1. Introduction
The existing communication systems over sea, such as automatic identification system (AIS) [1], only provide low data-rate non real-time services such as ship identification, positioning, and email, etc. Satellite system is able to cope with such service requirements, but the cost will be high. For the purpose of enhancing the service quality of maritime communications, radio transmission technologies (RTT) for ship ad-hoc network (SANET), based on the Recommendation ITU-R 1842-1 (VHF band) [2], are designed in our former study [3]. Moreover, in this paper, to solve the tradeoff between the congested spectrum band and the growing demands of high data-rate, real-time services (e.g., multi-media services and video surveillance), we employ the cognitive radio (CR) technique to support SANET opportunistically access other primary network’s spectrum band.
CR technique is considered as a potential solution to solve spectrum inefficiency problem. It is an intelligent wireless communication system that can change its transmitter parameters based on the interaction with its surrounding environment. By exploiting under-utilized or un-utilized spectrum band opportunistically, it provides a large amount of new spectrum availablities. There are two main entities in CR systems, namely primary user (PU) and secondary user (SU). PU has the license to operate in a certain spectrum band, while SU is permitted to transmitt or receive signal on this spectrum band only when PU is inactive to avoid interference to PU. Therefore, the basic idea of CR network is that the SU needs to vacate the band once the PU is detected [4]. Consequently, spectrum sensing, the process of detecting PU by sensing the surrounding radio environment, is the key technique to ensure SUs would not interfere to PUs. However, hidden primary user problem may decrease the spectrum sensing accurracy of a SU. Therefore, a cooperative spectrum sensing scheme is required to reliably solve the hidden primary user problem, providing a precise sensing result. In this paper, we implement the CR technique to the SANET, and propose a decentralized cooperative spectrum sensing algorithm for the SANET-CR system to avoid hidden primary user problem.
The rest of this paper is orgnized as follows. In section 2, we propose a SANET-CR system, and introduce the hidden primary user problem in the system. Section 3 reviews the local spectrum sensing methods under Rician fading environment and some traditional cooperative spectrum sensing schemes. Our proposed self-weighted decentralized cooperative spectrum sensing (SWDCSS) algorithm is presented in section 4. The simulation results and analyses are shown in section 5, and finally, we conclude this paper in section 6.
2. Proposed SANET-CR System
In our former work [5], we proposed three senarios for maritime wireless communications which is applied to the locations of harbor, shore, and ocean, respectively. Based on these senarios, in this paper we propose a SANET-CR system, where the CR technology is introduced into the SANET system, to make SANET users be able to opportunistically access the licensed spectrum band of other networks. In this way, the tradeoff between the congested spectrum band and the growing demands of high data-rate and real-time services can be managed. Fig. 1 depicts the architecture of the SANET-CR system, which describes SANET as the secondary network (SN). Around harbor or shore, we consider the terrestrial cellular network as primary network (PN). In the ocean enviroment, other maritime ad-hoc networks could be identified as primary networks. When SANET is closer to seashore, it accesses the unused spectrum band of the terrestrial cellular network for its own communications. Similarly, when SANET is in the ocean, it can use the licensed band of other maritime ad-hoc network opportuniscally. Notice that the SUs in SANET should not interfere the PU when they occupy the spectrum band of PN. Therefore, SUs should discover available spectrum holes over a wide frequency range for their transmission, and keep surveilling the spectrum band during the transmission. Once the presence of PUs is detected, SUs should evacuate the spectrum band so as to avoid inteference to PN.
Fig. 1.The architecture of SANET-CR system.
However, due to the hidden primary user problem caused mainly by the position of SUs in maritime environment, the SUs may not detect the existence of PUs, then it inevitably introduces the interference into PNs. As an example illustrated in Fig. 2, the left and right dot circles represent the operating ranges of the base-station (BS) in primary network and SU-A in secondary network, respectively. When PU-A is under the sleep mode, SU-A can occupy the spectrum band of PU-A to communicate with other SUs. Once BS sends the wakeup signal to PU-A, SU-A cannot detect the signal from BS to PU-A since SU-A is out of the operating ranges of BS. SU-A, however, continues to transmit signal over the spectrum band of PU-A, causing unwanted interference to PU-A. To solve such a hidden node problem, the cooperative spectrum sensing, where several SUs, e.g., including SU-B in Fig. 2, collaborate to detect the PU, is used to reduce the missing probability of detecting the PU. Due to the dynamic topology nature and no central entity for data fusion of SANET, in this paper, we propose a self-weighted decentralized cooperative spectrum sensing (SWDCSS) algorithm based on the SU notifications to reliably solve the hidden primary user problem for SANET.
