A Cluster-based Energy Efficient MAC Protocol for ... - Semantic Scholar

1 downloads 0 Views 2MB Size Report
There are many MAC protocols that have been developed for wireless sensor networks. Typical examples include the time division multiple access (TDMA), ...
Globecom 2012 - Ad Hoc and Sensor Networking Symposium

A Cluster-based Energy Efficient MAC Protocol for Multi-hop Cognitive Radio Sensor Networks Yanchao Xu, Chengyu Wu, Chen He, Lingge Jiang Department of Electronic Engineering Shanghai Jiao Tong University, China Email: {very, jerry916, chenhe, lgjiang}@sjtu.edu.cn

Abstract—In this paper,we concentrate on the design, implementation and performance evaluation of the medium access control (MAC) for the multi-hop sensor networks. Being operated in unlicensed bands, the battery-operated sensor nodes need to sense available spectrum and consume precious energy. Taking account of the cognitive radio as well as the traditional characters of the wireless sensor networks, we design a MAC protocol for the cognitive radio sensor network (CRSN) that supplies schemes for solution. We propose a cluster-based protocol (KoN-MAC) for the battery-operated sensor nodes in multi-hop networks, which allows sensor nodes to dynamically select an interference free channel for data communication. We evaluate the performance of the proposed protocol by NS-2 and the simulation results show that our protocol can achieve better energy gains as well as higher throughput, less delay and lower packet loss radio.

I. I NTRODUCTION Wireless sensor networks (WSN) are being used in many applications,such as battlefield surveillance,environment monitoring etc [1]. WSN normally consists of a large number of distributed nodes that organize themselves into a multi-hop wireless network, while each node is embedded processors and low-power radios and is normally battery operated. These nodes coordinate to perform a certain required task. There are many MAC protocols that have been developed for wireless sensor networks. Typical examples include the time division multiple access (TDMA), code division multiple access (CDMA), and contention-based protocols like IEEE 802.11 [2] and IEEE 802.15.4 [3]. There are some attributes that should be considered in designing a MAC protocol for WSN. As the battery-operated nodes are often very difficult to change or recharge, the most important attribute is the energy efficiency. Another important attribute is that the protocol should be fit the topology of the WSN, which here means that the protocol should be adapted to the multi-hop WSN ,instead of the networks with just few nodes. Other important attributes include the throughput, latency and packet loss etc, just as in other wireless networks. Cluster-based MAC protocols can be a good solution to satisfy those requirements. Authors in [4] examine some typical proposed clustering algorithms for WSN and briefly discuss the operations of these algorithms, such as the LCA, LCA2, LEACH and HEED etc. The paper provides a basis for research in clustering schemes for WSN. Authors in [5] propose a clustered on-demand multi-channel MAC protocol (COM-MAC) for wireless multimedia sensor networks. The

978-1-4673-0921-9/12/$31.00 ©2012 IEEE

operation of proposed protocol consists of three sessions to support energy-efficient, high-throughput, and reliable data transmission. As a solution to allow WSN work in unlicensed bands with the primary users, the cognitive radio (CR) technology can be considered to make the sensor nodes access available spectrum [6]. Thus, besides those traditional requirements stated above, the cognitive radio technology should also be considered for the design of a MAC protocol for the cognitive radio sensor network (CRSN). Compared with the number of the traditional MAC protocols, there are fewer MAC protocols for CRSN. Taking account of both the energy efficiency and primary user (PU) protection, authors in [7] propose a new channel management scheme which heightens the energy efficiency while not greatly disturbs the PU. The CRSN with the proposed scheme adaptively selects its operation mode according to the channel-sensing outcome. There are also some schemes proposed in literatures [8]–[10] to mainly improve energy efficiency in CRSN. Authors in [8] propose an energy-saving sensing architecture for multi-resolution spectrum sensing, and the scheme in [9] clusters the sensor nodes to reduce the energy consumption during the sensing-result reporting. The channel-sensing scheme proposed in [10] saves energy by choosing the optimal sleep period and sensing parameters. In this paper, combined with the traditional WSN’s sleep mechanism, we mainly focus on the research on the multihop CRSN. A medium access scheme is designed for nodes in every cluster to provide a contention free communication. And a channel-choosing scheme is also proposed to lower the energy consumed in successfully transmitting a packet. By modeling mathematically our scheme, we prove the solution of the scheme follows the Principle of Optimality of the Dynamic Programming and decreases the computational complexity of the solution. We will show our proposed protocol can reduce the energy cost and extend the life of the networks as well as achieve higher throughput, less delay and lower packet loss radio. This paper is organized as follows: in section II, we describe the problem and present the model of the network; in section III, we propose our protocol in detail and mathematically analysis our proposed protocol; in section IV, we evaluate the performance of the proposed protocol through simulation experiments using NS-2. Finally, we conclude in section V.

