D-MACRN: A Decentralized MAC protocol for Ad Hoc Cognitive Radio Networks

D-MACRN: A Decentralized MAC protocol for Ad Hoc Cognitive Radio Networks Sonu Kumar Mishra

Teng Joon Lim

Dept. of Electronics and Electrical Engineering Indian Institute of Technology Guwahati, Assam 781039, India Email: [email protected]

Dept. of Electrical and Computer Engineering National University of Singapore, Singapore Email: [email protected]

Abstract—Cognitive Radio has emerged as a promising technology for the efficient utilization of the radio electromagnetic spectrum and to address the problem of spectrum shortage. Having a common control channel (CCC) over which control signals can be exchanged has several drawbacks including bandwidth inefficiency, susceptibility to denial of service (DoS) attacks, and channel saturation. This paper proposes a decentralized MAC protocol which can operate in a cognitive network without any CCC. In addition, the protocol does not require any slot timing synchronization or complex channel-hopping sequence; it supports multiple rendezvous and quicker integration of a newcomer node into the network. Through extensive simulations on a MATLAB event driven simulator, we show that the proposed protocol outperforms the conventional SYN-MAC protocol for networks without a CCC, in terms of a lower time-to-rendezvous (TTR) and a higher throughput.

I. I NTRODUCTION Spectrum regulation through the conventional fixed spectrum policy often leads to inefficient utilization of the radio frequency bands [1]. It results in spectrum holes (or white spaces) in the licensed band on the one hand, and congestion in the unlicensed ISM band on the other. The concept of cognitive radio or dynamic spectrum access (DSA) [2], has emerged as a promising technology for circumventing both problems – inefficient use of licensed bands, and overutilization of unlicensed bands. In machine-type or machineto-machine (M2M) communications for realizing the Internet of Things (IoT), cognitive radio is also increasingly recognized as a practical channel access method in an environment of infrequent transmissions of short messages. The licensed users are known as primary users (PUs) and the opportunistic users as secondary users (SUs). SUs are capable of sensing the channel licensed to the PUs, and can locate the spectrum holes. On finding the spectrum holes, the SUs adjust their communication parameters and utilize the channel for their own communication. As PUs are the licensed owners of the spectrum, they can commence their transmission anytime irrespective of the usage of the channel by the SUs. SUs on the other hand are obliged to vacate the channel as soon as they detect the arrival of the licensed user of the channel, thus causing minimal or no interference to the primary traffic. In cognitive radio ad hoc networks (CRAHN), the secondary users may be dispersed over a large area, and hence the

interference to and from the PUs and the channel availability can be non-uniform across the network. Thus, it will be challenging to find a common channel which can be used for control data exchange. Moreover using a fixed thin slice of an unlicensed band as the CCC often leads to various undesirable outcomes. First, the scaling of the cognitive network causes the saturation of CCC due to the increased control message traffic. This is known as channel saturation [3]. Second, the dedicated control channel can be blocked by a malicious node deliberately flooding the CCC. This is known as a Denial of Service (DoS) attack [4]. The situation becomes worse in a Multi-Hop CRN where another problem can be a multichannel-hidden terminal problem [5]. Thus, MAC protocols for CRNs must ideally be free from any CCC requirement. The rest of the paper is organized as follows. Section II describes the prior work related to the MAC protocols without CCC. In section III we present the system model being studied. In section IV we propose our own MAC protocol, which we call Decentralized MAC for Ad hoc CRN (D-MACRN). Section V presents the simulated performance of D-MACRN, and comparison with SYN-MAC [6]. Finally, in section VI, we conclude the paper. II. R ELATED W ORK There has been a considerable amount of research on MAC protocol for CRAHNs. On the basis of a CCC, CRAHN MAC protocols can be classified into three categories: dedicated CCC (D-CCC), non-dedicated CCC (ND-CCC), and non-CCC (N-CCC) [9]. The D-CCC protocols [11–14] assume the availability of a dedicated CCC – a licensed channel or a portion of some unlicensed band, ISM for instance. Control information is exchanged by using this D-CCC, making network startup, node joining, and information exchange easier. In ND-CCC protocols [15–19] one of available channels is chosen as an ND-CCC. Large networks are divided into clusters that have different ND-CCCs. Both D-CCC and ND-CCC suffer from potential channel saturation and DoS attack problems. As D-CCC and ND-CCC protocols are relatively well understood, they are not considered in this paper. Existing N-CCC protocols are relatively fewer in number. In the Synchronized MAC (SYN-MAC) protocol [6], time is divided into slots; the number of slots is equal to the number of channels.

