Abstract
Recent improvements in communication technology have undergone a significant shift over the last two decades, with state-of-the-art communication equipment, standards, and protocols simplifying the lives of consumers everywhere. For more than a decade, advancements in communication technology have mostly focused on increasing the speed with which information can be delivered and retrieved from anywhere in the globe at any time of day or night, regardless of location. Four-generation (4G) communication technologies, which have already been developed and implemented, are used to offer users with seamless access to multimedia content at transmission rates of 100 megabits per second (Mbps). It is becoming more vital to create new technologies in order to meet the growing need for faster speed as well as a variety of other advanced features. 5G networks have just recently been built as a result of extensive research and development. This has resulted in the gradual replacement of existing 4G services with new 5G networks, which are capable of transmitting multimedia content such as audio-video and high definition images, among other things, at data transmission rates in the gigabyte range or higher (up to several gigabits per second). Further recent development, in addition to the Internet of Things (IoT), which was made possible by future communication technology, is the Internet of Things-based social network. Aspects of this include the ability to connect and expanding Internet connectivity to all physical devices that consumers use to access common commercial and industrial services available on the Internet. In spite of this, with the advancement of existing high-speed communication networks, the effective interaction of devices with their inputs and responses via the Internet may be made possible through 5G Internet of Things networks. This new generation of automation and communication systems has emerged as innovative platforms for the next generation of automation and communication systems to be developed further in the future. M2M data may be utilised to more efficiently distribute resources if machine learning (ML) and optimum cell clustering are applied to the situation. It is because of this heterogeneity that the ML is able to make the best use of the remaining resources of the M2M network in order to optimise efficiency. Over the last several years, the shortage of radio frequency spectrum has proven to be the most challenging hindrance to wireless communication. This has occurred from the large number of high-frequency devices that need significant amounts of bandwidth allowance. Cognitive radio networks have been designed to meet this higher demand as a result of this increased demand.
1. Introduction
The exponential rise of wireless communications has led in an ever-increasing need for the deployment of wireless services in both licensed and unlicensed frequency spectrum, resulting in an ever-increasing demand for licensed and unlicensed frequency spectrum [1]. Several studies have indicated that the practice of fixed spectrum assignment results in inefficient spectrum usage, which supports this claim. Cognitive radio (CR) is a new technology that has arisen to handle this problem in a cost-effective and efficient manner [2]. Increased spectral efficiency may be achieved by the use of CR by gaining access to the intermittent intervals in empty frequency bands, which are known as white space or spectrum holes [1, 2]. In the most basic sense, the essential role of each secondary user (SU) in CR networks is to establish whether or not the licensed users, also known as primary users (PUs), are present and, if the PUs are missing, to ascertain the empty spectrum [3]. In order to do this, SUs engage in spectrum sensing, which is defined as the process of perceiving the radio frequency environment [4]. In order for SUs to be effective, they must not produce damaging interference to PUs, which is the fundamental purpose of spectrum sensing. SUs must do this by moving to a band that is accessible or by minimizing the interference to PUs to a level that is acceptable to them [3, 4]. The second goal of spectrum sensing is for SUs to locate and exploit the spectrum gaps as effectively as possible in order to fulfil the throughput and quality-of-service (QoS) needs of the network. So the detection performance in spectrum sensing is critical to the performance of primary and CR networks, as well as the overall performance of the system [5]. This fact emphasizes the need of a spectrum sensing algorithm in a cognitive radio network, which is described in detail below [6]. When working in noisy surroundings, multipath fading, and shadowing effects are present, the detection performance of classic spectrum sensing techniques is often degraded [7].
When it comes to wireless communication systems, cognitive radio has earned a prominent position among the extensively explored paradigms. This is due to the fact that it has the potential to improve spectrum efficiency while simultaneously increasing network capacity [8]. It is possible to boost the total network capacity by sharing spectrum across several wireless communication systems without causing interference [7, 8].