Fig. 2.Hidden primary user problem in SANET-CR system.
3. Spectrum Sensing under Flat Fading Channel
3.1 Selection of the Local Spectrum Sensing Methods
Before performing cooperative spectrum sensing, every SU should detect whether an active PU is existing in its local sensing area. There are three widely used spectrum sensing methods: matched filter detection, energy detection, and cyclostationary feature detection [6,7].
Under the consideration of that the noise in maritime environment is relatively steady and the low computational complexity of the energy detection method is required, we consider the energy detection as our local spectrum sensing method.
3.2 Performance of Energy Detection under Flat Fading Channel
The classic energy detection method is based on the following hypothesis model [4]:
where y(t) is the signal received by the SU, x(t) is the transmitted signal of the PU, h represents the gain of the channel, and n(t) denotes a zero-mean additive Gaussian noise (AWGN). H0 states that there is no PU signal in the detected spectrum band, while H1 indicates that PU signal exists.
Define the detection probability (Pd) and false alarm probability (Pf) as:
where Y is the signal power received by the SU, λ is the threshold to determine whether the PU signal exists or not. Then the missing probability (Pm) can be expressed as:
The expressions of both Pd and Pf under AWGN channel can be obtained as in [8]:
where Qu(.,.) is the generalized Marcum Q-function, Γ(.) & Γ(.,.) are the complete and incomplete gamma functions, respectively. u denotes the time bandwidth product and γ represents the SNR.
Considering that Pf is independent on SNR since it is defined for the case of no signal transmission, the expression of Pf under any fading channel remains the same. In this paper, we focus on Pd and Pm only. Equation (5) gives Pd under non-fading environment based on the instantaneous received SNR, where h is deterministic. When h varies due to fading, corresponding Pd must be calculated by
where fγ(x) is the probability distribution function (pdf) of SNR under fading [9]. According to [10], the measured SANET flat fading channel follows the Rician distribution with the K-factor of 15 dB. When the signal experiences Rician fading, the pdf of γ will be
where K indicates the Rician K-factor, denotes the average SNR, and Iv(.) is the v-th-order modified Bessel function of the first kind. Then the corresponding Pd in (9) in the case of the Rician channel is obtained by averaging (5) over (8) as in [8]
3.3 Traditional Cooperative Spectrum Sensing Schemes
As illustrated in section 2, a cooperative spectrum sensing scheme is required to solve the hidden primary user problem in SANET-CR system. The main idea of the cooperative sensing techniques is that each SU can individually sense the channel and exchange their local sensing results to decide if the medium is available [11]. As the example in Fig. 2, SU-A is far from BS, resulting missing detection of signal from BS to PU-A. However, SU-B is much closer to BS which has high opportunity to detect the signal of the primary network. By the support of SU-B, SU-A can vacate the spectrum band of PU-A so as to avoid interference.
Various kinds of cooperative spectrum sensing algorithms have been proposed in [12]-[17]. Generally, the cooperative spectrum sensing algorithms can be classified into two categories [11]. The first one is the centralized cooperative spectrum sensing (CCSS), in which a central entity gathers information from all the SUs to make a decision on the medium status. Then the central entity transmits its decision back to the receivers. The second one is the decentralized cooperative spectrum sensing (DCSS), in which the SUs share their sensing information and make their own decisions. For the CCSS, such as in [12], SUs send their local sensing results to a data fusion unit, which combines and processes all the sensing results. Then the data fusion unit transmits its decision back to SUs. However, CCSS is not suitable for SANET, since that SANET has a dynamic topology and usually there is no central entity available for data fusion in maritime communication systems. In [13], a DCSS based on SU notification protocol, namely embedded spectrally agile radio protocol for evacuation (ESCAPE), is proposed to disseminate the evacuation information among the local SUs. Even though some SUs cannot detect the presence of PU due to the hidden primary user problem, they can vacate the spectrum band efficiently when they receive the notification signal from other SUs. However, the ESCAPE is vulnerable to the inaccuracy of sensing results.
