Content uploaded by Andre Puschmann
Author content
All content in this area was uploaded by Andre Puschmann on Jan 08, 2015
Content may be subject to copyright.
SWITCH: A Multichannel MAC Protocol for
Cognitive Radio Ad Hoc Networks
Mohamed A. Kalil, Andr´
e Puschmann and Andreas Mitschele-Thiel
Integrated Communication Systems Group, Ilmenau University of Technology, Ilmenau, Germany
Email: mohamed.abdrabou, andre.puschmann, mitsch@tu-ilmenau.de
Abstract—Cognitive radio (CR) technology will empower wire-
less devices with the capabilities to dynamically exploit opportu-
nities in both licensed and unlicensed spectra. Thus, the spectrum
shortage problem that occurs due to the ever-increasing of wire-
less devices can be handled. In CR ad hoc network, a secondary
user (SU) is allowed to utilize a channel of the primary system
provided the channel is idle from primary user (PU) activity. In
this environment, the way the SU copes with a sudden appearance
of the PU is the most important feature of distributed CR-
MAC protocols. In this paper, a multichannel CR-MAC protocol,
which reacts efficiently to PU appearance, is developed. The new
protocol is named opportunistic Spectrum access WITh backup
CHannel (SWITCH). The SWITCH protocol is a decentralized,
asynchronous, and contention-based MAC protocol for CR ad hoc
networks. The proposed protocol operates over both licensed and
unlicensed spectra. In addition, the concept of backup channel
is introduced and employed to make the SU extremely robust
to the appearance of PUs. The simulation results show that
SWITCH accomplishes 91% throughput gain over other CR-
MAC protocols.
I. INT RO DUC TI O N
In the existing literature, several MAC protocols have
been extensively studied in the context of classical ad hoc
networks. However, they cannot be applied directly to CR
ad hoc networks, which have some unique characteristics
that clearly distinguish them from their classical counterparts.
Firstly, the SUs should have the ability to seek adaptively and
dynamically for opportunities in both licensed channels (LCs)
and unlicensed channels (UCs). Secondly, the SUs should react
efficiently to the sudden and the consecutive appearance of
PUs. These two characteristics make the design of an efficient
CR-MAC protocol a challenge.
An extensive survey of CR-MAC protocols has been given
in [1]. According to this survey, CR-MAC protocols can be
classified in: contention-based protocols, time slotted protocols
and hybrid protocols. Contention-based MAC protocols are
based on the classical CSMA/CA principle. In addition, one
or more features are added to these protocols to be adapted to
the CR environment. For contention-based MAC protocols,
no synchronization is needed between the SUs to access
the available channels. The DCA-MAC protocol [2] is an
example of this class. Time-slotted MAC protocols need a
global synchronization between SUs. Therefore, the time is
divided in slots for both the control and the data transmission.
Examples of this class are the Cognitive MAC (C-MAC)
protocol [3] and the Opportunistic Spectrum Access (OSA-
MAC) protocol [4]. Hybrid MAC protocols use a partially
slotted transmission, in which the control signaling generally
occurs over synchronized time slots. In addition, the fol-
lowing data transmission may have random channel access
schemes, without time synchronization. The SYNchronized
MAC (SYN-MAC) protocol [5] is an example from this class.
Most of aforementioned protocols operate over LCs only.
In [6], the authors have categorized MAC protocols for CR
into two major groups according to the way the SU copes with
the sudden appearance of the PU: 1) MAC protocols that en-
able buffering of SU connections preempted by the PU arrival,
and 2) MAC protocols that enable switching of SU connections
to a vacant channel when the SU preempted. The disadvantage
of the former group is that the SU buffers its connection even
if there is another free channel. Furthermore, it may happen
that the SU will not be able to re-establish its connection
after buffering because of continuous PU transmissions which
leads to a high delay. The disadvantage of the latter group is
the control message overhead between transmitter/receiver pair
to access the new channel. However, this problem is already
considered by the Opportunistic Spectrum Access with Backup
channel (OSAB) concept [7].
To benefit from the OSAB concept, a flexible MAC pro-
tocol, that coordinates the access to the medium (LCs and
UCs), should be developed. Therefore, the main goal of this
paper is introducing such a protocol. The proposed protocol is
called opportunistic Spectrum access WITh backup CHannel
(SWITCH) protocol. The SWITCH protocol is a decentralized,
asynchronous, and contention based MAC protocol for CR ad
hoc networks. The proposed protocol operates over both LCs
and UCs. In addition, the concept of Backup Channel (BC) is
introduced and employed to make the SU extremely robust to
the appearance of PUs.
