Cluster based cooperative sensing_-A survey

Cluster Based Cooperative Sensing:-A Survey Tanuja Satish Dhope (Shendkar)*, Dina Simunic** *Faculty of Electrical Engineering and Computing University of Zagreb, Croatia ** IEEE Senior Member, Faculty of Electrical Engineering and Computing University of Zagreb,Croatia *tanuja__dhope@yahoo.com, **Dina.Simunic@ fer.hr Abstract- The complexity of wireless system requires a careful design, especially related to bandwidth and energy efficiency. The energy efficiency is getting more and more on importance, due to increasing penetration of various wireless systems in different battery-oriented applications, as well as due to the more conscious global view on the need for "greening the Earth". Bandwidth efficiency is very important parameter, because it relates to frequency spectrum, which is naturally limited resource. The cognitive radio has been proposed as the future technology to meet the ever increasing demand of the radio spectrum by allocating the spectrum dynamically to allow unlicensed access on non-interfering basis.Cooperative spectrum sensing has been proposed to combat the multipath fading, shadowing and receiver uncertainty problem improving the detection performance by exploiting spatial diversity but at the cost of increase in cooperation overhead such as extra sensing time, delay, energy and operations devoted to cooperative sensing. Cluster based cooperative sensing can improve the performance and reduce the computational cost. In recent years, many methods of cooperative spectrum sensing have been proposed based on the clustering technique. In this paper, we provide an extensive review of cluster-based cooperative spectrum sensing. Keywords: Cooperative sensing, Cluster based sensing, Soft combining,Hard combining ,Soften Hard combining I. INTRODUCTION The existing static approach of spectrum management model is not efficient to cater the present and future requirement of spectrum for new wireless applications like mobile computers and personal digital assistants (PDA) etc. which demands for higher throughput and higher mobility [1][2].ITU predicted that 1720 MHz spectrum will be required by year 2020. The introduction of cognitive radios (CRs) [3] sheds new light on the unavailability of spectrum for new wireless service [4][5]. CR (unlicensed /secondary user) are based on the concept of dynamic spectrum access; an autonomous wireless devices able to optimize, learn and reason upon different network information. Spectrum sensing is crucial task in CR i.e. scanning the available spectrum bands and finding a suitable spectrum hole for their own communication purposes. Moreover, CRs must also be able to detect the arrival of other users/licensed users/primary users (PUs) in the band they are currently communicating in and perform spectrum mobility (change the channel) in order to minimize the possible interference in the network. Performance of any spectrum sensing algorithm is indicated by two metrics: by a probability of detection, 𝑃𝑑 , which is the probability of the algorithm correctly detecting the presence of the PU and by a probability of false alarm, 𝑃𝑓𝑎 , which defines probability of the algorithm mistakenly declaring the presence of PU leading to underutilization of available spectrum opportunities. The higher 𝑃𝑑 is, the less interference CR users may cause to primary user. Consequently, low 𝑃𝑓𝑎 and high 𝑃𝑑 means excellent sensing performance. However, 𝑃𝑑 and 𝑃𝑓𝑎 are positively correlated in practical. So improving the sensing performance means pursuing a higher 𝑃𝑑 when given the value of 𝑃𝑓𝑎 or try to achieve a lower 𝑃𝑓𝑎 versus a fixed 𝑃𝑑 [6] [7][8][9]. Cooperative spectrum sensing has been proposed [10] to combat the multipath fading, shadowing and receiver uncertainty problem, for improving the detection performance of single CR user by exploiting spatial diversity. Based on how cooperating CR users share the sensing data in the network, the cooperative sensing is divided into three categories [11]: Centralized /partial cooperative, Distributed /total cooperative and relay–assisted. In centralized cooperative sensing, a central identity called fusion center (FC), combines the received local sensing information, determines the presence of PUs, and diffuses the decision back to cooperating CR users. As shown in Fig. 1a, CR4 is the FC and CR0–CR3 are cooperating CR users, performs local sensing on the selected licensed channel through sensing channel. For data reporting, all CR users send the sensing results to FC through reporting channel. Note that partial cooperative sensing can occur in either centralized or distributed CR networks. In