4. Proposed Decentralized Cooperative Spectrum Sensing Algorithm
4.1 Proposed SWDCSS Algorithm
In this section, we propose a self-weighted decentralized cooperative spectrum sensing (SWDCSS) algorithm to effectively reduce the missing probability. The proposed algorithm is a kind of the DCSS algorithms, which matches with the dynamic SANET environment (without the unit of central data fusion). In addition, a weighting factor Wr(n) is introduced into our algorithm to further reduce the missing probability of the SANET-CR system, where n denotes the index of the sensing process and r indicates the r-th SU. The weighting factor is generated based on the historical sensing data, and the details of the proposed SWDCSS algorithm is illustrated in Fig. 3. There are total 7 phases in the proposed SWDCSS algorithm as follows.
Fig. 3.Flow chart of the proposed SWDCSS algorithm.
4.2 Analysis on the Proposed SWDCSS Algorithm
In (10), the weighting factor Wr(n) is defined as the ratio of the detection probability of the last sensing process and the average value of all the historical detection probabilities. A bigger Wr(n) means that the SU has higher probability of detecting PU. The received SNR is affected by the node location and the fading environment [18], and generally the correctness of SU’s sensing results is increased with a higher received SNR. Therefore, in SANET-CR system, the location of SU is the most important factor that influences the correctness of the sensing results. The historical sensing data can reflect the change of a SU position. By this consideration, we use historical sensing data to calculate the weighting factor to weight every SU’s local sensing result.
In our algorithm, if any SU detects the signal of the PU, it then notifies other SUs to vacate the PU’s spectrum band. Assuming there are R SUs in the SANET, the detection probability Qd and the missing probability Qm of the SANET-CR system can be expressed as:
Consider that there are tens of SUs in local area, and they perform local spectrum sensing independently. In this case, the collisions may occur if the multiple notification beacons are transmitted at the same time. The slotted ALOHA algorithm is used in our algorithm to cope with the collisions [19]. Assume that there are R SUs, s beacons in one packet, and l SUs attempt to send notification beacon at the same instance, which becomes occupancy number of one beacon. The expected value of the number of beacons with occupancy number l is given as:
When the occupancy number l = 1, the notification beacon is successfully broadcasted without collision. Thus we have
As a result, the missing probability in (13) becomes
To the best knowledge of the authors, most of decentralized cooperative spectrum sensing algorithms are not proposed based on the historical sensing accuracy [15,16]. The adoption of weighting the local sensing result makes the Qm in (13) and (16) decreased further compared to the ESCAPE algorithm. Moreover, applying slotted ALOHA algorithm makes Qm in (16) closer to the practical detection performance than that in (13).
Table 1.Simulation parameters
Fig. 4.Simulation scenario of the SANET-CR system. PUs by filled-in circles and SUs by hollow circles.
5. Simulation and Discussion
5.1 Simulation Scenario and Parameters
As shown in Fig. 4, we consider a scenario of 120 km by 120 km area with 4 hops in the area i.e., 30 km per hop. X-axes represents the distance from the seashore to the ocean, and Y-axes represents the distance along the seashore. The filled-in circles represent the ships which are communicating via PN, and the hollow circles represent the ships which are communicating as SUs. The first 30 km along the X-axes is the near-shore area where the density of ships is the highest. As the distance increases in X-axes, the density of ships decreases. This assumption is based on the fact that the number of ships are more in the near-shore area, while less in the ocean. In this paper, we consider all the ships (PUs and SUs) are deployed with the same distributation.
Fig. 5.Complementary ROC curves of the proposed SWDCSS with different SNR under Rician channel with K = 15 dB.
Within the region of the first hop (first 30 km), the PN becomes terrestrial cellular network, whose carrier frequency is 2.075 GHz in this paper. In the next 3 hops (far away from seashore), the PN networks can be other types of networks such as AIS and satellite communication systems, and the carrier frequency and the licensed frequency band can be totally different from those in the first hop. Since that there are more ships located in the first hop and so the probabiltity of introducing interference by the hidden primary user is higher, we take the network scenario in the first hop area (first 30 km) as an example to evaluate the performance of the proposed SWDCSS algorithm. In the first hop area, assume that the number of ships (i.e., SUs) adopted in the simulation is 30. The SANET system bandwidth is set to 50 KHz according to Recommendation ITU-R 1842-1 [2], and the Rician K factor is set to 15 dB as described in section 3. In the simulation, we predefine a set of false alarm probability Pf as 0 < Pf < 1, to generate the complementary receiver operating characteristic (ROC) curves, and then observe the simulation performances. The threshold λ is set according to [8]:
where μ is the degree of freedom, and
Fig. 6.Complementary ROC curves of the proposed SWDCSS with SNR = 5dB under Rician channel with different K factor.