In each slot all the stations listen to the same channel. The secondary users contend on the channel corresponding to the present time slot. The major drawback of this protocol is that only one CR pair can start communication in one time slot. This increases the time to rendezvous and thus adversely affects the throughput. In Dynamic Hopping MAC (DH-MAC) protocol [7], each SU hops in a synchronized manner following a unique channel-hopping (CH) sequence. To communicate, the source SU changes its CH sequence to the CH sequence of the destination SU so that both of them hop to the same channel in every time interval. The control and data packets are exchanged on the common operation channel provided the channel is free. There is a need for synchronization in slot timing, and an overhead of CH sequence generation. CH sequences are generated with the help of CH sequence parameters; the CH sequence parameters of every node must also be known to every other node. In Decentralized Cognitive MAC (DC-MAC) protocol [8], the channel access is carried out by a combination of physical layer carrier sensing and historical statistics. Some prior knowledge of the primary traffic is assumed. The primary users are assumed to communicate according to a synchronous slot structure. The protocol synchronizes the SUs, and optimally decides on the channels to sense and transmit on, after including the probability of sensing errors. In Concurrent Access MAC (CA-MAC) protocol [9], the most commonly available channel is used for channel contention so that the maximum number of nodes can get the channel reservation information, while the channel which is available with the least number of nodes is selected for data transmission, in order to boost the chances of winning the contention. This protocol assumes synchronization in slot timing, and that every node knows the channel availability of all the nodes in the network. The Single-Radio Adaptive Channel (SRAC) protocol [10] is based on the cross-channel communication. Each node selects a stable receive channel among available channels for receiving data and maintains a list of channels used by the neighbors for receiving. Changes in a node’s receive channel have to be made known to the neighboring nodes through a notification packet on their receive channels. SRAC does not require network-wide synchronization in slot timing, but a new node can start transmission only after it obtains the list of receive channels of its neighbors. As can be seen from the above survey, most known decentralized MAC protocols for cognitive radio networks require synchronization of some form, such as in slot timing or in maintaining a table of channels available at a node’s neighbors. Without a CCC, achieving and maintaining synchronization will be challenging and thus such protocols may not be practical. The proposed protocol, D-MACRN, is designed to minimize the synchronization requirements, and yield shorter network setup and node admission times. In addition, it does not assume any prior knowledge of primary traffic, and there is no computational overhead of CH sequence generation.