Distributed cooperative sensing, there is no FC for making the cooperative decision as shown in Fig. 1b. In this case, CR users communicate among themselves and converge to a unified decision on the presence or absence of PUs by iterations. In centralized CR networks, a CR base station (BS) is naturally the FC. Alternatively, in CR ad-hoc networks (CRAHNs) where a CR base station (CRBS) is not present, any CR user can act as a FC to coordinate cooperative sensing and combine the sensing information from the cooperating neighbors The sensing results can be improve by relaying local spectrum sensing result of one CR having not good channel state to other CRs having good channel state. In Fig. 1c, CR0 and CR3, weak report channel.CR1 and CR2, who have a strong report channel, can serve as relays to assist in forwarding the sensing results from CR0 and CR3 to the FC. For example, [12] uses relaying based on the Amplify-and- Forward (AF) cooperation protocol in order to reduce the detection time. Further the distributed cooperative sensing can 2012 International Conference on Communication, Information & Computing Technology (ICCICT), Oct. 19-20, Mumbai, India 978-1-4577-2078-9/12/$26.00©2011 IEEE 1 Figure 1.Cooperative Sensing a) Partial Cooperation b) Total Cooperation c) Total Cooperation via Relay –assistance [11] be classified based on various mathematical transformations [13] of the received data. In [13], multiple SUs are used to infer on the structure of the received signals using Random Matrix Theory (RMT). The SUs share information among them making the scheme not dependable on the knowledge of the noise statistics or its variance, but relying on the behaviour of the largest and the smallest eigenvalue of random matrices. But the conventional cooperative sensing described above has some drawbacks. Due to fading, the channels between CRBS and CR users are not reliable. As a result, reporting errors are inevitable. Also the global decision may not apply to all the CR users in the situation in which CR users are distributed in different spectral circumstance and also reporting to CRBS consumes large amount of energy and introduces control and transmission overhead. In this paper extensive survey of cluster based sensing is given. Section II describes the various combining Techniques in cooperative sensing based on the bandwidth requirement of control channel. Section III gives detailed idea about the cluster based sensing, classification of cluster based sensing viz a) overhead reduction b) performance gain and c) combined metrics based approaches which are followed by conclusions and future direction in section IV. II. COMBINING TECHNIQUES IN COOPERATIVE SENSING The sensing results reported to the FC or shared with neighbouring users can be combined in three different ways demanding upon the control channel bandwidth requirement:  Soft Combining: If there is no constraint on the bandwidth utilised by control channel then CR users can transmit the entire local sensing samples or the complete local test statistics for soft decision.  Hard Combining: If bandwidth utilised by the control channel is restricted and to avoid the communication overhead at the FC, CR users make a local decision and transmit the one bit decision for hard combining.  Softened hard combining: CR users can quantize the local sensing results and send only the quantized data for soft combining to alleviate control channel communication overhead. This type of combining outperforms the hard combination with respect to probability of detection and takes less overhead compared to soft combination. A. Soft combining: In the soft combination, sensing nodes send their sensed information directly to FC without making any decisions. The FC makes the decision based on this received information. The soft combination gives better performance than hard combination. This is true if and only if CRs are tightly synchronized in which case they can collectively overcome the SNRwall (in case of energy detector). The physical noise uncertainty gives a lower bound on signal strength that a user can reliably detect. This lower bound is increased further to keep the probability of false alarm tolerable. Existing receiver diversity techniques such as linear combining (LC), equal gain combining (EGC) and maximal ratio combining (MRC) can be utilized [14]. Due to the computational complexity of the LRT-based fusion methods that involves quadratic forms, an efficient linear combination of local test statistics is proposed in [14]. In [15], an optimal soft combination scheme based on NP criterion is proposed to combine the weighted local observations. The proposed scheme