Performance Evaluation of Cooperative Sensing via IEEE Radio

Transcription

1 Performance Evaluation of Cooperative Sensing via IEEE Radio Tahir Akram, Horst Hellbrück Lübeck University of Applied Sciences, Germany, Department of Electrical Engineering and Computer Science, Abstract Spectrum Sensing is one of the important tasks for the wireless devices but due to fading, shadowing and noise the performance of individual spectrum sensing devices is not ideal. Cooperative Sensing is seen as a way to improve the performance of individual spectrum sensing devices resultantly improving the efficient utilization of radio bandwidth and minimizing the interference among wireless devices. State of the art are extensive simulations and analysis on cooperative sensing although there are also number of performance evaluations of various fusion rules of cooperative sensing using software defined radios and FPGAs. The limitation of previous work is that they do not address the question how we can improve the overall performance of real systems with cooperative sensing. To the best of our knowledge, this is the first experimental work which presents cooperative sensing protocols with standard radios and evaluates the system performance using cooperative sensing. With IEEE equipped radio devices we model primary, secondary and cooperating users. We implement cooperative sensing protocols, setup a scenario, perform measurements and compare system performance with and without cooperative sensing. All the experiments are automated with the wisebed testbed software. The evaluation results of cooperative sensing protocols indicate new challenges for optimization and provide awareness to the problem of improving the overall system performance. Index Terms Cooperative Sensing; Testbed; IEEE Radio, Heterogeneous Radio Environment I. INTRODUCTION Wireless communication imposes more challenges than wired communication. One of the growing problems is the increasing number of collisions for concurrent transmissions due to increase in number of devices with wireless interfaces and still limited number of frequency bands. We use the term goodput which is the percentage of the existing gross data rate for the wireless channel that is effectively used for transmission of bits that are received successfully without errors. For wireless transmission collisions waste the existing resource as goodput decreases with each collision. Even without collisions or other interference physical layer goodput is reduced by overhead. For a setup with a single transmitter and a single receiver this comprises at least the physical layer preamble for bit and frame synchronization at the beginning of a frame. Additional overhead includes switching from listening or idle mode to transmission mode, CRC, optional carrier sensing and timeouts at the end of the frame. For wireless transmission a goodput between % and 90% at physical layer is realistic. In a heterogeneous radio environment like the ISM band collisions further degrade the performance. One of the performance risks is the hidden node problem where carrier sensing of a transmitter fails when two potential transmitters cannot sense each other but produce collisions of signal at the receivers. In order to solve this problem the receiver or in our approach other cooperating nodes can detect the channel state and avoid collisions by informing the potential transmitter of the busy channel. Vice versa the cooperating nodes inform the potential transmitter of idle channel and the opportunity to transmit without collisions. This approach is called cooperative sensing and is a candidate for improving performance on the physical layer goodput. In this work, we consider receiver as passive element therefore we do not use it for cooperation. To the best of our knowledge we are the first to provide an implementation for cooperative spectrum sensing for IEEE sensor hardware and demonstrate the effectiveness of this approach in a real measurement setup. Our contributions comprise: We present a novel implementation of cooperative sensing MAC protocols using IEEE equipped radio devices for a single hop case. We provide a testbed scenario and measurement results for cooperative sensing. We evaluate the system performance with and without cooperative sensing and demonstrate the effectiveness of spectrum sensing on real sensor hardware. We provide problem awareness of improving the system performance by cooperative sensing. The rest of the paper is structured as follows. The next section discusses related work. Section III describes briefly the testbed and the radio device we used in this work. Section IV describes details of the selected cooperative sensing protocols. Section V introduces details of the evaluation scenario. Section VI defines the evaluation criterion. In Section VII measurement results are discussed. Finally, the paper concludes with a short summary and future research direction in Section VIII. II. RELATED WORK An extensive list of work in simulations and analysis for cooperative sensing can be found in the survey paper of Akyildiz et al. [1]. In that paper, the authors state that