The remainder of this paper is organized as follows. The
proposed protocol is described in details in Section II. The
cognitive cycle of the proposed protocol is presented in Section
III. The performance of the proposed protocol is evaluated by
simulation in Section IV. Next, in Section V, we present and
discuss selected results from our analysis. In Section VI, we
summarize the paper.
II. SW ITCH PROTOCO L
In this section, the SWITCH protocol is described in de-
tail. Firstly, the design features are presented. Secondly, the
assumptions are listed. Finally, the basic protocol operation is
given.
A. Design features
OSAB [7] [8] is an abstract concept that does not answer
questions such as: How does the transmitter/receiver pair
coordinate access to the available spectrum? How does the
SU cope with the sudden appearance of the PU? Thus, a
detailed MAC protocol is needed to answer the aforementioned
questions. This was a motivation for us to develop SWITCH.
For amending the first issue mentioned above, SWITCH
is a contention-based MAC protocol to coordinate the access
to the available channels. Contention-based MAC protocols
are asynchronous MAC protocols. This feature makes this
class an appropriate candidate for designing a MAC protocol
for CR ad hoc networks. In addition, this class utilizes a
Common Control Channel (CCC) as a rendezvous channel
for the exchange of control packets for the whole network.
Thus, all nodes in the networks are aware of the spectrum
availability in their vicinity. The End-to-End Reconfigurability
(E2R) project [9] has shown that CCC is very suited for CR
networks. On the contrary, time slotted protocols and hybrid
MAC protocols need synchronization among the nodes in the
network which is quite a challenge in an environment that
lacks a centralized entity.
To handle the second issue, SWITCH uses the BCs con-
cept proposed by OSAB. The BC is negotiated between the
transmitter and receiver prior to the actual data transmission.
Thus, when a PU appears (For simplicity, we assume that
PU appearance is sensed by both transmitter and receiver),
both transmitter and receiver switch to the BC without addi-
tional control messages. This minimizes the control overhead
required to find a new channel in the case of PUs appearance.
Furthermore, all nodes in the transmission range of both nodes
are informed about such a switch and therefore, the number
of data collisions is reduced.
B. Assumptions
SWITCH is developed based on the following assumptions:
•Two types of users affect the SU’s performance: PUs and
Classical Users (CUs) which are wireless devices without
cognitive radio capabilities such as devices using the
conventional standards e.g. IEEE 802.11 and Bluetooth.
•Two types of channels are assumed: a CCC and data
channels. The CCC is used as a rendezvous channel by
SUs for coordinating access to the medium. The selection
of the CCC is beyond the scope of this paper and we
assume that is statically assigned. The data channels
are of two types: LCs and UCs. The maximum number
of LCs and UCs are C1and C2respectively. The C1
channels are used as operating channels in the case of
PUs absence. In addition, the LCs are shared between
PUs and SUs with high priority for PUs to access the
channels. The C2channels are used as BCs in the case of
PUs appearance (Note: if there are no free channel from
UCs, the channel with the least PU activity is selected as
a BC). The UCs are shared between SUs and CUs with
equal priority.
•Each SU is equipped with two transceivers (TRx): The
first transceiver, TRx1, is devoted to operating over
the CCC. The second transceiver, TRx2, consists of a
Software Defined Radio (SDR) module. The SDR module
can tune to any of the available channels, LCs and
UCs, to sense for the unused spectrum and moreover
receive/transmit the SUs packets.
III. C OGN ITIVE CY CLE O F SWITCH
To facilitate the description of the proposed protocol, we
present a simplified cognition cycle for the SWITCH proto-
col. This cycle contains the following components: spectrum
sensing, spectrum allocation, spectrum sharing and spectrum
mobility. The SWITCH cognitive cycle is consistent with the
generic cognition cycle presented in [1].