reduces to EGC at high SNR and reduces to MRC at low SNR. If all SUs have identical energy detectors and the received signals are modelled as correlated lognormal random variables, then a Linear-Quadratic (LQ) fusion strategy based on a deflection criterion that takes into account the correlation among the nodes proves to significantly outperform other fusion strategies under the mentioned assumptions [16]. [17] proves that the optimal fusion role at the FC is the half-voting rule if energy detection is used by the secondary users locally. B. Hard decision: In decision fusion, each user sends its one-bit to FC, to make the final decision based on following fusion rule. Specifically, 2012 International Conference on Communication, Information & Computing Technology (ICCICT), Oct. 19-20, Mumbai, India 978-1-4577-2078-9/12/$26.00©2011 IEEE 2 if each user only sends one-bit decision (“1” for signal present and “0” for signal absent) and no other information is available at the FC. Fusion rules [11] are : B.1.Logical –OR Rule: If one of the decisions is “1,” the final decision is “1.” Assuming that all decisions are independent, and then the probability of detection and probability of false alarm of the final decision are 𝐶𝑑 = 1 − (1 − 𝑃𝑑 ,𝑚 ) 𝑛 1 (1) 𝐶𝑓𝑎 = 1 − (1 − 𝑃𝑓𝑎 ,𝑚 ) 𝑛 1 (2) where n is number of CR users B.2. Logical –OR Rule: If and only if all decisions are “1,” the final decision is “1.” The probability of detection and probability of false alarm of the final decision are 𝐶𝑑 = (1 − 𝑃𝑑 ,𝑚 ) 𝑛 1 (3) 𝐶𝑓𝑎 = (1 − 𝑃𝑓𝑎 ,𝑚 ) 𝑛 1 (4) B.3. “k out of n” Rule: If and only if k decisions or more are “1”s, the final decision is “1.” This includes “Logical-OR (LO)” (k = 1), “Logical-AND (LA)” (k = n), and “Majority” (k= n/2) as special cases. The probability of detection and probability of false alarm for total number of „n‟ SUs are given as 𝑃𝑑 = 𝑛 𝑘 + 𝑚 𝑛−𝑘𝑚=0 1 − 𝑃𝑑 ,𝑚 𝑞 (1 − 𝑃𝑑 ,𝑚 ) 𝑘+𝑚 (5) 𝑤ℎ𝑒𝑟𝑒 𝑞 = 𝑛 − 𝑘 − 𝑚 𝑃𝑓𝑎 = 𝑛 𝑘 + 𝑚 𝑛−𝑘𝑚=0 1 − 𝑃𝑓𝑎 ,𝑚 𝑞 (1 − 𝑃𝑓𝑎 ,𝑚 ) 𝑘+𝑚 (6) Where 𝑃𝑑 ,𝑚 and 𝑃𝑓𝑎 ,𝑚 are probability of detection and probability of false alarm of the m th CR respectively which can be calculated for energy detection using following formulae with 𝑄 is 𝑄 function: 𝑃𝑑 = 𝑄 𝑁 ( 𝛽𝐸𝐷− 𝜎𝑠 2+ 𝜎𝜂 2 ) 𝜎𝑠 2+ 𝜎𝜂 2 (7) 𝑃𝑓𝑎 = 𝑄 𝑁 ( 𝛽𝐸𝐷− 𝜎𝜂 2 ) 𝜎𝜂 2 (8) C. Softened Hard Combining : Soft combining demands a wider bandwidth for the control channel [18], requires more overhead than the hard combination scheme. A softened 2-bit hard combining scheme is proposed in [15] for energy detector requires less overhead than soft combination and has greater performance gain than hard combination. The main idea behind softened 2-bit hard combining scheme is to divide the whole range of observed energy into more than two regions and to assign different weights to these regions. By doing this, nodes that observe higher energies in upper regions have greater weights than nodes that observe lower energies in lower regions. The thresholds are determined by using Neyman-Pearson criterion (to meet the target overall false alarm probability of all nodes in the network) and optimizing the detection performance. In [19] a new softened 3-bit hard combination scheme for collaborative spectrum sensing is proposed. In this case the whole range of observed energy is divided into eight regions. Each node sends to FC a 3-bit information that indicates the region in which its observed energy fell. The proposed scheme is superior to the traditional hard combination schemes in false alarm reduction. The detection performance of the 3-bit hard combination scheme can be improved with little additional cost by increasing the number of averaged PSDs. Sensor selection based on the knowledge of sensor positions is presented in [20]. In [21], iterative algorithms to estimate the probabilities of detection and false alarm of sensing nodes are proposed. However, these papers derive their results based on the assumption that if the local decision is the same as the final fused decision, it is the correct decision. However, it is important to emphasize that the proposed methods in [21] do not select the best CRs, but only assign weights to their decisions. Moreover, it is also noteworthy that the estimation of detection and false alarm probabilities would require a large number of samples. III. CLUSTER BASED COOPERATIVE SENSING In order to support data fusion through efficient network organization, CR users can be distributed into some small groups called „clusters‟. Each cluster has a coordinator called Cluster Head (CH) and the other CR users are cluster members. The spectrum sensing is performed on a hierarchical structure, through two levels of CR users cooperation. The low level one is conducted within the cluster and the high level one is executed among CHs. The benefits we can get by adopting cluster structure in cooperative spectrum sensing list as follows  Sensing performance improvement: we can improve the detection probability or decrease the error probability so that more reliable sensing results are acquired.  