537

II. S YSTEM M ODEL A. Characters of WSN Equipped with a low-level and low-power wireless communication module, a large number of sensor nodes form a multi-hop network that constitutes a WSN in order to deliver collected data to a resource center or a sink. Battery-operated sensor nodes all have their own sleep-wake strategy,and sensor nodes enter into sleep mode to save energy consumption every a certain period, while the nodes leave the sleep mode when to send or receive packets

Sink P2

CH CG

Fig. 1: The structure of cluster-based WSN

B. Multi-channel Access There are different methods to select an available channel, and we category the methods of channel access into two kinds, a)without control channel, which is called the static channel allocation. b)with control channel, which is called the dynamic channel allocation. In a multi-channel network, there exists the traditional hidden terminals and exposed terminals problems, and there are also some other new problems, such as the multichannel hidden terminals [11]. It is caused by the fact that the nodes equipped with a single transceiver can only either transmit or receive on the control channel or the data channel while cannot transmit and receive at the same time.

Based on the description of the previous section, the main problem can be summarized as follows: with the presence of primary users, secondary users can effectively access the channel and communicate with each other, while alleviating the multi-hop multi-channel hidden terminal interference and minimizing the energy consumption. Here we can make a classification and model of those problems, as shown in Table II. TABLE II: The model of CRSN Problem Node energy consumption

C. Multi-hop Networks Without considering spatial differences, wireless sensor network can be mathematically abstracted into a graph with the vertex nodes, which can be called as a flat structure. As the scale of the network grows larger, network performance will decline rapidly, authors in [12] theoretically analyzed the wireless multi-hop network capacity issues, and concluded that the capacity of wireless multi-hop network is inversely proportional with the number of the sensor nodes. And with the number of nodes in wireless sensor networks increasing, the network management and control costs will gradually increase, thereby the overall network performance will be greatly declined. D. Cognitive Radio The presence of primary users will affect the performance of the secondary users (SU), and there are quite some problems to make the secondary users opportunely access the spectrum without affecting the primary users. Table I shows the cognitive radio issues that are needed to be addressed in this paper. Hereafter, unless otherwise specified, the second user and the wireless sensor nodes are the same in concept. TABLE I: Characters of Cognitive Radio Characteristics Cooperative sensing Notification of PU Dynamic Channel Selection Channel-switching Mechanism

Description To improve the reliability of spectrum sensing, decision fusion on the outcomes collected from neighbor nodes is performed. Once a PU is detected, the SUs need to communicate the signal of PU with each other. Channels are selected in adaptation to channel availability for packet transmission When a channel is re-occupied by a PU, the SU activities must be able to switch to a backup channel.

Multi-hop and Multi-channel

Channel Poll and Selection

Model Wireless sensor nodes enter into sleep mode in a period to save energy consumption, and cognitive wireless sensor network nodes also need to leave the sleep mode when to sense channel. Therefore, nodes need their own sleep-wake strategy. From the aspect of multi-hop,the essential reason of multi-channel hidden terminal problem is that the two-hop neighbor nodes choose the same channel. Therefore, a good coordination between the nodes can effectively alleviate the problem. It consumes energy during channel poll. To select the best available channel as well as minimum the energy consumption, a reasonable channel poll and selection process should consider the node sleep-wake strategy (Joint DCS and Sleep-Wake Strategy).