CR allows users to access a variety of air interfaces and pick the most appropriate options depending on the operating circumstances and communication requirements [9]. The use of CR approaches has shown to be advantageous in a number of wireless communication systems [10]. In mobile and satellite communication networks, CR methods are becoming more prevalent [11] going into great depth on the history of the approaches, the reasons for their acceptance, and the specifics of how they work [10, 11]. It offers an overview of how the CR technology is being used in several additional wireless communication systems, including public safety and emergency networks, the Internet of Things, military communications, and aeronautical communications, among other applications [11]. Spectrum sensing is a fundamental building component of cognitive radio technologies [12]. Identifying spectrum utilization and establishing whether or not main users are present in a specific geographical region is the process of spectrum identification and determination. While spectrum sensing is still primarily concerned with measuring radio frequency energy over the spectrum, it has evolved into a more general term that includes obtaining characteristics of spectrum usage across dimensions such as space, time, frequency, and code in the context of cognitive radio (CR) [13]. Spectrum sensing has taken on new dimensions as a result of the development of opportunistic spectrum access and cognitive radio ideas [8, 12, 13], and it is unquestionably one of the most difficult problems to solve in cognitive radio systems. The most significant difficulties connected with spectrum sensing include noisy uncertainty circumstances, the availability of adequate hardware platforms, the detection of concealed primary users, the sensing length and frequency, and the detection of spread spectrum primary users. The key obstacles in cooperative sensing are the reduction of cooperative overheads and the development of a decision fusion technique [8, 12]. The energy detection (ED) method is straightforward to develop and does not need any prior information on the part of the main user; nonetheless, it is not ideal for usage in noisy or unpredictable environments. Compared to other algorithms, the sensing algorithm is more precise, but it increases computing complexity [8, 12, 14]. Due to the fading and shadowing effects, a single CR user may be unable to notice the presence of a licensed (primary) user at times. In such circumstances, cooperative spectrum sensing may aid in the improvement of the sensing process as well as the production of more trustworthy data. Because the conventional sensing method takes into account judgments from all SUs [15, 16], it increases cooperation overhead. The majority of methods in the literature are aimed towards assessing the existence or absence of a principal user signal in the band under consideration. However, in particular applications, such as military applications, it is necessary to determine the modulation type of the received signal in order to perform effective jamming/antijamming of the signal and interpretation of the enemy signal [8, 15, 17]. Using energy detection methods, we can estimate the energy of the received signal and compare it to a set threshold value that has been specified. A technique with such broad applicability lacks sufficient dependability and may result in incorrect detection when sensing a wideband spectrum. Wideband sensing technologies that are now in use are based on the Nyquist principle, which necessitates large sample rates. Communication systems may employ many RF frontends at the same time in order to perceive the wideband radio spectrum, but this can result in lengthy processing times, high hardware costs, and computational complexity [18, 19]. Because of the limits and the need for development in spectrum sensing algorithms, the primary objective for the proposed work in this study is to improve the detection performance of sensing algorithms and to offer better spectrum sensing algorithms that are more efficient. Measuring the use of the radio spectrum campaigns is being carried out all around the globe to estimate the amount of spectrum white space available, and the findings reveal that there is plenty of white space available in a variety of interest bands [15, 20–22]. If such information is accessible for every geographical place, it will be very beneficial for new and forthcoming wireless apps to have access to it. The study work provided here has been expanded to include a quantification of the radio frequency white spaces accessible in the Pune area. A literature analysis revealed that there is a need for improvement in the performance of spectrum sensing algorithms, and the issue description and research goals for this study are established as follows. The problem is defined as follows: improving the performance of spectrum sensing algorithms in cognitive radio networks in order to increase the chance of detecting and identifying the principal user [23]. An investigation and overview of the many parameters that are employed in the performance analysis of spectrum sensing methods are the primary objectives.
Because of poor use of the restricted spectrum, which accommodates users who do not have spectrum licenses, dynamic spectrum access approaches are required. The CR users have been granted permission to temporarily utilise the unused licensed spectrum. The spectrum management capabilities of CR allow SUs to choose the most appropriate frequency from a variety of available bands [24]. The quality of service (QoS) fluctuates in response to the time-varying channel characteristics. Spectrum sensing, spectrum decision-making, spectrum sharing, and spectrum mobility are the four issues that must be overcome in order to achieve efficient spectrum management. A number of related difficulties were addressed, including the common control channel problem, the power management method, spectrum sensing and scheduling approach, the hidden and exposed terminal problem, and mobility [25]. It provides a high-level overview of the technological and regulatory problems associated with opportunistic channel access (OSA). It is possible to coexist successfully with licensed and unlicensed users in the same frequency range according to the OSA standard.
