information provided by
a technical matter for achieving QoS, but must also reflect the spectrum sensing and spectrum mobility. However, more
military hierarchies of the users, indicating an impact from the research is needed to ensure efficient spectrum-aware
application layer. communication in military CRAHNs and to propose efficient
protocols.
In order to exploit the mutual influence of the OSI layers,
efficient cross-layer techniques must be implemented. In [6], REFERENCES
an architecture is presented, in which a cross-layer interface is [1] I.F. Akyildiz, W.Y. Lee, M.C. Vuran, and S. Mohanty, “NeXt
introduced that allows for exchanging messages between all generation/dynamic spectrum access/cognitive radio wireless networks:
elements of the radio (layers, sensing, cognitive manager, etc.). A survey”, Elsevier Computer Networks, vol. 50, no. 13, pp. 2127–
That facilitates e.g. the NET layer to take influence on the PHY 2159, Sept. 2006.
layer without sending a message through the MAC layer. [2] B. F. Lo, “A survey of common control channel design in cognitive
radio networks”, Elsevier Physical Communication, Volume 4, Issue 1,
F. Coexistence between networks pp. 26-39, March, 2011.
[3] I. F. Akyildiz, B. F. Lo, R. Balakrishnan, “Cooperative spectrum sensing
In NATO operations, it often occurs that several nations use in cognitive radio networks: A survey”, Elsevier Physical
different waveforms in the vicinity of each other. Today Communication, Volume 4, Issue 1, pp. 40–62, March 2011.
interferences between those systems are avoided by giving [4] D. Rauschen, S. Couturier, “Cooperative Spectrum Sensing for a Real-
each nation a set of frequencies for which they have an Time Cognitive Radio Demonstrator”, NATO STO IST-123 RSY-029
exclusive right to use, but still there are unconfirmed reports of Symposium on Cognitive Radio and Future Networks, The Hague,
interference between own forces. When CRAHNs will be NLD, May 2014.
fielded in the future, this exclusiveness will probably be [5] X. Liu, Z. Ding, “ESCAPE: A Channel Evacuation Protocol for
Spectrum-Agile Networks”, 2nd IEEE International Symposium on New
abolished for the sake of spectrum sharing. Therefore, new Frontiers in Dynamic Spectrum Access Networks (DySPAN), Dublin,
ways of avoiding interference between systems from different IRL, April 2007.
nations need to be introduced. [6] L. Rose, R. Massin, L. Vijayandran, M. Debbah, C. Le Martret,
“CORASMA Program on Cognitive Radio for Tactical Networks: High
A promising approach is that even though networks from Fidelity Simulator and First Results on Dynamic Frequency Allocation”,
different nations do not exchange user data, control information Proc. of Military Communications Conference (MILCOM), San Diego,
might be exchanged to ensure coexistence alongside each other CA, USA, November 2013.
[7]. This exchange requires some kind of CCC, and with this
[7] T. Ulversøy, “Waveform Coexistence Features”, Presentation at RTO- [26] O. Hussein, T. Saadawi, “Ant routing algorithm for mobile ad hoc
IST-104 meeting #5, October 2014. networks (ARAMA)”, Conference Proceedings of the 2003 IEEE
[8] M. Cesana, F. Cuomo, E. Ekici, “Routing in cognitive radio networks: International Performance, Computing, and Communications
Challenges and solutions”, Challenges and solutions, Ad Hoc Networks, Conference, 2003.
2010. [27] G. Cheng, W. Liu, Y. Li, W. Cheng, “Joint on-demand routing and
[9] S. Sengupta, K. P. Subbalakshmi, “Open Research Issues in Multi-Hop spectrum assignment in cognitive radio networks”, in Proceedings of the
Cognitive Radio Networks”, IEEE Communications Magazine, April IEEE ICC, June 2007.
2013. [28] J.-J. Lee and J. Lim, “Cognitive Routing for Multi-hop Mobile
[10] E. Berg, M.A Fecko, S. Samtani, C. Lacatus, M. Patel, “Cognitive Cognitive Radio Ad Hoc Networks”, Journal of Communications and
Topology Control Based On Game Theory”, The 2010 Military Networks, vol. 16, no. 2, April 2014.
Communications Conference, 2010. [29] T. Le, V. Rabsatt, M. Gerl, “Cognitive Routing with the ETX Metric”,
[11] M. Kartheek, Vinod Sharma, “Providing QoS in a Cognitive Radio 13th Annual Mediterranean Ad Hoc Networking Workshop (MED-
Network”, 2012 Fourth International Conference on Communication HOC-NET), 2014.
Systems and Networks (COMSNETS), 2012. [30] H. Zeng et al.: “From Spectrum Agility to Network Agility: Proactive
[12] T. Chen, H. Zhang, G. M. Maggio, I. Chlamtac, “Topology Management and Adaptive Reconfiguration for Reliable Communication in Tactical
in CogMesh: A Cluster-Based Cognitive Radio Mesh Network”, Proc. Networks”, 2013 IEEE Military Communications Conference
IEEE ICC 2007, pp. 6516-6521, June 2007. MILCOM, 2013.