A. Spectrum sensing
Spectrum sensing is an essential component of the cognitive
cycle of the SWITCH protocol. It is used to identify unused
channels regardless of the fact that these channels are LCs
or UCs. In this paper, we assume that SUs use cooperative
spectrum sensing as a spectrum sensing strategy. In this
strategy, the sensing results (i.e. available channels from LCs
and UCs) are combined from all SUs in the network. Thus,
the chance of missing signals from PUs, CUs, and other
SUs can be reduced which leads to better utilization of the
available spectrum. To achieve this goal, coordination and
cooperation between both transceivers (i.e. TRx1 and TRx2),
employed by each SU, are essential to sense available channels
and to distribute the sensing information among SUs. The
coordination between both transceivers can be done by using
SDR transceiver, TRx2, to sense one of Cchannels randomly,
say k-th channel, (1≤k≤C;C=C1+C2). Afterwards,
the SU tunes to its TRx1 to inform other neighbors about the
availability of this channel over the CCC.
B. Spectrum allocation
The accuracy of spectrum allocation process has a great im-
pact on both the network throughput and the overall spectrum
utilization. In SWITCH, there are two spectrum allocation data
structures, Neighbors Channel List (NCL), and Free Channel
List (FCL). The NCL is used by each node Xto keep record
of the channels occupied by neighboring nodes. The NCL
is constructed by listening to control messages sent on the
CCC. The data structure for the NCL can be described as
follows: NCL(i).node presents the neighboring node iof
node X.NCL(i).ch.no indicates the channel used by node
i.NCL(i).ch.index presents the type of the channel (LC or
UC). This field has two values, 1 or 2, which indicates the
channel type, LC or UC, respectively. NC L(i).time shows
how long NCL(i).ch.no will be occupied?
The FCL contains the available channels in the transmission
range of the node (i.e. channels not used by other neigh-
bors). A node updates its NCL and FCL, once it receives
a new control messages. The data structure for the FCL of
node A can be described as follows: F C L(i).ch.no presents
the channel number. F CL(i).ch.index shows the type of
the channel (LC or UC). F C L(i).ch.priority indicates the
priority of each channel to be used by the node. Each
channel may be assigned one of three priorities: L, M or H
which presents the channel always has low, moderate or high
priority to be used, respectively. The priority is assigned to
a channel according to PU and CU activities. The channel
with least PU and CU activities is given the highest priority
(i.e. F C L(i).ch.priority =H) to be the data channel. If
there are more than one channel with high priority, then
the data channel is selected randomly. After maintaining the
FCL, the next logical step is the selection of the Proposed
Data Channel (PDC) and Backup Channel (BC) for data
transmission preparation. The PDC is selected firstly from the
LCs (i.e. channels with ch.index = 1) as mentioned before.
The transmitter checks its FCL and selects the first channel
with the least PU activity (i.e. F C L(i).ch.priority =H) as
the PDC. The BC is selected firstly from the UCs (i.e. channels
with ch.index = 2). If all UCs are busy, the second channel
with the least PU activity from the LCs, is selected as a BC.
C. Spectrum sharing
In this section, we describe the spectrum sharing process of
SWITCH. First, we introduce the control packet format. Sec-
ond, we present the two handshake modes used by SWITCH.
1) Control packets format: The control packet format
of SWITCH is similar to the IEEE 802.11 packet format.
However, some modifications are added to support the CR
operation. For Request To Send (RTS): Three more fields are
added to the packet format of the original RTS: PDC, BC and
FCL fields. For Clear To Send (CTS): Two fields are added
to the packet format of the original CTS: The Selected Data
Channel (SDC) and BC fields which indicate the data channel
selected by the receiver and the BC suggested in the case of
PUs appearance, respectively. In addition, a new packet named
Notification To Reserve (NTR) is added. This packet has the
same format as CTS. The NTR is sent by the transmitter to
its neighbors only in the case that the PDC or/and BC carried
by the RTS control message is not equal to the SDC or/and
BC carried by the CTS.
2) Handshake process: SWITCH has two modes of
handshake: Two-way RTS/CTS handshake and Three-way
RTS/CTS/NTR handshake. The usage of each mode depends
completely on: 1) the channel availability in both the trans-
mitter and receiver sides, and 2) the activity of PUs, CUs and
other SUs.
Figure 1. shows an example of the spectrum sharing process.
Suppose that we have five SUs: A, B, C, D and E. Each
user constructs its FCL during the spectrum sensing process.
There are two types of channels: four LCs and two UCs.
One of the LCs (Ch.no = 1) is selected as a CCC. We
assume that two LCs (i.e. Ch.no = 2 and Ch.no = 3) and
one UC (i.e.Ch.no = 5) are available for transmission. Both
Ch.no = 4 from LCs and Ch.no = 6 from UCs are busy.