TABLE I F REE C HANNEL TABLE AT NODE x SUs

Channel 1

Channel 2

Channel 3

Channel 4

Channel 5

SU 1 SU 2 SU 3 SU 4 SU 5 SU 6 SU 7 SU 8 SU 9 SU 10

1 1 0 0 1 0 1 0 0 1

1 1 0 0 0 1 0 0 1 1

0 1 0 1 0 1 1 1 0 0

0 1 0 0 1 0 0 1 0 1

1 0 0 1 1 1 0 0 1 0

III. S YSTEM M ODEL We consider a cognitive radio network with N channels and Ms secondary users. Each SU has only one transceiver, so it can either transmit or receive but not both simultaneously. Additionally, each SU has a sensor which can sense a finite, say Mo , number of channels simultaneously. Each SU randomly and independently selects Mo out of the N channels to sense in each frame. An SU can transmit on only one channel at a time. When the SU is transmitting, it will suspend its channel sensing function because its own transmitted signal would generally over-power any other signal, which therefore cannot be detected reliably. When the SU is idle (i.e. not transmitting or receiving), its sensor keeps sensing its selected Mo channels. When it detects that another SU is attempting to communicate with it on one of the channels, it will receive on that channel until transmission is complete. While receiving on one channel, we assume that the SU does not sense, transmit or receive on other channels. Thus, the SU can receive on only one of the Mo channels at a time. Moreover, each SU maintains a free channel table (FCT), which contains the channel-sensing results at other secondary nodes. Table 1 shows an FCT in which each row corresponds to an SU and each column corresponds to a channel. { FCT(i, j) =

1, if channel j is free at node i; 0, if channel j is busy at node i.

(1)

Each row of FCT is termed as a free channel vector (FCV). The purpose of FCT is to have a rough idea regarding how many SUs can potentially contend for a particular channel. As will be clarified later, no additional transmissions are required for updating the FCTs, and they are used only in the extended form of the protocol. The basic D-MACRN protocol does not involve FCT; it can be understood without understanding the how the FCTs are maintained. SUs do not know about the slot timings of other SUs, i.e. there is no synchronization in slot timing among the SUs, in the network being considered for the protocol. The PUs can reoccupy their channels irrespective of any ongoing secondary transmission.

START Ch_option = all

YES

NO transmission on any of its No channels?

NO

0 YES N_f ~=0

NO

YES NO NO

YES

YES YES

NO

RTS NO

CTS

DATA

ACK

YES

YES

NO

Fig. 1. A flow diagram for channel contention and data transmission by a node that has a DATA packet to transmit

IV. P ROPOSED D-MACRN P ROTOCOL In this section we present the proposed scheme, based on the system model described in the previous section. A. Network Initialization The FCTs are initialized with all ones. The proposed protocol does not require any slot timing synchronization, and the new nodes do not need any information about their neighbors. So a new node can independently enter the network and start contending for the channels immediately. Hence, network set up and new node admission are simpler than any of the existing non-CCC protocols. B. Channel Selection and Contention Suppose that a secondary node x has a packet to be transmitted to node y. It senses all the N channels, Mo at a time, and locates the channels which are free. Then it randomly selects No = min(Nf , Mo ) channels from the set of Nf free channels. The node x starts contending on all the No channels simultaneously using CSMA/CA. On similar grounds with 802.11 WLAN [21], it starts random back-off counters on all the No channels together. The moment one of the counters goes to zero, all other counters are frozen. Then it sends an

Fig. 2. A flow diagram for an idle node which either does not have any DATA packet to transmit, or is waiting for CTS, DATA or ACK packet after sending RTS, CTS or DATA packet

RTS (request to send) packet on the corresponding channel, Ci , and waits for a CTS (clear to send) packet. If Ci belongs to the set of channels which the sensors of node y are sensing, and if the node y is idle (neither transmitting nor receiving), it goes into receiving mode and starts listening to Ci . If the RTS packet is received without any error, node y transmits a CTS packet destined for node x. When node x receives the CTS from node y, the RTS-CTS handshake is accomplished, and node x commences its DATA transmission. Otherwise, the RTS packet from node x does not reach node y. In such a case, the node x discards the counter corresponding to the channel Ci and resumes the counters on the other channels. If the rendezvous is not achieved within Tround seconds after the free channels are found, the channels are assumed to be unusable for connecting to node y and the process is re-started, with node x selecting a different set of No channels for contention. The SUs which do not have any packet to transmit also hop to a different set of No channels if they do not get any RTS packet addressed to them, in Tround seconds after the free channels are found. Tround is chosen large enough so that a rendezvous can be completed within Tround if node x and node y have at least one channel common in the set of No channels they chose. Erroneous RTS packets are simply discarded.