[13] A. Capone, L. Fratta, F. Martignon, “Bandwidth estimation schemes for [31] J. Mitola, “Software radio architecture: a mathematical perspective,”
TCP over wireless networks”, IEEE Transactions on Mobile Computing IEEE Journal on Selected Areas in Communications, vol.17, no.4, pp.
3 (2), 2004. 514-538, April 1999.
[14] C. Chou, S. Shankar, H. Kim, K.G. Shin, “What and how much to gain [32] 802.11af-2013 - IEEE Standard for Information technology -
by spectrum agility?”, IEEE Journal on Selected Areas in Telecommunications and information exchange between systems - Local
Communications 25 (3), 2007. and metropolitan area networks - Specific requirements - Part 11:
Wireless LAN Medium Access Control (MAC) and Physical Layer
[15] K. R. Chowdhury, M. D. Felice, I.F. Akyildiz, “TCP CRAHN: A
(PHY) Specifications Amendment 5: Television White Spaces (TVWS)
Transport Control Protocol for Cognitive Radio Ad Hoc Networks”,
Operation.
IEEE Transactions on Mobile Computing, vol. 12, no. 4, April 2013.
[33] 802.22-2011 - IEEE Standard for Local and metropolitan area networks
[16] J. Liu and S. Singh, “ATCP: TCP for Mobile Ad Hoc Networks”, IEEE - Specific requirements - Part 22: Cognitive Wireless RAN Medium
J. Selected Areas of Comm., vol. 19, no. 7, pp. 1300-1315, July 2001.
Access Control (MAC) and Physical Layer (PHY) specifications:
[17] A. K. Al-Ali, K. Chowdhury: “TFRC-CR: An Equation-based Transport Policies and procedures for operation in the TV Bands.
Protocol for Cognitive Radio Networks“, 2013 International Conference [34] 802.22a-2014 - IEEE Standard for Information Technology -
on Computing, Networking and Communications (ICNC) Workshop on
Telecommunications and information exchange between systems
Computing, Networking and Communications. Wireless Regional Area Networks (WRAN) - Specific requirements -
[18] N. Mansoor, A. K. M. M. Islam, S. Baharun, S. Komaki, K. Wada, Part 22: Cognitive Wireless RAN Medium Access Control (MAC) and
“CoAd: A Cluster Based Ad hoc Cognitive Radio Networks Physical Layer (PHY) Specifications: Policies and Procedures for
Architecture With Broadcasting Protocol”, 2013 International Operation in the TV Bands Amendment 1: Management and Control
Conference on Informatics, Electronics & Vision (ICIEV), Dhaka, 2013. Plane Interfaces and Procedures and Enhancement to the Management
[19] J. Zhao, G. Cao, “Robust topology control in multi-hop cognitive radio Information Base (MIB).
networks”, Proceedings IEEE INFOCOM, 2012. [35] 802.19.1-2014 - IEEE Standard for Information technology -
[20] Cordeiro, K. Challapali, D. Birru, and N. Sai Shankar, “IEEE 802.22: Telecommunications and information exchange between systems - Local
the first worldwide wireless standard based on cognitive radios,” in and metropolitan area networks - Specific requirements - Part 19: TV
IEEE DySPAN, 2005. White Space Coexistence Methods.
[21] W. Feng, J. Cao, C. Zhang, and C. Liu, “Joint optimization of spec-trum [36] https://datatracker.ietf.org/doc/draft-ietf-paws-protocol, as of 2014-11-
handoff scheduling and routing in multi-hop multi-radio cognitive 19.
networks,” in IEEE ICDCS, 2009. [37] ETSI EN 301 598 V1.1.1 - White Space Devices (WSD); Wireless
[22] T. Clausen, P. Jacquet, “Optimized Link State Routing Protocol Access Systems operating in the 470 MHz to 790 MHz TV broadcast
(OLSR)”, RFC 3626, 2003. band; Harmonized EN covering the essential requirements of article 3.2
[23] C. Perkins, E. Belding-Royer, S. Das, “Ad hoc On-Demand Distance of the R&TTE Directive.
Vector (AODV) Routing”, RFC 3561, 2003. [38] M. Skjegstad, B. Ellingsater, T. Maseng, J. Crowcroft, O. Kure, “Large-
[24] Y. Liu, Z. Mi, J.-F. Zhang, X. Qu, “Improvement of ETX metric base on scale distributed Internet-based discovery mechanism for dynamic
OLSR”, 2010 International Conference on Wireless Communications spectrum allocation,” 2014 IEEE International Symposium on Dynamic
and Signal Processing (WCSP), Suzhou, 2010. Spectrum Access Networks (DYSPAN), pp. 404-415, 1-4 April 2014.
[25] E. Royer and C.-K. Toh, “A Review of Current Routing Protocols for
Ad Hoc Wireless Networks,” IEEE Pers. Commun., 1999.

View publication stats