III. D ESIGN OF THE KO N-MAC P ROTOCOL Figure 1 is a cluster-based structure and there have been quite a number of researches on the cluster-based protocol in WSN, and we select a simple cluster algorithm shown in [13]. In the cluster-based structure, a certain node will be elected as a cluster head (CH), and communication between adjacent cluster nodes will be conducted by the gateway nodes (CG), and the rest of the nodes in the cluster are cluster members (CM). For traditional WSN, there are some wellknown clustering protocols, such as smallest ID clustering algorithm [13],LEACH [14],TopDisc [15] etc. As shown above, multi-channel hidden terminal problem is essentially caused by the two-hop neighbor nodes. While in the cluster-based structure, as shown in Figure 1, two-hop neighbor nodes will be either within the same cluster, or in the adjacent cluster. There is no multi-channel hidden terminal problem within a same cluster, as the communication within the same cluster is scheduled by the CH. And the adjacent clusters try to select different channels to ease multi-channel hidden terminal problem. Our proposed protocol is schedule-based protocol, which will use split phase. Figure 2 is a complete superframe interval

538

of a CH, which is composed of channel sense and selection phase (CSSP), channel schedule phase (CSP), data transmission phase (DTP) and sleep phase (SP).The superframe structure of the other nodes is a little different from CH’s, which will be introduced later in detail. Interval

the rest nodes of the cluster. The latter one is prepared for the cooperative sensing, which means that the intra-cluster members transmit their sensed results to the CH and CH makes the fusion of those results. The cooperative sensing is performed by the data fusion method. In a cluster of M members, if every node’s sensed result is Di (i = 1, 2, . . . M ), the final result is performed by the data fusion Ψ, D = Ψ(D1 , D2 , . . . DM ).

CSP

CSSP

DTP

SP

Fig. 2: The structure of cluster-based WSN

A. Channel Weight Before the introduction of the proposed MAC protocol, the concept of channel weight will be introduced here. As is shown above, the multi-channel hidden terminal problem is caused by the fact that two-hop neighbor nodes select the same channel and the cognitive radio is that different channels have different usage probabilities for SUs with the presence of the PUs. Therefore, the channel weight can be used to distinguish the channels and make the nodes select the better channel. Every node shall maintain its own channel weight table and the change of the channel weights is caused by the channel’s states sensed by the nodes, as is shown in Table III. TABLE III: The States of Channel State

Description

Idle Busy Communication Collision

Happens in CSSP i.e. the SUs find the channel available to access Happens in CSSP i.e. the SUs find the presence of the PUs in the channel Happens in DTP i.e. SUs use the channel to transmit data successfully Happens in DTP i.e. PUs or other SUs appear when a SU is transmitting

The channel weight w will change differently as the state of the channel changes. If channel is idle, w will be increased by Widle . If busy, w will be decreased by Wbusy . If communication, w will be increased by Wcom . If collision,w will will be decreased by Wsec or Wpri respectively. With the help of channel weights, the adjacent clusters can easily select different channels to transmit data, which mitigates the multichannel hidden terminal problem.

General channel sense protocols are to sense all the N channels. To decline energy consumption, the period of the sleep mode of the nodes should be longer. Thus it is reasonable to propose a Joint DCS and Sleep-Wake Strategy to make the nodes select the best channels as well as minimum the energy consumption. Here we adopt the method that nodes just sense a subset of the N channels set N . The subset is called the Polled-Channel Set S, in which k S k= K, (K ≤ N ) and nodes can just sense the channels listed in S. And it is just the reason why we call our proposed protocol as KoN-MAC. Here a channel can be denoted by (chi , wi ), i = 1, 2 . . . N , in which wi is the channel weight of channel chi . Assume that the PU on any channel follows i.i.d. λ poisson process, then during the process of channel sense, the available probability of a certain channel for SUs is pi = eλTp ,in which Tp is the duration of the single channel sense slot. And then the channel can be denoted by (chi , wi , pi ), i = 1, 2 . . . N . We use A to denote the set of available channels after the whole process of channel sense, then k P (k A k= k) = CN

j=k Y

plj

j=1,lj ∈N

Y

(1 − plm ).