C. Virtual Carrier Sensing and Free Channel Vector The carrier sensing is performed at both the physical layer and MAC layer. The duration field in the RTS and CTS packet indicates how long the channel will be occupied for the successful transmission of the DATA packet. Nodes, other than the intended destination, receiving the CTS packets use the embedded duration field to adjust their network allocation vectors (NAV) and refrain from using the channel until the NAV expires. This MAC sublayer sensing is referred to as Virtual carrier sensing [21]. In addition, the RTS and CTS packets in D-MACRN also carry FCVs of the transmitter. The FCV indicates the carriersensing results at the transmitter just before the transmission of the RTS or CTS packet. In the extended version of DMACRN, all the nodes receiving the RTS or CTS packets update their FCTs using the FCV contained in the packets. The SU k updates the ith row in its FCT when an RTS or CTS packet from SU i is received at k. As an example of how the FCTs are updated, assume that node a has to send a CTS to node b. There are 10 (say) channels in total, out of which 1, 3, 5, 7 are free at node a. Node a will make a 1x10 vector, called FCV, with 1s at 1st , 3rd , 5th , and 7th position, and zero otherwise. This whole vector will be embedded in the CTS packet. When node b receives the CTS, it uses the embedded FCV to update the ath row in its FCT. After this modification, the entries of ath row are not touched until a new RTS or CTS packet from node a is received at node b. Similarly, the y th row in the FCT of node x corresponds to the carrier-sensing results of node y the last time an RTS or CTS from node y was received at node x; its FCT may not reflect the current channel occupancy state at node y. So, the node x does not use its FCT directly for channel selection. It is incorporated indirectly, which is explained in section E. D. Arrival of Primary Users Since the primary users are the licensed owners of the channels they do not coordinate their transmissions with the secondary network and hence primary signals can appear at any time. We make the practical assumption that the arrival of a PU can be detected by an SU only after the latter completes any ongoing transmission. After the arrival of the primary is detected, the SU vacates the channel and the primary reoccupies its channel for its own transmission. Even if, unfortunately, an old RTS-CTS packet comes in the way of the primary transmission, being very small as compared to the primary packet size, the interference caused is negligible. If the primary arrives after the rendezvous between x and y is complete, the node x uses the FCV of the received CTS packet from node y to decide which channels to hop to next. Subsequent contentions and data transmissions take place on those channels and not on the channel that has just been sensed to be occupied. In addition, the updated FCVs allow the sensed primary activity to be propagated to nodes that are shadowed from the primary transmitter and cannot sense its activity reliably.

Fig. 1 shows the flow chart for a node that has a DATA packet to transmit. Fig. 2 shows the same for an idle node which either does not have any DATA packet to transmit, or is waiting for CTS, DATA or ACK packet after sending RTS, CTS or DATA packet. E. Proposed Extension: TF-IDF Now, we propose an enhanced scheme for channel selection. In selecting No channels out of Nf free channels, instead of giving equal probability of selection to all channels, we can give higher weight to the channels which are more likely to be usable. Term frequency-Inverse document frequency (TF-IDF) [20] is a commonly used metric in text-mining for assigning importance to the terms occurring in a huge corpus (group of documents). Term Frequency (TF) is the frequency of a particular term in a document. For a particular document, the TF of a term i is given by: No of occurrences of term i TFi = (2) maxj (No of occurrences of term j) The higher the TF of a term in the document, the higher is its importance. But the TF of commonly used pronouns and auxiliary verbs are also very high. To get rid of such unimportant terms, the concept of IDF is introduced. The IDF of a term i is given by: ( ) D IDFi = log (3) Di where D is the total number of documents in a corpus and Di is the number of documents containing the term i. A higher TFi × IDFi product implies that the term i has been used frequently but in only a subset of all documents, so it is not likely to be a commonly used unimportant word. Borrowing the above mentioned technique from [20], we derive the probability of selection of channel Ci . If we consider documents and terms analogous to the SUs and channels, we can modify the definition of TF and IDF as following: No of times Ci caused (x, y) rendezvous maxj (No of times Cj caused (x, y) rendezvous) (4) A larger TFi implies that the channel Ci has been a better fit in the past for secondary node x and y. This could be due to channel Ci being relatively unused at node x’s location for instance. Thus, intuitively this channel is more likely to successfully the present packet also. Then we define ( ) No of secondary users IDFi = log (5) No of SUs having Ci available TFi =