2 cooperative sensing improves performance of sensing devices but we also need to address synchronization, control channel and energy consumption in the evaluations which is still an open research question. There are also a limited number of results for implementation of cooperative sensing either using Software Defined Radio (SDR) or FPGAs. In the following, we discuss the state of the art. In [2] Yusof et al. present a MAC protocol for cooperative sensing implemented with Software Defined Radio (SDR). That work assumes constant primary user traffic during the active frame slots and evaluates the detection probability with various fusion rules. The authors ignore the aspect of energy consumption and the impact of synchronization and control channel in the evaluation. In [3], the authors perform evaluation of fusion rules of cooperative sensing with energy detector using FPGAs whereas in [4], the authors perform evaluation of cooperative sensing with Roy s Largest Root Test algorithm. In both articles, the authors ignore synchronization and control channel in the implementations. Summarizing the previous experimental works, their focus is on evaluation of fusion rules with various radio platforms (FPGA, SDRs). But no work evaluates the system performance of cooperative sensing with the standard radios. In our previous work, [5] we performed analytical work and evaluated the performance of cooperative sensing schemes in heterogeneous radio environments by ignoring synchronization, control channel and energy consumption aspects. In [6], we presented and validated our IEEE sensor node based testbed by measurements. In this work, we perform evaluation of selected cooperative sensing protocols with the testbed. For evaluation of system performance, we also incorporate energy consumption and impact of synchronization and control channel which were not addressed in previous work. We perform evaluation of system performance for hidden node scenario. Although there are already solutions for hidden node problem for the systems having same radio technology (RTS/CTS in )[7] here we address the system having heterogeneous radio environments (Bluetooth, IEEE and IEEE ). In the next section, we briefly describe the testbed and the radio device for implementation and evaluation of cooperative sensing in this article. III. TESTBED AND THE RADIO DEVICE Figure 1 illustrates the testbed. Each radio device is connected to a PC called Gateway which is further connected to Portal Server via wireless or wired LAN. The Testbed User connects to the Portal Server via the Internet. The Portal Server forwards data or program of the Testbed User to the respective Gateway which then forwards it to the radio device. The Portal Server also receives the results of the experiment from the Gateway and forwards to the Testbed User. Further details of the Testbed are given in [6]. In the next paragraph we describe the hardware for the radio device and its interface with the Gateway. Figure 2 provides the details of the radio device and its interface with the Gateway. The radio device is based on Radio Device AT86RF231 RFgChip Gateway LAN Portal Server Internet Fig. 1. Testbed Conceptual Overview [6] SPIg Interface ATXMEGA128 Microcontroller JTAGg Interface Fig. 2. Radio Device [6] Programmingg/gDatag AVRg Dragon RS232g Interface Testbed User Debugg/gDatag/guSensinggResults( fromg Gateway toggateway AT86RF231 RF chip and an ATxmega128A1 microcontroller. The RF chip is a 2.4 GHz low power transceiver. It is compliant to IEEE although only physical layer functionality is used in this work. The Received Signal Strength Indicator (RSSI) value is updated by the hardware every 2µs and stored in a register of the RF chip. We use this RSSI value to perform spectrum sensing for secondary and cooperating user. If the measured RSSI value is larger than a threshold, the channel is declared as busy otherwise it is declared free. Offset-QPSK (O-QPSK) is the modulation scheme and different standard and non-standard data rates (2kbps up to 2Mbps) can be selected. Different data rates for Primary User (PU) and Secondary User (SU) data transmissions distinguish between both transmissions at the receivers. Primary and secondary user transmits data at 2 kbps and 0 kbps respectively. The RF chip is controlled via SPI interface of the microcontroller. The radio device is connected to the gateway via two interfaces, from gateway via JTAG interface for programming and to the gateway via RS232. The serial interface serves for debugging the radio device, sending parameters for programs of the radio device and retrieving experiment results from the radio devices. Primary and secondary user transmit the same frame format [6] in different data rates. The header of physical layer frame of IEEE consists of preamble sequence, Start Frame Delimiter (SF D) and Frame Length (F L). Primary and secondary user receivers can distinguish the transmitted frames of primary and secondary user by data rate and F L. In the next section, we describe implementation details of the selected cooperative sensing protocols.