(2)

lm ∈N

Let KE be the expectation of the number of the available channels, then

KE =

=

k=N X k=1 k=N X k=1

kP (k A k= k) k kCN

j=k Y j=1,lj ∈N

(3) plj

Y

(1 − plm ).

lm ∈N

In KoN-MAC, different from the general channel sense protocols, a node can get available channels just from sensing the channels in S instead of sensing all the channels in N , which decreases the number of sensed channels by N − K. To get such a S, a channel-sensing strategy is proposed hereafter. Let SK be the set that consists of K channels,

B. Channel Sense and Selection Phase (CSSP) CPSP Channel Poll

Tansmitting slot

(1)

Tp Single channle sense slot

Fig. 3: The slot structure of CSSP

As is shown in Figure 3, CSSP is composed of transmitting slots and channel-sensing slots.The former transmitting slot is used by the CH to transmit its own channel weight table to

SK = {(ch1 , p1 , w1 ) . . . (chi , pi , wi ) . . . (chK , pK , wK )}. (4) In the channel-sensing strategy, SK is composed of K steps of channel selection, among which the i-th step of channel selection is (chi , xi ),in which chi is the selected channel and xi represents the effect of the (i − 1)-th step of channel selection on the i-th one. Here the energy consumption of a certain channel is calculated in the way of transmitting a

539

packet. Assume the energy required by every single channel poll is Ep , and a node’s energy of transmitting packet on the selected channel (chi , wi , pi ) is ETr .As for ETr , the node cannot promise successfully transmit data because of the presence of PU or other SUs, then ETr should be a function of pi and wi i.e.ETr (pi , wi ). As the definition of SK , those channels that are sensed earlier (assume that they are also sensed idle) are more easily used by nodes, i.e. the order of the channels sensed in SK reflects the priority of the channels. Therefore, the i-th step of channel selection strategy is actually effected by the (i − 1)-th step, which is denoted by Ee (xi−1 ).Then the definition of the energy consumption of the i-th step strategy is, ei (chi , xi ) = Ep + Et (pi , wi ) + Ee (xi−1 ).

(5)

Therefore, the energy consumption of the whole channelsensing strategy SK , E1,K [SK , x1 ] =

K X

ei (chi , xi ).

(6)

The subscript of E1,K [SK , x1 ] means the process of the first step of channel-sensing strategy to the K-th one. As for the ∗ , the follow equation is satisfied, optimal SK ∗ E1,K [SK , x∗1 ] = min E1,K [SK , x1 ]. ∀SK ∈N

(7)

∗ is enumeration, which A direct way to get the optimal SK K needs CN K! times of calculations. To decrease the computational complexity, equation (7) needs to be resolved in another way. Firstly, the follow equation can be got from (6),

E1,K [SK , x1 ] =

Tcs

Ttr

Fig. 4: The slot structure of CSP and DTP

(chi , pi , wi ) listed in A, the channel (chi , pi , wi ) is still available with the probability Pki = e−λi (Tcs +(k−1)Ttr ) as the PU may appear in the k-th Ttr slot.It is better that if the kth node select a channel with higher Pki . Taking account of the channel weights, we modify the set A in the following way. The channel in A is denoted by (chi , wi Pki ),in which P wi = wi / wj ∈A wj is the mean of the weighted values of the channel weights.Then the channels in A is sorted by the value of wi Pki . IV. SIMULATION

i=1

K X

DTP

CSP

In this section, we evaluate the performance of the proposed cluster-based KoN-MAC through simulation experiments using NS-2. We investigate the performance in terms of network throughput, packet loss, transmission delay and the average energy consumed in transmitting a packet (average energy per packet). The performance of KoN-MAC is compared with a protocol (CR-COM),which is transformed from [5] by adding the channel-sensing process into COM-MAC. At the same time, the performance of KoN-MAC is also compared in different channel-sensing strategy sets SK with different values of K. TABLE IV: Simulation Parameters

ei (chi , xi )

i=1

= e1 (ch1 , x1 ) +

K X

(8) ei (chi , xi )