A larger IDFi implies that fewer secondary users are contending for the channel Ci , therefore the probability that the secondary node x will win the contention is higher. The denominator in (5) is obtained from node x’s FCT. The probability of selection of channel Ci is given as: TFi × IDFi (6) pi = ∑N j=1 TFj × IDFj

0.015

1600 1400 Throughput (packets/sec)

Time to Rendezvous (sec)

D−MACRN SYN−MAC

0.01

0.005

1200 1000 800 600 400 D−MACRN SYN−MAC

200 0

1

2

3 4 5 6 7 8 Number of channels sensed simulatneously, Mo

9

0

10

Fig. 3. Plot of time to rendezvous, in the absence of primary users, vs number of channels which the secondary nodes can contend for simultaneously

1

2

3 4 5 6 7 8 Number of channels sensed simulatneously, Mo

10

Fig. 4. Plot of average throughput, in the absence of primary users, vs number of channels which the secondary nodes can contend for simultaneously

0.05

V. S IMULATION

D−MACRN SYN−MAC

0.045 0.04 Time to Rendezvous (sec)

In this section we compare the performance of the proposed D-MACRN protocol with the conventional SYN-MAC protocol [6] for cognitive radio networks without a common control channel. The performance metric chosen for comparison are: Average throughput and Time to Rendezvous. Average throughput is defined as the number of data packets successfully transmitted to the intended destinations per unit time. Time to Rendezvous is the time taken to complete the RTS-CTS handshake between the transmitter and receiver. The simulation tool used is a MATLAB event driven simulator, which we developed ourselves. The total number of channels is 10, and there are 20 secondary nodes randomly distributed in 200x200 area. 10 nodes are randomly chosen as the sources and the other 10 as the destinations. The source nodes operate in asymptotic conditions, i.e. they always have a packet to transmit. Tround is chosen to be equal to 1.6ms, approximately twice the time required for the RTS to reach a destination, followed by a CTS to reach the source. The primary traffic model used in the simulations is: whenever a primary node transmits, its transmission time varies uniformly between 10ms and 20ms. First, we observe the performance in the absence of any primary node, i.e the channels are always free of the primary transmissions. The objective is to decide a suitable value of Mo , the maximum number of channels which can be contended for simultaneously by the SUs. For lower values of Mo , an increase in value increases the probability that the source and destination have a channel in common. But for higher values of Mo , Nf becomes the bottleneck, and thus leads to no further increase in the value of No . Hence, the performance gain for higher values of Mo is negligible. Fig. 3 shows the time to rendezvous (TTR) vs Mo , in absence of any primary transmission. The TTR decreases sharply with the increase in Mo , for the initial values of Mo ; later the decrease becomes gradual. Fig. 4 shows the average throughput vs Mo . The throughput increases sharply with the

9

0.035 0.03 0.025 0.02 0.015 0.01 0.005 0

0

0.1

0.2 0.3 0.4 0.5 Channel Utilization by primary nodes

0.6

0.7

Fig. 5. Plot of time to rendezvous, in the presence of primary users, vs the channel utilization by the primary users. SUs contend for 4 out of 10 channels simultaneously.

increase in Mo , for the initial values of Mo , but later the increase becomes negligible. Thus we choose Mo as 4; for higher values, the gain in performance is outweighed by the increase in the hardware cost. Now, we observe the performance of the network employing the two protocols, D-MACRN and SYN-MAC, in the presence of primary transmissions. Fig. 5 shows the variation of TTR vs the primary utilization of the channels. It can be observed that the proposed protocol, D-MACRN, outperforms the conventional SYN-MAC protocol. D-MACRN also yields a higher throughput as compared to SYN-MAC, Fig. 6. For the sake of clarity, the curves corresponding to the enhanced D-MACRN, the one involving TF-IDF, is not shown in these figures. We study the behavior of the extended D-MACRN wrt. the basic D-MACRN in fig. 7 separately. Fig. 7 shows the variation of TTR vs the primary utilization of the channels. It can be observed that the extended D-MACRN witnesses a slight improvement in TTR as compared to the basic DMACRN.