3 IV. COOPERATIVE SENSING PROTOCOLS For performance evaluation, we developed and implemented two protocols based on the ideas presented in [5]. In that work, we performed analytical evaluation ignoring the problem of control channel, synchronization and energy consumption. We evaluated two cooperative sensing schemes; Detection Performance Enhancement (DPE) and Sensing Frequency Enhancement (SFE). The idea of DPE scheme is to enhance the detection performance of secondary user with cooperating nodes. This is achieved by fusing the sensing results at secondary and cooperating users according to a rule (OR, AND and MAJORITY) [8]. The idea of SFE scheme is to introduce more sensing events performed by cooperating nodes between two sensing events of the secondary user. In this work, we implement both schemes in two protocols called DPE and SFE protocol based on schemes presented in [5]. For both protocols we define the following setup, model and terminology. PU TX radio device models primary user transmission. T ON represents the time during which the radio channel is busy due to PU TX transmission. Whereas T OF F represents the time where the radio channel is idle and therefore not utilized by PU TX. Be reminded that in our implementation primary user transmits deterministically whereas secondary user transmits opportunistically. In order to model secondary user transmission, we introduce a SU TX radio device. The cooperating nodes (CN 1, CN 2 ) perform spectrum sensing and report their results to SU TX. The cooperating nodes use a dedicated control channel [9] to report the result of their spectrum sensing to the secondary user. Primary and secondary users transmit data on the data channel. Furthermore, cooperating node and SU TX perform spectrum sensing also on the data channel. The radio devices implement a software clock based on the internal HW clock of the microcontroller. Due to crystal characteristics the clock drifts among the various radio devices. Therefore, we synchronize clocks among SU TX and the cooperating nodes by transmitting periodic beacon frames. We address the single hop scenario as mentioned in Section I. Beacon frames are sent by SU TX with time period T Beacon on the control channel. Cooperating nodes receive beacon frames and cooperating nodes adjust their local clock to SU TX. Figure 3(a) provides an overview of the first cooperative sensing protocol (DPE). In this protocol, SU TX uses a window of size (N T w ) to receive the sensing information frames sent by the cooperating nodes. Where N is the number of cooperating nodes present in the system and T w is the receiving window size for one cooperating node. SU TX performs spectrum sensing after time which is called Sensing Period [5]. Sensing Period, is the time interval between two local sensing events at SU TX. The cooperating nodes also perform spectrum sensing with time period of and report their result to SU TX as described earlier in this section. All the times in Figure 3(a) are local to the respective radio devices. After receiving sensing information frames from the cooperating nodes, SU TX fuses its local sensing information and those of cooperating nodes according to particular fusion rule (OR, AND or MAJORITY). SU TX transmits if the cumulative decision declares the data channel as free. In case SU TX decides to transmit, the transmission lasts until the next sensing event. The channel and state switching times of the RF chip are not shown in Figure 3(a). Besides all communication among SU TX and cooperating nodes are on the control channel and are performed at 2kbps. The beacon frame and sensing information frame duration is 1 ms in the current setup. Figure 3(b) provides an overview of the second cooperative sensing protocol (SFE). For SFE, SU TX synchronizes cooperating nodes via periodic Beacon Frames as for DPE. We use a dedicated control channel also in this protocol as used in DPE protocol. The time interval between two local sensing events at SU TX is defined as. The time interval between cooperative (at CN 1, CN 2 ) and local sensing event (at SU TX) is given by Besides, we use the term blocking time in later sections which is related to the time during which SU TX is not allowed to transmit on the data channel. This time is due to channel and state switching times of the RF chip, synchronization and control channel establishment. Additionally, we also implement a scenario with noncooperative sensing without cooperating nodes in the setup. SU TX performs spectrum sensing with time period, and transmits based upon the result of the sensing [6]. We name the setup with non-cooperative sensing as NO-COOP. In the next section, we describe details of the scenario we used to evaluate both cooperative and non-cooperative sensing protocols. Ts N+1. V. SCENARIO SETUP Figure 4 shows the details of the scenario we set up for evaluation in the testbed located on a floor of our university. The testbed consists of six mobile and six fixed nodes with equipped radio devices. More details of the testbed and the nodes are available in [6]. We use the mobile testbed nodes (PU TX, CN 1, SU TX, PU RX, SU RX) as well as fixed testbed nodes (CN 2, Start) in the scenario. PU RX is used to receive the primary user transmitted frames whereas SU RX is used to receive the secondary user transmitted frames. To distinguish between primary and secondary user transmissions we differ the field frame length F L of IEEE physical layer frame and the data rate [6] between the two transmissions as described earlier. The testbed does not ensure synchronized start of software on the radio devices, therefore we use a dedicated synch channel different than data and control channel to synchronize start of the experiments. A dedicated Start node is used to send the starting frame on the synch channel. The software running on all radio devices are synchronized to this starting frame. SU TX in our setup is in a room where radio conditions were not good therefore nodes relay the start frame sent by the Start node. The experiment duration is T exp which is managed by the local timers of the radio devices.