i=2

= e1 (ch1 , x1 ) + E2,K [SK /S1 , x2 ]. With the help of (8), (7) can be resolved as (9). It can be concluded that the solve of (9) follows the Principle of Optimality of the Dynamic Programming(DP ). The definition of DP is that a problem is said to satisfy the Principle of Optimality if the sub-solutions of an optimal solution of the problem are themselves optimal solutions for their sub-problems correspondingly. Therefore, the solution of the ∗ optimal SK can be conducted with the DP algorithm, which introduces a recurrence algorithm and decreases the computational complexity. C. CSP and DTP The CSP is the phase that CH allocates channels and transmitting data slots for the rest nodes of cluster. As for CH, it is transmitter, and for others, it is receiver. The DTP is a phase that non-CHs transmit data in their own slot. Figure 4 is the slot structure of CSP and DTP. If the k-th node of the cluster (the k-th Ttr duration slot is allocated in DTP correspondingly) will use channel

Parameter

Value

Channel occupancy Channel number Channel capacity Time for one channel sense Power in receive mode Power in channel sense Power in transmit mode Node transmission range CBR packet size Time for one simulation run Queue length

0.4/0.5/0.6 10 2Mbps 0.4ms 58.9mW 58.9mW 46.5mW 250m 210Byte 30s 50

PUs in different channels have different channel usage patterns which follows i.i.d ON/OFF random process [16]. There are three kinds of channel usage probabilities 0.4/0.5/0.6 that a certain channel is occupied by PUs. We assume there are 10 channels in all, including one common control channel and nine data channels, and the capacity of each channel is 2Mbps, while it takes 0.4ms to sense a single channel [17]. The energy consumption model in [18] is used. Each source node generates and transmits a constant-bit rate (CBR) data stream and the packet size is 210Byte. The maximum of data packet queue length in each node is 50. Hereafter, unless otherwise specified, each simulation run is performed for the duration of 30 seconds. Table IV lists the basic parameters in the simulation.

540

∗ E1,K [SK , x∗1 ] = min E1,K [SK , x1 ] ∀SK ∈N

= = =

Fig. 5: Throughput for various data rate

min

{E1,i [Si , x1 ] + Ei,K [SK /Si , xi ]}

min

{

min

{E1,i [Si , x1 ] +

∀SK ∈N ,∀i∈(1,K)

min

∀SK ∈N ,∀i∈(1,K) ∀SK /Si ∈N ∀SK ∈N ,∀i∈(1,K)

{E1,i [Si , x1 ] + Ei+1,K [SK /Si , xi ]}} min

∀SK /Si ∈N

Fig. 6: Packet loss for various data rate

Figure 5 shows the throughput performance of CR-COM and KoN-MAC(K=1,3,5,7,9) for diffrent data rate. It is clear that the throughput of both protocols increase as the data rate increases, and increase to a maximum at last. As to KoNMAC(K=1,3,5,7,9), the saturated throughput of KoN-MAC outperforms CR-COM, which is caused by two reasons: a) the total channel-sensing time of KoN-MAC is determined by the value of K, meaning that except K=9 , the channel-sensing time of KoN-MAC is less than the CR-COM. Therefore, the interval of KoN-MAC is shorter than CR-COM; b) instead of just supplying a single available channel in CR-COM, KoN-MAC can supply several available channels, which enables a certain node can transmit a packet more successfully during the DTP. For example, in KoN-MAC(K = 1), there are no other backup available channels in both protocols, however, the channel-sensing time and the whole interval of KoN-MAC is about 1/9 and 3/4 of CR-COM respectively. And figure 5 just shows the saturated throughput of KoNMAC (K=1) is about 1.3 times of CR-COM. As to KoNMAC(K=3,5,7,9), the saturated throughput outperforms the KoN-MAC(K=1). Among those KoN-MAC (K=3,5,7,9) , the saturated throughput increases as K decreases. This is because any value of K (K=3,5,7,9) can provide a channel-sensing strategy set for node to transmit successfully packet during its DTP. The saturated throughput of KoN-MAC with K=3 is best, indicating the KoN-MAC(K=3) can supply a channelsensing strategy set as good as those with K=5,7,9 while its channel-sense time is much shorter. Figure 6 shows the packet loss performance of CR-COM and KoN-MAC(K=1,3,5,7,9) for different data rate. As in Figure 5, Figure 6 shows that the packet loss ratio of KoN-MAC (K=1,3,5,7,9) outperforms that of CR-COM, of which the reason is approximately the same to the one described above. The