R EFERENCES

1500

Throughput (packets/sec)

D−MACRN SYN−MAC

1000

500

0

0

0.1

0.2 0.3 0.4 0.5 Channel Utlization by primary nodes

0.6

0.7

Fig. 6. Plot of average throughput, in the presence of primary users, vs the channel utilization by the primary users. SUs contend for 4 out of 10 channels simultaneously.

Time to Rendezvous (sec)

0.02

0.015

0.01

0.005 Basic D−MACRN Enhanced D−MACRN 0 0.1

0.2

0.3 0.4 0.5 Channel Utilization by primary nodes

0.6

0.7

Fig. 7. Plot for the comparison of performance (TTR) of the extended DMACRN wrt the basic D-MACRN protocol. SUs contend for 4 out of 10 channels simultaneously.

VI. C ONCLUSION We presented a MAC protocol for cognitive radio Ad Hoc networks without a common control channel. The proposed MAC protocol circumvents all the problems pertaining to common control channel, eg. channel saturation, Denial of Service attacks, etc. It requires no slot timing synchronization and a minimal synchronization in maintaining the FCT. A newcomer node can start communication as soon as it enters the network by choosing the channels randomly. Multiple rendezvous are possible at the same time, and hence a substantial decrease in TTR and an increase in throughput are achieved. There is no computational overhead of channel hopping sequence. The proposed protocol is compared with the the conventional SYN-MAC protocol using a MATLAB event driven simulator; it shows that D-MACRN incurs a much shorter time to rendezvous, and hence a higher average throughput than SYNMAC.