4 PU TX T ON TOFF SU Transmission Control Channel Data Channel Sensing Event Receiving Window PU TX T ON T OFF T Beacon SU TX T Beacon N T w Collision SU TX Collision 1 T w Collision CN 1 Beacon Frame CN 1 T 1 S N 1 Beacon Frame CN 2 Sensing Information Frames CN 2 T 2 S N 1 Sensing Information Frames (a) (b) Fig. 3. DPE and SFE Cooperative Sensing Protocols In the following, we describe the methodology of placement of various testbed nodes for the scenario. SU TX is placed at a location where it cannot detect the transmission from PU TX, that is; detection probability [5] P d 0 (hidden node problem). The cooperating nodes are placed at the locations where they can detect primary user transmission whereas both cooperative nodes can also communicate successfully in both directions with SU TX. SU RX is placed at a location where it can receive secondary user transmission but simultaneous transmission from primary user corrupts the reception of secondary user transmission. Similarly PU RX is placed at a location where it can receive primary user transmission but simultaneous secondary user transmission corrupts the reception of primary user transmission. The placement of testbed nodes according to the requirements specified here was not trivial as the radio channel conditions in the indoor environment was not very predictable. As reproducibility of the experiments was most important to allow the comparison between different implementations and settings we measured in the evenings to ensure a minimum number of movements (people) on the floor and to avoid additional unpredictalbe interferers (WLAN and Bluetooth). Besides, setting up a scenario for 0% < P d < 100% [6] did not ensure reproducibility of the experiments. That is why, we placed SU TX with P d 0% and cooperating nodes with P d 100%. In the next section, we describe details of the evaluation criterion for comparison of various cooperative and non-cooperative sensing. VI. EVALUATION CRITERION We define and calculate Goodput (bps) and T hroughput (bps) in bits per second for primary and secondary users in Equation 1 and Equation 2, Goodput (bps) = N received 8 T exp (1) 16m PU TX CN1 CN2 SU RX Start 27m PU RX SU TX Fig. 4. Scenario for Cooperative Sensing T hroughput (bps) = Nsent 8 T exp (2) where N received and N sent are the number of bytes which are received and transmitted by the primary or secondary user respectively. We use CRC to detect the integrity of the received frames. Therefore, preamble and CRC are not part of goodput or throughput as they are not useful part of data for the user. We use Equation 3 and Equation 4 to calculate the goodput in ratios (P U Goodput r and SU Goodput r ) where P U Goodput max and SU Goodput max are the reference maximum possible primary and secondary user goodput respectively. P U Goodput r = SU Goodput r = P U Goodput (bps) P U Goodput max (bps) SU Goodput (bps) SU Goodput max (bps) In heterogeneous radio environments or ISM band as mentioned in [5], each radio device has equal right to access the radio bandwidth. Therefore, both primary and secondary user goodput are equally improved in the system. In the following (3) (4)

5 discussion, we evaluate the overall system performance for the heterogeneous radio environment. Therefore, we define T otal Goodput as follows T otal Goodput = P U Goodput r + SU Goodput r (5) In state of the art [10], energy consumption is considered an important aspect in wireless networks, therefore we incorporate this aspect also in the evaluation of a cooperative sensing protocol. Due to collisions between primary and secondary user transmission, energy is wasted both at PU TX and SU TX which is calculated in our experimental setup as follows: W P U = W SU = P U T hroughput(bps) P U Goodput(bps) P U Goodput max (bps) SU T hroughput(bps) SU Goodput(bps) SU Goodput max (bps) Where the ratios (W P U and W SU ) represent the energy wasted due to collisions at PU TX and SU TX respectively. Besides, energy is also wasted due to transmission of Beacons and Sensing Information Frames at SU TX and cooperating nodes which is calculated as follows: (6) (7) W COOP = TCOOP T exp (8) Where the ratio, W COOP represent the energy wasted due to beacons and sensing information frames at SU TX and cooperating nodes. T COOP is time duration of all beacons and sensing information transmitted on the control channel during the experiment duration, T exp. Besides, note that W P U and W SU attributes to the energy wasted on the data channel whereas W COOP attributes to the energy wasted on the control channel. We calculate the total energy wastage, W as