This explains why those channels are not listed in the FCL.
Ch.no Ch.index Ch.priority
1 1 -
3 1 H
2 1 M
5 2 H
FCL for node B
FCL for node C
BC
DE
ATransmitter
Receiver
PU1
CU1
Ch.no Ch.index Ch.priority
1 1 -
3 1 H
2 1 M
5 2 H
Fig. 1. Node B communicates with node C depending on FCL
To establish a communication between B and C, the nodes use
one of the previously mentioned handshake modes.
Receiver
(node C)
Neighbors
(e.g. node E)
RTS RTS
CTS CTS
Neighbors
(e.g. node A)
Transmitter
(node B)
W_DATA
W_CTS
PDC=SDC
Switch
Sense
CCCCh3 from LCs
Receive
over LC
Transmit
over LC
Idle
Backoff
(RA,..., PDC=3, BC=5)
(RA,SDC=3, BC=5) (RA,SDC=3, BC=5)
DATADATA
ACKACK
t0
t1
t2
PU
appears
DATADATA
Transmit
over UC
Ch5 from UCs
Switch
Receive
over UC
W_ACK
(RA,..., PDC=3, BC=5)
Fig. 2. MSC for RTS/CTS handshake with SU interruption
In the two-way RTS/CTS handshake mode, the normal
handshake RTS/CTS, like IEEE 802.11 MAC, is used. Figure 2
presents the Message Sequence Chart (MSC) for this mode. In
this figure, the hexagon represents the state of both transmitter
and receiver. The RTS/CTS mode is used when the receiver
(node C) agrees with the transmitter’s (node B) proposal. In
Figure 2, the proposal from node B is Ch.no = 3 as PDC
and Ch.no = 5 as BC. On receiving RTS and CTS, the
neighboring nodes (e.g. node A and node E) update their
NCL accordingly. Afterwards, both node B and node C tune
their TRx2 to Ch.no = 3 and start data transmission. If a
PU appears during the data transmission between B and C,
both nodes wait for a time, TSwitch and then switch to the
BC (Ch.no = 5) from the UCs if available. TSwitch can be
defined as the time required by the SU to sense and switch to
the BC. This process is called spectrum mobility. This process
is presented by the Switch state in Figure 3. There is no need
to inform the neighboring nodes about such a switch since they
are already informed before by listening to the RTS and CTS.
The TSwitch time should be less than DIFS. If this condition
is not satisfied then there is a probability that another SU in
the vicinity of the transmitter wins the contention and thus
TABLE I
SIM UL ATION PARAMETER S
Parameter Value
Data rate 1 Mbps
Number of LCs and UCs varies
Transmission range for PUs 150m
Transmission range for SUs, CUs 50m
RTS size 24 Byte
CTS size 16 Byte
NTR size 16 Byte
DATA size 2300 Byte
ACK size 14 Byte
SIFS 10 µs
DIFS 50 µs
TSwitch 40 µs
the switching fails. If the BC is available, both users will not
perform any additional handoff since the UCs are free from
PUs. If the BC is not available, the transmitter/receiver pair
restart the negotiation process again.
The three-way RTS/CTS/NTR handshake is used when the
receiver (node C) did not agree with the transmitter’s (node
B) proposal. For example, it may happen that the proposal
from node B is Ch.no = 3 as PDC and C h.no = 5 as BC.
However, C h.no = 3 from LCs is not included in the FCL
of node C (i.e., Ch.no = 3 is busy). Therefore, node C will
match both its FCL with the FCL of node B to select a new
data channel. Based on this matching, C selects for example
Ch.no = 2 from LCs as the SDC and sends a CTS. Based
on this change on the data channel, a NTR control message
is sent by node B to inform its neighbors. The neighboring
nodes change the tentative reservation that happened for the
PDC as a response to the RTS with the new information carried
by the NTR message. If a PU appears on Ch.no = 2, both
nodes will wait for TSwitch and after that switch to the BC
from the UCs if available and follows the same procedure like
the aforementioned example. Remark: there is a possibility
that after performing a channel switching, the transmitter or
the receiver find the BC to be busy, this may occur due to
one of the following reasons: Conflicting reservations due to
loss of control packets or a CU occupies the BC during SU
transmission since we assume a soft reservation of the BC.
Soft reservation means that the BC can be utilized by other
CUs if there is no free UC.