To make the Monte-Carlo simulation results more reliable, we simulate 10,000 spectrum sensing trials, and then average out the simulation results to obtain the final performance curves.
5.2 Simulation Results and Discussion
Fig. 5 and Fig. 6 shows the performance of the SWDCSS algorithm on different interference and propagation environment. From Fig. 5, it is easy to observe that the missing probability Pm decreases as the received SNR increases. It tells that the proposed SWDCSS algorithm performs well in a relative low interference environment. As a result, we set the SNR= 5dB in all simulation scenarios. And we can observe from Fig. 6 that the missing probability Pm reduces as the K increases, i.e., Pm,(K=−40) > Pm,(K=9) > Pm,(K=15). It indicates that the SWDCSS algorithm performs better with a lager K vaule, which shows the asymptotic behavior as AWGN channel.
Fig. 7 compares the complementary receiver operating characteristic (ROC) curves of our proposed SWDCSS algorithm with the traditional ESCAPE algorithm in [13]. The solid lines represent the theory curves, while the other lines represent the simulation results. From Fig. 7, we can observe that the simulation results match the theory curves well. When there is a single SU, the system shows the worst performance due to the lack of information exchange among SUs. It is obvious that both cooperative algorithms, the proposed SWDCSS algorithm and ESCAPE algorithm, show better performance than the single SU case without cooperation. Furthermore, our proposed SWDCSS algorithm outperforms traditional ESCAPE algorithm since that the weighting factor has been adopted in our algorithm to further reduce the missing probability. As shown in Fig. 7, at the target false alarm probability Pf of 0.3184, the proposed SWDCSS algorithm shows the missing probability Pm of 0.0097, while the traditional ESCAPE algorithm shows the missing probability Pm of 0.0185. That is, the proposed algorithm decreases the missing probability by 47.5% compared with the ESCAPE algorithm.
Fig. 7.Complementary ROC curves of the proposed SWDCSS vs. traditional ESCAPE under Rician channel with K = 15 dB.
Fig. 8.Missing probability of the proposed SWDCSS vs. traditional ESCAPE under Rician channel with K = 15 dB.
Fig. 8 shows the comparison of missing probability of the proposed SWDCSS algorithm and the traditional ESCAPE algorithm. We expect to reduce the missing probability as low as possible, while keeping the false alarm probability in a relatively low level. For the simulations, we set the false alarm probability Pf equal to 0.1, and try to observe the tendency of missing probability by varying SNR. In Fig. 8, we see that the performance of the proposed SWDCSS algorithm is almost overlapped with that of traditional ESCAPE algorithm up to the SNR of -5 dB, since that the historical sensing data is not reliable in low SNR range. However, after the SNR of -5 dB, the missing probability of our proposed SWDCSS algorithm decreases more rapidly than that of the traditional ESCAPE algorithm, which suggests that the correctness of the sensing results is improved by introducing a proper weighting factor. As shown in Fig. 8, when the average received SNR is 5 dB, the traditional ESCAPE algorithm shows the missing probability Pm of 0.2530, while our proposed SWDCSS algorithm reaches Pm of 0.1833. That is, the proposed SWDCSS algorithm decreases the missing probability by 27% than the traditional ESCAPE algorithm.
6. Conclusions
In this paper, we propose a SANET-CR system to meet the requirement of growing demands of high data-rate and real-time services in maritime communications. By opportunistically utilizing the spectrum hole of primary network, the spectrum band of SANET can be efficiently expanded. However, SANET may cause unwanted interference to primary network if the active PUs of primary network are not detected due to the hidden primary user problem. To provide an applicable and reliable solution, we propose the self-weighted decentralized spectrum sensing scheme (SWDCSS), which does not need data fusion and prior knowledge of network’s topology. In the proposed SWDCSS, a weighting factor has been introduced to improve the correctness of the sensing results. Referring the analytical and simulation complementary receiver operating characteristic (ROC) curves, we observe that by weighting the SUs’ local sensing result according to their historical sensing data, the missing probability is significantly decreased and the detection probability is increased compared with the traditional ESCAPE algorithm, i.e., the interference from SANET-CR system to primary network can be dramatically decreased. Furthermore, the false alarm probability is also significantly decreased, which means that the SANET-CR system can use the spectrum hole a lot more reliably.
참고문헌
- Technical Clarifications of Recommendation ITU-R M.1371-1, "Technical characteristics for a universal shipborne automatic identification system using time division multiple access in the VHF maritime mobile band," Draft Edition 1.5, 2001.