follows: W = W P U + W SU + W COOP (9) We define the system performance in terms of Effectiveness which is calculated as follows: Effectiveness = T otal Goodput T otal Goodput + W (10) In the next section, first we describe the parameter settings of the experiments and then we discuss the measurement results we obtained after running the experiments for three nights. VII. RESULTS For the measurements results presented in this section, the experiment duration T exp for a single run is set to 0 s. The expected value of clock drift among various radio devices is 0.5 s per 100 s or 0.5 %. Therefore, we select the values T Beacon = 100 ms and T w = 2.4 ms which ensures more than 98 % successful reception of sensing information frames at SU TX in our setup. We investigated the system for T ON and T ON = T OF F in this work, we will investigate other cases in future. We select T ON = ms and T OF F = ms, therefore the system was analyzed for values of = 10 ms, ms, ms and ms. Note that the maximum time duration of IEEE physical layer frame is 4.2 ms (2kbps data rate) for primary user and 2.2 ms for secondary user (0kbps data rate). Multiple frames are used by primary and secondary user in various experimental settings. We selected IEEE channel number 25, 22 and 21 for data, control and synch channel respectively in the measurements. Figure 5 shows primary and secondary user goodput as calculated in Equation 3 and Equation 4 for both non-cooperative (NO-COOP) and cooperative sensing (DPE and SFE). All results are given with confidence level of 95%. With non-cooperative sensing, primary user goodput is 0% (see Figure 5 (a)). In this case, all primary user transmissions are corrupted due to concurrent secondary user transmissions as secondary user is not able to detect primary user transmissions (hidden node problem). Note, that the maximum achievable primary user goodput in this set up is % as the duty cycle for primary user transmission is also %. Secondary user goodput is maximum for noncooperative sensing (see Figure 5 (b)) as there is no blocking time which is there in cooperative sensing. Cooperative sensing protocols improve the primary user goodput but the secondary user goodput is also decreased. In the following we discuss the performance of each cooperative sensing protocol. The performance of DPE protocol varies according to the fusion rule. There is a large increase (0% to %) in primary user goodput due to the OR fusion rule. Due to high cumulative detection probability, secondary user is able to detect the primary user transmissions leading to very low number of collisions. There is medium level decrease (from 45% to 25%) in the secondary user goodput due to blocking time and high cumulative detection probability for this fusion rule. Figure 5 shows a small increase (0% to 5%) in the primary user goodput with AND fusion rule, as this fusion rule does not help to improve the cumulative detection probability which leads to higher number of collisions as is the case with non-cooperative sensing. Furthermore, a very small decrease in the secondary user goodput (from 45% to 38%) results due to blocking time. The performance of the MAJORITY fusion rule is close to the OR fusion rule. With SFE protocol we see medium level increase (0% to %) in primary user goodput but medium level decrease (45% to %) in secondary user goodput. Although the cooperative sensing events help to improve performance but there is always transmission after local sensing event at SU TX (due to hidden node problem) which collides with primary user transmission (see 1 in Figure 3(b)). Figure 6 shows the Effectiveness as calculated in Equation 10 for the system having different number of cooperating nodes (N = 1,2). We see that DPE-OR and DPE-MAJORITY outperforms all others because the collisions are minimum for these protocols which leads to higher T otal Goodput and lower energy wastage (W ). The performance of SFE protocol is medium as the number of collisions is less than NO- COOP but higher than DPE-OR and DPE-MAJORITY. The performance of DPE-AND is even worse than non-cooperative sensing. In this case the reduction in number of collisions is minimal as compared with non-cooperative sensing. Besides