IV. SIM ULATI ON
As saturation throughput is a major performance measure
to evaluate MAC protocols [10], we use it as the main
performance metric. The saturation throughput means that SUs
always have data packets in their queue to transmit.
The scenario used in our simulation can be described as
follows: There are 24 SUs, 12 CUs and 12 PUs. SUs are static.
Each two SUs establish a session. We assume that each SU
has always a packet in its queue to send. The SUs coexist with
both the PUs and CUs. Each SU in this network independently
generates traffic of fixed-size packets. Table I presents the
simulation parameters. All reported results are averages over
three different runs of the simulation. Each run is equal to
transmit 30,000 SU packets on aggregate.
SWITCH is comparatively evaluated along with CR-MAC
and DCA-MAC [2]. The CR-MAC protocol is a modified
version of the IEEE 802.11 MAC protocol. The modifica-
tion mimics and supports multichannel access methods. Like
SWITCH, CR-MAC also uses a dedicated CCC for control
packets exchange while using other channels for data commu-
nications. The data channels are assigned from the LCs only
and other UCs are ignored. CR-MAC uses always two-way
RTS/CTS handshake for coordinating access to the available
channels. The FCL of the transmitter is carried by the RTS.
Upon receiving RTS, the receiver matches this list with its
own FCL and based on that the data channel is selected. DCA-
MAC [2] uses always three-way RTS/CTS/Reservation(RES)
handshake for coordinating access to the available channels. In
addition, DCA-MAC is operating only over the LCs. SWITCH
has the following features compared to the two aforementioned
protocols
•it operates over both LCs and UCs,
•it reacts efficiently to the appearance of the PUs by using
the BC’s concept.
•it is flexible to use the two-way RTS/CTS handshake
or three-way RTS/CTS/NTR handshake according to the
channels availability in both the transmitter and receiver.
V. RE SULTS
Simulation results are given as a function of the PU traffic
load since the appearance of PUs is the most important event
that affects CR ad hoc networks.
Fig. 3. Throughput of the SUs as a function of PU traffic load and using
the LCs only: C1= 12 and C2= 0
Impact of PU traffic load: The PU traffic load has a great
impact on the performance of SUs since once a PU appears
in a channel occupied by an SU, the SU should vacate this
channel and determine another free one. Figure 3 shows the
saturation throughput of the SUs using the SWITCH protocol
compared to DCA-MAC and CR-MAC vs. the PU traffic
load and using the LCs only. The number of LCs is set to
12 channels. In this Figure, the impact of the UCs is not
shown. Obviously, when the PUs traffic load increase, the
throughput for the three MAC protocols decreases however
with different levels. Although the UCs are not used here, the
performance of SWITCH outperforms the performance of the
other two protocols because of the BC concept. The throughput
of SWITCH increases compared to CR-MAC and DCA-MAC
by 91% and 19%, respectively.
7UDIILFORDGRI38V
6DWX
U
DWLRQ7K
U
RXJKSXW0ESV
6:,7&+
&
&
&
&
ORZ&8WUDIILF
&
&
KLJK&8WUDIILF
'&$&
&
Fig. 4. Throughput of the SUs as a function of PU traffic load with different
number of UCs: For SWITCH, C1= 12,C2= 0 or C1= 10,C2= 2. For
DCA, C1= 12 and C2= 0
Impact of UCs: Figure 4 shows the throughput of the SUs
using the SWITCH protocol as a function of the PU traffic
load and using both LCs and UCs. In addition, we generate
different CU traffic loads in the UCs to investigate the effect
of CUs on the SWITCH protocol. To make a fair comparison
between SWITCH and DCA-MAC, we use the same number
of channels for each protocol. However, the type of channel
will be different from one protocol to another. For DCA-MAC,
C1= 12 and C2= 0 are used since this protocol operates
over the LCs only. For the SWITCH protocol, C1= 10 and
C2= 2. The Figure illustrates that the performance of the
SWITCH protocol outperforms DCA-MAC. Furthermore, it
outperforms the performance of the SWITCH protocol when
operating LCs only. This improvement is expected within low
CU traffic since the two UCs are utilized somehow exclusively
by the SUs. On the contrary, when the PUs appears in DCA-
MAC, the SU data transmission is interrupted and a new
transmission is established. This process continues till the SU
data transmission is completed. This explains the significant
improvement on the throughput for the SWITCH protocol
compared to DCA-MAC. For a high traffic load of CUs and
using both LCs and UCs, SWITCH increases the throughput
compared to DCA-MAC and SWITCH, using the LCs only, by
91.7% and 63.5%, respectively. Interestingly, although DCA-
MAC utilizes 12 LCs and SWITCH utilizes only 10 LCs and
two highly loaded UCs, SWITCH outperforms DCA-MAC.