Several optimization strategies, including convex optimization, have been employed in the literature to optimise the capacity of SUs. An optimum time-sharing approach and power allocation strategy for CR broadcast channels were given by the authors of existing study [26]. The optimum power optimization algorithms presented by existing study are designed to work within the restrictions of average transmit and interference power, respectively. (i)Performance assessment of an algorithm for spectrum sensing based on an energy detector(ii)A comparative examination of currently available spectrum sensing methods(iii)A two-step strategy was used in the development of an improved spectrum sensing algorithm(iv)Using a collaborative approach, the development of an improved spectrum sensing algorithm for the usage of white space is being pursued
2. Overall System Model and Methodology
Spectroscopic sensing is the method of getting information about the consumption of radio spectrum and the presence of main users in a given geographic region. Obtaining this information may be accomplished by the use of geolocation and databases, the use of beacons, or the use of local spectrum sensing at cognitive radios [3–5]. Despite the fact that spectrum sensing is traditionally used for measuring the spectrum usage, or energy of the radio frequency over the spectrum, when cognitive radio is considered, it is a more common term that refers to finding the spectrum usage information across multiple dimensions such as time (including space), frequency (including frequency), and code (including code). Identifying the kinds of signals present in the spectrum, as well as their modulation, waveform, bandwidth, and carrier frequency among other characteristics, is also required. However, in order to do this, stronger signal analysis methods with increased computing complexity are required [27].
A dual-path FSO system, with each path having two hops, is proposed for long-distance communication, with each path having two hops. At destination , the optical signals are merged with the help of the SSC diversity system. In the DSSC system, one of the two pathways is chosen based on its optimality. If the current SNR falls below a certain level, the other route is turned on. It is expected that the distributions for both branches are not similar since the relays on both channels operate in automatic failover mode (AF). For the hops of the S-RA-D connection, we assume that the Malaga (' M) distribution is used, and for the hops of the S-RB-D link, the GG distribution is used. The Malaga distribution simulates weak-to-strong turbulence, while the GG distribution has been generally recognized for FSO channel modelling for a long period of time. There are two nonregenerative nodes on pathways A and B, respectively, represented by the relays RA and RB. They first amplify the signal before passing it on to the next stage of transmission. Each FSO route is set with an IM/DD modulation scheme and an OOK modulation scheme, respectively. It is expected that full CSI is accessible at both the relays (RA and RB) and the destination node () [28].
In signal detection, detection probability refers to the likelihood of a particular signal being detected by the detector. Suppose a detector announces the presence of PU even when the PU is not really present. In such case, the probability of the detector declaring the presence of PU is 1. Second, the possibility that a detector would indicate the presence of PU when, in reality, PU is not present is referred to as the false alarm probability (FAP). Missed detection has a three-fold chance of occurring: when it comes to probability, it is defined as the possibility that the detector will declare PU to be absent when PU is really present [29]. In computing, it is the amount of time it takes for a secondary user to detect the existence of a primary user channel and to determine whether or not the channel is open. SUs experience a drop in throughput if the sensing time is too long. If the sensing time is too long, the data transmission period is lowered, leading in a reduction in throughput (sensor units). Roughly speaking, SNR is the ratio of signal intensity to noise. Because of the strength of the sent PU signal and the propagation environment, the received PU signal at the CR receiver has a signal-to-noise ratio (SNR) of 1. Both error probabilities are connected together using the sensing duration, signal-to-noise ratio, and detection threshold values, among other factors [30]. It has been shown that increasing the signal-to-noise ratio (SNR) may improve detection performance. When it comes to communication channels, it is defined as the number of successful messages that are sent over them at any given moment. “Throughput” refers to the quantity of data that can be processed by the algorithms that work together to collect and analyse data. There are other factors to consider, including overhead for cooperation and sensing duration. Detecting the modulation type of a PU signal is accomplished in the following ways: a desirable attribute in a PU signal is the capacity to detect the modulation type of the signal because; in certain applications, it is necessary to determine the modulation type of the signal in order to adjust the receiver settings properly [31]. Complex sensing algorithms that are difficult to implement are favored over simpler and more implementable sensing algorithms that are similarly energy efficient in terms of complexity and implementation. Therefore, the signal to interference and noise ratio (SINR) for an M2M device is represented as follows:
In order to assign the resource, it is necessary to verify the kth M2M capacity, which is specified in the form of the transmission rate stated in equations (3) and (4).