6 PU Goodput r 10 N = (a) DPE OR DPE AND DPE MAJORITY SFE NO COOP SU Goodput r N = 1 10 Fig. 5. Primary and Secondary User Goodput with Cooperative and Non- Cooperative Sensing Effectiveness N = 1 10 (a) Effectiveness N = 2 (b) 10 (b) DPE OR DPE AND DPE MAJORITY SFE NO COOP Fig. 6. Effectiveness with Cooperative and Non-Cooperative Sensing blocking time reduces T otal Goodput whereas beacons and sensing information frames increase energy waste compared to non-cooperative sensing. Table I lists the components of energy wastage (W) and T otal Goodput for non-cooperative (NO-COOP) and cooperative sensing (DPE-OR). For non-cooperative sensing, although T otal Goodput is already low (45%) but high energy wastage (W P U = %, W SU = 54%) further reduces Ef f ectiveness (31%). For DPE-OR, T otal Goodput is high (66%) and besides small energy wastage (W COOP = 11%, W P U = 10% and W SU = 7%) further increases Effectiveness (70%). When we compare the system performance with an increasing number of cooperating nodes (From N = 1 to N = 2), we do not see any significant change. This is because additional cooperating node (in N = 2) does not help significantly in reduction of collisions in our setup (see Figure 4 and Section V). For = 10 ms and N = 2, the blocking time increases for SFE protocol that no time is left for transmission for secondary user after the cooperative sensing events, therefore we do not TABLE I Effectiveness, =10MS, N=1 DPE-OR Parameter Value Effectiveness 70% T otal Goodput 66% W P U 10% W SU 7% W COOP 11% NO-COOP Parameter Value Effectiveness 31% T otal Goodput 45% W P U % W SU 54% W COOP 0% evaluate the system with these values. With this discussion, we show that for systematic improvement of performance of real systems with cooperative sensing, we need a complete understanding of the problem. Besides considering fusion rules we also need to consider the impact of synchronization, control channel and energy consumption in the evaluations. More work is needed here to find optimal solutions. VIII. CONCLUSION AND FUTURE WORK In this work, we present a novel approach for evaluation and implementation of cooperative sensing with a real sensor network based testbed. We provide awareness to the problem of improving system performance with cooperative sensing. Initial results presented in this work show that the improvement in overall system performance with cooperative sensing is encouraging. In future, we will analyse the system performance with probabilistic primary user traffic pattern, furthermore we will also perform optimizations based upon the prediction of primary user traffic. As the results presented in this paper show that an increase in number of cooperative nodes does not help improve system performance therefore we orient our research direction and will consider using cooperative sensing in the mobile scenario. ACKNOWLEDGMENT This work has been supported by the Federal Ministry of Education and Research of Germany: Förderkennz. 17N3809, SoFT and German Academic Exchange Service (DAAD) and Higher Education Commission (HEC) of Pakistan. REFERENCES [1] I. Akyildiz, B. F. Lo, and R. Balakrishnan, Cooperative Spectrum Sensing in Cognitive Radio Networks: A Survey, Physical Communication (Elsevier) Journal, vol. 4, no. 1, pp. 62, Mar. 11. [2] S. S. Yusof, K. K. Rashid, N. Latiff, N. Fisal, M. Sarijari, and R.A.Rashid, TDMA based Cooperative Sensing using SDR Platform for Cognitive Radio, in Proceedings of the 18th Asia Pacific Communications Conference, APCC, 12. [3] S. U. S. Srinu, S.L. Sabat, FPGA implementation of Cooperative Spectrum Sensing for Cognitive Radio Networks, in International Workshop on Cognitive Wireless Systems (UKIWCWS) 10, 10. [4] R. S. Yoshimura, F. S. Mathilde, J. P. Dantas, V. A. de S. Jr.., J. H. da Cruz Jr., J. J. Bazzo, and D. C. Melgarejo, A USRP based Scheme for Cooperative Sensing Networks, in Anais do 4 Workshop de Redes de Acesso em Banda Larga - WRA 14, Brazil, pp [5] T. Akram, T. Esemann, and H. Hellbrück, Performance Evaluation Metric for Cooperative Sensing in Heterogeneous Radio Environments, in Proceedings of the European Wireless Conference, Guildford, UK, 13. [6] T. Akram, T. Esemann, T. Teubler, and H. Hellbrueck, A Reusable and Extendable Testbed for Implementation and Evaluation of Cooperative Sensing, in Performance Monitoring, Measurement and Evaluation of Heterogeneous Wireless and Wired Networks Workshop, PM2HW2N 13, 6-10 November 13, Barcelona. ACM, 13, pp [7] N. Jinal and P. Kaur, An Efficient Method to Avoid Hidden Node Problem in Ad-Hoc Network, vol. 80, pp , 14. [8] A. Ghasemi and E. S. Sousa, Opportunistic Spectrum Access in Fading Channels Through Collaborative Sensing, JCM, vol. 2, no. 2, pp , 07. [9] B. F. Lo, A Survey of Common Control Channel Design in Cognitive Radio Networks, Phys. Commun., vol. 4, no. 1, pp , Mar. 11. [10] D. Feng, C. Jiang, G. Lim, L. J. Cimini, and G. Feng, A Survey of Energy Efficient Wireless Communications, IEEE Communications Surveys, vol. 15, pp , 13.