(9)

Ei+1,K [SK /Si , xi ]}

Fig. 7: Average delay for various data rate

packet loss ratio of CR-COM and KoN-MAC(K=3)increases dramatically when the data rate is larger than 80kps and 170kps respectively. The value of K has approximately the same effect on packet loss ratio as on network throughput, which shows that except K=1(no backup available channels) the packet loss ratio of KoN-MAC decreases as K decreases. Figure 7 shows the average delay performance of CRCOM and KoN-MAC(K=1,3,5,7,9) for various data rate. As in Figure 5, Figure 7 shows that the average delay of KoNMAC(K=1,3,5,7,9) outperforms CR-COM, which is mainly caused by the length of interval determined by the channelsensing time and the probability of transmitting successfully a packet. The average delay of CR-COM increases dramatically at the data rate that is larger than 80kps and stays around 0.45s at last. Except KoN-MAC(K=1), the average delay of KoNMAC decreases as K decreases. As to both of the protocols, the average delay will reach its max value when data rate is larger than a certain value, which is because of the packet queue length of any node is finite (the maximum is 50).

541

Fig. 8: Average energy per packet for various data rate

evaluation shows that our proposed KoN-MAC can achieve better energy gains as well as higher throughput, less delay and lower packet loss. ACKNOWLEDGMENT The research was supported by the National Natural Science Foundation of China Grant No. 60832009, 61172067, and Fujitsu Research Development Center Co. Ltd.. R EFERENCES Fig. 9: Average energy per packet for various data rate

Figure 8 shows the average energy per packet performance of CR-COM and KoN-MAC(K=1,3,5,7,9) for various data rate. It shows that when the data rate is low (lower than 40kps), the average energy per packet of KoN-MAC is much lower than CR-COM. When the data rate increases, the average energy per packet of the both protocols will stay around a stable value. To observe the difference more clearly, we zoom out part of the Figure 8 (the part whose data rate is higher than 40kps) and get Figure 9. Figure 9 shows that the average energy per packet of CR-COM is the highest while the average energy per packet of KoN-MAC is lower than that of CR-COM and decreases as K decreases, indicating that the average energy per packet of KoN-MAC(K=1) is the lowest (about 0.065mw). For KoN-MAC(K=3) whose performance of network throughput, packet loss and average delay is the best, its average energy per packet is around 0.068mw, which is about 1.04 times than that of KoN-MAC with K=1. This is because of the different channel sensing strategy sets used in KoN-MAC with different value of K have a great effect on the overhead during the CSSP, among which a smaller channel sensing strategy set will introduce less overhead and consume less energy. Considering all of Figure 5, 6, 7, 8, and 9, the comprehensive performance of KoN-MAC(K=3) is better. Its saturated throughput is 1.03 times of KoN-MAC(K=5) and 1.99 times of CR-COM. Its maximum average delay is 0.938 times of KoN-MAC(K=5) and 0.489 times of CR-COM. Its average energy per packet is 1.04 times of KoN-MAC (K=1) and 0.648 times of CR-COM. V. C ONCLUSION We have proposed a cluster-based CRSN MAC protocol (KoN-MAC) for the multi-hop cognitive radio wireless sensor networks. We first design a medium access scheme for nodes in every cluster to supply a contention free communication, which is specified for the multi-hop and multi-channel WSN. Then, we design a channel-sensing scheme to lower the energy consumed in transmitting a packet, which is based on the fact that the CR sensor nodes will consume higher energy if CR sensor nodes sense more channels. Our proposed scheme makes CR sensor nodes just sense fewer channels to get the available channels. We model and analysis mathematically our scheme, and evaluate the performance using NS-2. The