- Technical Clarifications of Recommendation ITU-R M.1842-1, "Characteristics of VHF radio systems and equipment for the exchange of data and electronic mail in the maritime mobile service RR Appendix 18 channels," June 2009.
- B. Hui, K.H. Joen, K.H. Chang, S.G. Kim, J.W. Park, and Y.K. Lim, "Design of radio transmission technologies for VHF band ship ad-hoc network," in Proc. of ICTC, Sept. 2011, pp. 626-629.
- I.F. Akylidz, W.Y. Lee, and K.R. Chowdhury, "CRANS: Cognitive radio ad hoc networks," Ad Hoc Networks, vol. 7, no. 5, pp. 810-836, July 2009. https://doi.org/10.1016/j.adhoc.2009.01.001
- Y.B. Kim, J.H. Kim, Y.P. Wang, K.H. Chang, J.W. Park, and Y.K. Lim, "Application scenarios of nautical ad-hoc network for maritime communications," in Proc. of IEEE Oceans, Oct. 2009, pp. 1-4.
- D.D. Ariananda, M.K. Lakshmanan, and H. Nikookar, "A survey on spectrum sensing techniques for cognitive radio," in Proc. of CogART, May 2009, pp. 74-79.
- H. Sun, A. Nallanathan, C.X. Wang, and Y. Chen, "Wideband spectrum sensing for cognitive radio networks: a survey," IEEE Wireless Communications, vol. 20, issue 2, pp. 74-81, April 2013.
- F. Digham, M. Alouini, and M. Simon, "On the energy detection of unknown signals over fading channels," in Proc. of IEEE ICC, vol. 5, May 2003, pp. 3575-3579.
- A. Ghasemi and E.S. Sousa, "Collaborative spectrum sensing for opportunistic access in fading environments," in Proc. of DySPAN, Nov. 2005, pp. 131-136.
- K.H. Joen, B. Hui, K.H. Chang, S.G. Kim, S.M. Kim, and Y.K. Lim, "Performance analysis of channel compensation and channel coding techniques based on measured maritime wireless channel in VHF-band ship ad-hoc network," KICS Journal, vol. 36, no. 5, pp. 517-529, May 2011. https://doi.org/10.7840/KICS.2011.36B.5.517
- Y. Zhang, J. Zheng, and H.H. Chen, Cognitive Radio Networks: Architectures, Protocols, and Standards. CRC Press, Boca Raton, 2010.
- Z. Quan, S. Cui, and A.H. Sayed, "Optimal linear cooperation for spectrum sensing in cognitive radio networks," IEEE Journal of Selected Topics in Signal Processing, vol. 2, pp. 28-40, Feb. 2008. https://doi.org/10.1109/JSTSP.2007.914882
- X. Liu and Z. Ding, "ESCAPE: a channel evacuation protocol for spectrum-agile networks," in Proc. of IEEE DySPAN, April 2007, pp. 292-302.
- J.X. Xue, L. Li, S.K. Chen, and L.H. Huang, "Exponential weighted collaborative spectrum sensing based on SNR," in Proc. of ICMT, Oct. 2010, pp. 1-4.
- X. Huang, N. Han, G. Zheng, S. Sohn, and J. Kim. "Weighted-collaborative spectrum sensing in cognitive radio," in Proc. of CHINACOM, Aug. 2007, pp. 110-114.
- W. Zhang, Z. Wang, Y. Guo, H. Liu, Y. Chen, J. Mitola III, "Distributed Cooperative Spectrum Sensing based on Weighted Average Consensus," in proc. of IEEE Globecom, Dec. 2011, pp. 1-6.
- B.L. To, M. Otmani, T.M.T. Nguyen, and A. Fladenmuller, "Decentralized cooperative spectrum sensing for cognitive radio ad-hoc network: Hidden terminal awareness approach," in Proc. of ISSPA, July 2012, pp. 146-151.
- W. Ejaz, N. Hasan, H.S. Kim, and M.A. Azam, "Fully distributed cooperative spectrum sensing for cognitive radio ad hoc network," in Proc. of FIT, Dec. 2011, pp. 9-13.
- P.W. Graaf and J.S. Lehnert, "Performance comparison of a slotted ALOHA DA/SSMA network and a multichannel narrow-band slotted ALOHA network," IEEE Transactions on Communications, vol. 46, issue 4, pp. 544-552, April 1998. https://doi.org/10.1109/26.664310