We can explain that as follows. Even if the two UCs are
highly loaded, there is a chance for the SUs to access the
UCs since all users in the UCs have the same priority to access
the channels. In addition, the concept of BC reduces the time
needed to establish a new connection.
VI. CO N CL USI ON
In this paper, a decentralized, asynchronous, and contention-
based MAC protocol named SWITCH, has been developed.
Using simulation, we were able to compare SWITCH with
other MAC protocols in a cohesive manner. We draw important
conclusions from our study such as: 1) A combination of
channels from LCs and UCs as a spectrum environment for
CR ad hoc devices is a better approach for efficient utilization
of the available spectrum, 2) The concept of BC minimizes
the overhead needed to maintain the SUs link in the case of
PU appearance. In future work, the SWITCH protocol will be
implemented using a SDR testbed developed at our university.
REF ER ENC ES
[1] I. F. Akyildiz, W.-Y. Lee, and K. R. Chowdhury, “Crahns: Cognitive
radio ad hoc networks,” Ad Hoc Networks Journal, vol. 7, no. 5, pp.
810–836, July 2009.
[2] P. Pawelczak, R. Venkatesha Prasad, L. Xia, and I. Niemegeers, “Cog-
nitive radio emergency networks - requirements and design,” in Proc.
of the 1st IEEE International Symposium on New Frontiers in Dynamic
Spectrum Access Networks (DySPAN 2005), Maryland, USA, November
2005, pp. 601–606.
[3] C. Cordeiro and K. Challapali, “C-mac: A cognitive mac protocol for
multi-channel wireless networks,” in Proc. of the 2nd IEEE International
Symposium on New Frontiers in Dynamic Spectrum Access Networks
(DySPAN 2007), Dublin, Ireland, April 2007, pp. 147–157.
[4] L. Le and E. Hossain, “Osa-mac: A mac protocol for opportunistic
spectrum access in cognitive radio networks,” in Proc. of the IEEE
Wireless Communications and Networking Conference (WCNC 2008),
Las Vegas, USA, April 2008, pp. 1426–1430.
[5] Y. R. Kondareddy and P. Agrawal, “Synchronized mac protocol for
multi-hop cognitive radio networks,” in Proc. of the IEEE International
Conference on Communications (ICC 2008), Beijing, China, 19-23 May
2008, pp. 3198–3202.
[6] J. Park, P. Pawelczak, and D. Cabric, “To buffer or to switch: Design
of multichannel mac for osa ad hoc networks,” in Proc. of the IEEE
International Symposium on New Frontiers in Dynamic Spectrum Access
Networks (DySPAN 2010), Singapore, 6-9 April 2010, pp. 1–10.
[7] M. A. Kalil, H. Al-Mahdi, and A. Mitschele-Thiel, “Analysis of oppor-
tunistic spectrum access in cognitive ad hoc networks,” in Proc. of the
16th International Conference on Analytical and Stochastic Modelling
Techniques and Applications (ASMTA 2009), Madrid, Spain, 9–12 June
2009, pp. 9–12.
[8] H. Al-Mahdi, M. A. Kalil, F. Liers, and A. Mitschele-Thiel, “Increasing
spectrum capacity for ad hoc networks using cognitive radios: An
analytical model,” IEEE Communications Letters, vol. 13, no. 9, pp.
676–678, September 2009.
[9] P. Cordier, P. Houze, S. B. Jemaa, O. Simon, D. Bourse, D. Grandblaise,
K. Moessner, J. Luo, C. Kloeck, K. Tsagkaris, R. Agusti, N. Olaziregi,
Z. Boufidis, E. Buracchini, P. Goria, and A. Trogolo, “E2r cognitive
pilot channel concept,” in Proc. of the 15th IST Mobile and Wireless
Communications Summit, Myconos, Greece, 4-6 June 2006, pp. 1–4.
[10] J. Mo, H.-S. So, and J. Walrand, “Comparison of multi-channel mac
protocols,” IEEE Transactions on Mobile Computing, vol. 7, no. 1, pp.
50–65, Janaury 2008.