Sensing overhead reduction: Sensing overhead including energy consumption, time delay and bandwidth occupation can be reduced. Clustering technique is widely used in the research of ad- hoc network. It has been proved that clustering technique could reduce packet collisions and channel contention, leading to better network throughput under high load. Additionally, 2012 International Conference on Communication, Information & Computing Technology (ICCICT), Oct. 19-20, Mumbai, India 978-1-4577-2078-9/12/$26.00©2011 IEEE 3 energy efficiency could be achieved at the same time [11].The [11] mentioned different clustering methods for user selection depending on the availability of location information.  Random clustering: Used to randomly divide the SUs into clusters of equal size when the positions of both CR users and PUs are not available.  Reference-based clustering: clusters are formed based on SUs positions with respect to a given reference.  Statistical clustering: clusters are formed by using statistical information and the proximities of SUs when only the positions of SUs are known.  Distance-based clustering: only „k out of n’ SUs closer to the PU in a cluster participate in cooperative sensing when the positions of both SUs and PUs are known. A.System Model The generalised system model for cluster based sensing is shown in Fig.2.There is one primary transmitter (PUTx), one CRBS and number of CR users / SU, the location of SUs is randomly distributed. The SUs suffer different fading characteristics and have various reporting channel gain. Generally, the SUs in the same cluster are close enough to each other so that the channel between them can be regarded as perfect. Figure 2. Cluster Based Cooperative Sensing B. Classification of Cluster Based Sensing based on:  Overhead reduction based approach: The goal is to minimize the sensing overhead for e.g time cost, band occupation and energy consumption.  Performance gain based approach: This approach utilizes the user selection diversity or using appropriate fusion scheme for cluster based architecture.  Combined metrics based approach: Compromising between sensing overhead and sensing performance. B.1.Overhead reduction based approach: Most of the SUs are battery–operated mobile terminals and if conventional cooperative spectrum sensing is used wherein every SU reports its observations individually to the FC, affecting the lifetime of SU. Energy consumption, time costs and band occupation are sensing overhead in cooperative spectrum sensing scheme. In [22], to reduce the transmission data and more bandwidth requirement, the double threshold fusion scheme is proposed. Here there is a no decision region in the middle of the other two detection regions. If received energy falls into the no decision region, SUs don't give any decision. It is known that long-distance signal transmission consumes larger amount of energy than that of short-distance. A Hierarchical Spectrum Sharing Network is proposed in [23] based on the hybrid model in which multiple PUs are considered. In order to reduce the reporting distance, SUs close to FC report to it directly while SUs far way from FC form the cluster and then data is forwarded to FC in a multi- hop way. Here a „degree of correlation‟ a parameter is used to specify the spectral similarity of a cluster. For higher value of degree of correlation, less energy is consumed in transmitting data to CH. SUs with high spectral correlation are grouped into one cluster to perform cooperative sensing. In [24] to save the energy ultimately the battery power of SUs during reporting process, scheduled time slots is introduced in control channel. CHs remain quiet i.e in sleeping state until the time slots assigned to them for collecting and reporting decision come. But this leads to additional time delay. In [25], SUs have a same average SNR in the same cluster. Then, in each cluster, the SU which has the largest SNR is selected as CH. The SUs send their local observation to their respective CH. The CHs apply 2 bit softened hard combination as mentioned in section 2.3 to make cluster decision and the individual CH decision are sent to CRBS which makes global decision based on OR rule(conventional hard combining method). This approach gives better performance compared to conventional hard