Then, the energy efficiency of an M2M device, denoted by the symbol Ek in equation (5), is computed based on two factors: the power used by the M2M device and the device’s capacity, where it is defined as the ratio of the device’s capacity to the power spent. where
The architecture of network is given in Figure 1.
3. Proposed Method
represents the reconfiguration threshold of M2M at time , is the likelihood that a user is in the idle model, and is set to 0.1 in the first instance.
The adaptation of resource allocation during idle time is represented by the value .
In addition, the threshold level-based resource allocation enables the M2M network to be rebuilt heterogeneously in order to allow for better resource allocation, which contains the following sections:
Coordination between the network and MAC layers is performed via the use of ANFIS-assisted resource allocation, which calculates the most appropriate rule for allocating resources to an M2M device based on its configuration [32].
An alert message is then sent to the sleep mode cell in the mobility-based environment, in which the location information of the cell is needed right away. During a static cell’s lifetime, the frequency of resource allocation exchange rates is completely governed by the amount of time that cell has been occupied. It is engaged in response to network stability, and it is used, among other things, to alter the pace at which resources are allocated on an interval basis [33].
Achieving resource allocation across PU and SU clusters that are linked together using M2M architecture is the responsibility of the settling phase. The synchronization of the M2M stability controllers, which is a critical computation, is also computed by this programme. When combined with machine learning (ML), optimal cell clustering provides a technique for resource allocation that is based on the sensitive information provided by M2M. Because of the heterogeneity of the cells, the ML makes the best possible use of the remaining resources of the M2M network to maximize efficiency [34].
Maintenance is performed by scattered controllers, who gather information on how heterogeneous the M2M network is throughout this period. In the next phase, the settling phase determines the average amount of energy present in the network’s accessible resources by using the following equation [35]:
3.1. Forwarding Phase
When the controller is installed, the PU configures the communication settings for the controller. The ANFIS controller monitors the hello packet sharing as well as the actual data delivery in iterative fashion.
The threshold value governs how PU generates reports based on the information it receives. During this phase, the required data is collected, analyzed, and provided to the appropriate PUs as soon as possible [36]. The feature of data transmission is maintained in multipath communication between clusters, according to the findings of the research. It is possible to broadcast directly to the BS-SU in order to avoid transmission delays at certain periods. Finally, the MAC layer ensures that the controller is sent without delay as much as possible.
4. Performance Evaluation
According to the findings of the literature research, the detection performance of spectrum sensing algorithms is often affected by the noise uncertainty issue, poor signal-to-noise ratio (SNR), multipath fading, and shadowing in wireless channels [12]. Due to the inefficiency of single user detection sensing algorithms in dealing with these challenges, it is necessary to investigate cooperative sensing and wideband sensing methods [22]. Recognizing the modulation type of a received signal may be just as crucial in certain situations as determining whether or not a signal is present in a specific frequency range [37]. In addition to testing in a simulation environment, the implementation and performance analysis of sensing algorithms on a hardware platform are crucial [38].
Accordingly, the next chapter provides an in-depth look at how considerable research was undertaken to meet the stated objectives. A simulation and testing phase began with a few spectral sensing techniques like energy detection and matching filter detection [39]. We wanted to gain a more profound knowledge of how sensing algorithms work in general and the many factors that influence their performance [40]. In order to address the shortcomings of the sensing algorithms, a number of upgrades are offered, and the performance of these improved spectrum sensing algorithms is evaluated in detail [41].
As previously described in the preceding chapter, two novel improved sensing algorithms have been suggested in this study. Following extensive testing, it has become clear that the suggested methods provide improved detection performance. The investigation is also being expanded in order to determine the availability of white space in the radio frequency spectrum [42].
The experimental work is divided into four major sections, each of which is described below. (1)A study of the performance of the spectrum sensing algorithms(2)Analysis of the enhanced spectrum sensing algorithms’ performance(3)Evaluation of the proposed spectrum sensing algorithms in terms of performance(4)Determination of the amount of RF white space in the Pune area
The results of the above-mentioned studies and quantification are presented in further detail in the following sections. Communication toolbox was used to simulate sensing algorithms in the MATLAB programme (version 14.1), which was used to simulate the algorithms [43]. The parameters for the simulation are listed below.