[1] Q. Li, C. Li, and J. Shi, “Research and application on applicationoriented event-based middleware of wsns,” in Proc. Second Int Networks Security Wireless Communications and Trusted Computing (NSWCTC) Conf, vol. 2, 2010, pp. 498–501. [2] B. Alawieh, Y. Zhang, C. Assi, and H. Mouftah, “Improving the performance of power-aware multi-rate ieee 802.11 in multihop wireless networks,” IET Communications, vol. 5, no. 6, pp. 775–788, 2011. [3] T. R. Park, T. H. Kim, J. Y. Choi, S. Choi, and W. H. Kwon, “Throughput and energy consumption analysis of ieee 802.15.4 slotted csma/ca,” Electronics Letters, vol. 41, no. 18, pp. 1017–1019, 2005. [4] O. Boyinbode, H. Le, A. Mbogho, M. Takizawa, and R. Poliah, “A survey on clustering algorithms for wireless sensor networks,” in Proc. 13th Int Network-Based Information Systems (NBiS) Conf, 2010, pp. 358–364. [5] C. Li, P. Wang, H.-H. Chen, and M. Guizani, “A cluster based ondemand multi-channel mac protocol for wireless multimedia sensor networks,” in Proc. IEEE Int. Conf. Communications ICC ’08, 2008, pp. 2371–2376. [6] O. Akan, O. Karli, and O. Ergul, “Cognitive radio sensor networks,” IEEE Network, vol. 23, no. 4, pp. 34–40, 2009. [7] J. A. Han, W. S. Jeon, and D. G. Jeong, “Energy-efficient channel management scheme for cognitive radio sensor networks,” IEEE Transactions on Vehicular Technology, vol. 60, no. 4, pp. 1905–1910, 2011. [8] S. Izumi, K. Tsuruda, T. Takeuchi, H. Lee, H. Kawaguchi, and M. Yoshimoto, “A low-power multi resolution spectrum sensing (mrss) architecture for a wireless sensor network with cognitive radio,” in Proc. Fourth Int Sensor Technologies and Applications (SENSORCOMM) Conf, 2010, pp. 39–44. [9] W. Xia, S. Wang, W. Liu, and W. Chen, “Cluster-based energy efficient cooperative spectrum sensing in cognitive radios,” in Proc. 5th Int. Conf. Wireless Communications, Networking and Mobile Computing WiCom ’09, 2009, pp. 1–4. [10] S. Maleki, A. Pandharipande, and G. Leus, “Energy-efficient distributed spectrum sensing for cognitive sensor networks,” IEEE Sensors Journal, vol. 11, no. 3, pp. 565–573, 2011. [11] J. S. Pathmasuntharam, A. Das, and A. K. Gupta, “Primary channel assignment based mac (pcam) - a multi-channel mac protocol for multihop wireless networks,” in Proc. WCNC Wireless Communications and Networking Conf. 2004 IEEE, vol. 2, 2004, pp. 1110–1115. [12] P. Gupta and P. R. Kumar, “The capacity of wireless networks,” IEEE Transactions on Information Theory, vol. 46, no. 2, pp. 388–404, 2000. [13] C. R. Lin and M. Gerla, “A distributed architecture for multimedia in dynamic wireless networks,” in Proc. IEEE Global Telecommunications Conf. GLOBECOM ’95, vol. 2, 1995, pp. 1468–1472. [14] M. Avvenuti and A. Vecchio, “Adaptability in the b-mac+ protocol,” in Proc. Int. Symp. Parallel and Distributed Processing with Applications ISPA ’08, 2008, pp. 946–951. [15] B. Deb, S. Bhatnagar, and B. Nath, “Multi-resolution state retrieval in sensor networks,” in Proc. First IEEE Sensor Network Protocols and Applications 2003 IEEE Int. Workshop, 2003, pp. 19–29. [16] X. Zhang and H. Su, “Cream-mac: Cognitive radio-enabled multichannel mac protocol over dynamic spectrum access networks,” IEEE Journal of Selected Topics in Signal Processing, vol. 5, no. 1, pp. 110– 123, 2011. [17] Y.-C. Liang, Y. Zeng, E. C. Y. Peh, and A. T. Hoang, “Sensingthroughput tradeoff for cognitive radio networks,” IEEE Transactions on Wireless Communications, vol. 7, no. 4, pp. 1326–1337, 2008. [18] J. Ansari, T. Ang, and P. Mahonen, “Spectrum agile medium access control protocol for wireless sensor networks,” in Proc. 7th Annual IEEE Communications Society Conf. Sensor Mesh and Ad Hoc Communications and Networks (SECON), 2010, pp. 1–9.

542