[1] Federal Communication Commission. Spectrum Policy Task Force Report. FCC 02-155, 2002, Washington, DC. [2] J. Mitola III and G. Q. Maguire Jr, “Cognitive Radio: Making Software Radios More Personal,” IEEE Personal Communication, vol. 6, no. 4, Aug. 1999, pp.13-18. [3] J. So and N. Vaidya, “Multi-Channel MAC for Ad Hoc Networks: Handling Multi-Channel Hidden Terminals Using A Single Transceiver,” in Proc. ACM Int’l Symp Mobile Ad Hoc Net. and Comput. (MobiHoc) 2004., Tokyo, Japan, May 2004. [4] K. Bian and J.-M. Park, “MAC-layer misbehaviors in multi-hop cognitive radio networks,” 2006 US - Korea Conf. Science, Tech, Entrepreneurship (UKC 2006), Aug. 2006 [5] H. Nan, S.-J. Yoo and T.-I. Hyon, “Distributed Coordinated Spectrum Sharing MAC Protocol for Cognitive Radio,” IEEE Int’l Symp. New Frontiers Dynamic Spectrum Access Networks (DySPAN) 2007, pp. 240– 249, Dublin, Ireland, April 2007 [6] Y. R. Kondareddy and P. Agrawal, “Synchronized MAC protocol for multi-hop cognitive radio networks,” in Proc. IEEE Int’l Conf. Comms.(ICC) 2008, Beijing, China, May 2008. [7] C.-F. Shih, T. Y. Wu and W. Liao, “DH-MAC: a dynamic channel hopping MAC protocol for cognitive radio networks,” in Proc. IEEE Int’l Conf. Comms (ICC) 2010, Cape Town, South Africa, May 2010 [8] Q. Zhao, L. Tong, A. Swami and Y. Chen, “Decentralized cognitive MAC for opportunistic spectrum access in ad hoc networks: A POMDP framework,” IEEE J Sel Area Comms., vol. 25, no. 3, April 2007, pp. 589–600. [9] S. K. Timalsina, S. Moh, I. Chung and M. Kang, “A concurrent access MAC protocol for cognitive radio ad hoc networks without common control channel,” EURASIP Journal Advances Signal Processing 2013, 2013:69. [10] L. Ma, C.-F. Shen and B. Ryu, “Single-radio adaptive channel algorithm for spectrum agile wireless ad hoc networks,” in Proc. IEEE Int’l Symp. New Frontiers Dynamic Spectrum Access Networks (DySPAN) 2007, pp. 547–558, Dublin, Ireland, April 2007. [11] L. Ma, X. Han and C.-C. Shen, “Dynamic Open Spectrum Sharing MAC Protocol for Wireless Ad Hoc Networks,” in Proc. IEEE Int’l Symp. New Frontiers Dynamic Spectrum Access Network (DySPAN) 2005, pp. 203– 213, Baltimore, Maryland, USA, Nov. 2005. [12] J. Jia, Q. Zhang and X. Shen, “HC-MAC: A hardware-constrained cognitive MAC for efficient spectrum management,” IEEE J Sel Area Comms., vol. 26, no. 1, Jan. 2008, pp. 106–117. [13] H. Su and X. Zhang, “Cross-layer based opportunistic MAC protocols for QoS provisionings over cognitive radio wireless networks,” IEEE J Sel Area Comms, vol. 26, no. 1, Jan. 2008, pp. 118–129. [14] B. Hamdaoui and K. G. Shin, “OS-MAC: An efficient MAC protocol for spectrum agile wireless networks,” IEEE Trans. Mob. Comp., vol. 7, no. 8, August 2008, pp. 915–930. [15] C. Cordeiro and K. Challapali, “C-MAC: A cognitive MAC protocol for multi-channel wireless networks,” In Proc. Int’l Symp. New Frontiers Dynamic Spectrum Access Networks (DySPAN) 2007, pp. 147–157, Dublin, Ireland, April 2007. [16] C.-S. Hsu, Y.-S. Chen and C.-E. He, “An efficient dynamic adjusting MAC protocol for multichannel cognitive wireless networks,” in Proc. IEEE Int’l Conf. Wireless Comm., Networking and Info. Security (WCNIS) 2010, Beijing, China, June 2010. [17] G. P. Joshi, S. W. Kim and B.-S. Kim, “An efficient MAC protocol for improving the network throughput for cognitive radio networks,” in Proc. Intl. Conf. Next Gen. Mob. Apps., Services Techno.(NGMAST) 2009, Cardiff, Wales, UK, Sept. 2009. [18] T. Chen, H. Zhang, G. M. Maggio and I. Chlamtac, “CogMesh: A cluster-based cognitive radio network,” in Proc. IEEE Int’l Symp New Frontiers Dynamic Spectrum Access Networks (DySPAN) 2007, pp. 168– 178, Dublin, Ireland, April 2007. [19] J. Zhao, H. Zheng and G.-H. Yang, “Distributed coordination in dynamic spectrum allocation networks,” in Proc.IEEE Int’l Symp New Frontiers Dynamic Spectrum Access Networks (DySPAN) 2005, pp. 259–268, Baltimore, Maryland, USA, Nov. 2005. [20] A. Rehman, H. A. Babri and M. Saeed, “Feature extraction algorithms for classification of text documents,” Int’l Conf. Comput. and Info. Tech. (ICCIT) 2012, Al-Madinah Al-Munawwarah, Saudi Arabia, March 2012. [21] A. L. Garcia and I. Widjaja,“Communication Networks: Fundamental Concepts and Key Architecture”

The network we chose was such that all SUs could sense the presence of PUs with the same probability. Therefore, the enhanced D-MACRN could not produce any advantage over Basic D-MACRN. If the protocol is simulated in a network in which PUs have different probabilities of getting detected at different SUs, the Enhanced D-MACRN will produce better results.