Data to be entered: BPSK/QPSK are two random signal modulation schemes
Model of the channel: AWGN (additive white Gaussian noise) receiver SNR ranges from -20 dB to 10 dB
As previously noted, energy detection is one of the most straightforward spectrum sensing techniques since it does not need any prior information of the main user [44]. The energy of the received signal is computed, and the result is compared to a specified threshold in this method of signal processing. If the received energy signal is larger than the threshold, it is determined that a main user is in the vicinity of the receiver. This portion of experimentation is subdivided into four subsections listed below [45]. (1)Energy detection (ED) with a predetermined threshold(2)The use of a dynamic threshold for energy detection(3)MFD is an acronym for matched filter-based detection (MFD)(4)Evaluation of the ED and MFD algorithms in comparison
Specifically, the primary goal of this investigation is to investigate the relationship between signal-to-noise ratio (SNR), sample quantity, and noise uncertainty on the receiver operating characteristics (ROC) of a cognitive radio receiver [46].
The procedure is as follows: (1)For , -12, and -15 dB and a given number of samples , the probability of detection (Pd) is depicted as a function of the probability of false alarm (Pf)(2)After that, the ROC curve is presented for a constant SNR value of -10 dB with ranging between 500 and 1500 points(3)Finally, to investigate the influence of noise uncertainty, the receiver operating characteristic (ROC) curves are displayed for , , and a noise uncertainty parameter, increasing from 1 to 1.05. Where =1 means that there is no noise uncertainty and =1.05 implies that there is the greatest noise uncertainty, we get the following
Conclusions: the impact of increasing SNR values on the detection effectiveness is shown in Figure 1. It can be observed in the picture that when the signal-to-noise ratio (SNR) grows, the probability of detection, Pd, increases as well for a constant value of the probability of false alarm, Pf, indicating that the detection performance improves as the SNR increases. When Pf is 0.1, a 5 dB increase in the SNR value resulted in a 50% increase in the power density [47].
The USRP IQ sampling rate is set to 1 million samples per second, and the active antenna of the USRP is RX1. The USRP strength is set to 25 decibels.
The suggested EWSS algorithm with gradient and double threshold is implemented on a CR receiver based on the National Instruments USRP.
The algorithm is programmed with the appropriate frequency range for scanning, which is then executed. In this experiment, the GSM band between 890 and 915 MHz is scanned and recorded.
The scanned bandwidth is separated into a number of subbands, each of equal size, after scanning the whole range of frequencies (in this case, 890-915 MHz). It is then computed and shown what is known as the gradient of the spectrum in each subband as a function of the total number of subbands.
It is determined by how changes in gradient transcend the higher and lower thresholds in a subband whether or not a main user is present in that subband [48].
The CWSS (conventional wideband spectrum sensing) algorithm is used to concurrently scan the required band in order to do a comparison between the two algorithms.
Constrained optimization is the process of maximising or minimising a particular objective/cost function while taking into account a variety of limitations. If the objective function and the constraints are both linear in nature, the issue is transformed into linear programming, and it becomes quite simple to identify the global optimum. However, when the issue is convex, alternative approaches are used in order to discover the global optimum solution.
The gradient technique and Newton’s method are the two most often used approaches for solving an unconstrained convex problem. After each iteration of the gradient method, the initial viable point travels in the direction of the gradient towards the optimum value, as shown in the diagram. In comparison to the gradient approach, Newton’s method is quicker in terms of convergence; nonetheless, it involves the calculation of Hessian of the objective function as part of the convergence process.
The interior point approach, the ellipsoid method, and the projected gradient algorithm are all used to solve restricted convex problems. When the search point’s border is reached, a penalty is applied to the objective function, which is known as the inner point method. The ellipsoid technique generates a series of ellipses in the feasible set, which is then evaluated. At each iteration, the volume of the ellipse shrinks, resulting in the largest enclosure of the convex function possible.
Whenever any of the constraints or the objective function are nonlinear, the optimization problem is classified as nonlinear programming (NLP). The interior point technique, genetic algorithm, and simulated annealing are the three most extensively used methods for global optimization of NLP in practice today. The maximum and lowest values of a multivariable function are determined by the Lagrange multiplier. The Lagrange function has been improved.
In the GSM band, it is very difficult to verify that a channel reported by spectrum sensing was indeed occupied by a main user in the first place. This is due to the fact that channels are dynamically assigned to various users depending on their needs. As a result, for the sake of comparison, the proposed EWSS algorithm and the CWSS algorithm are applied to the FM band where the existence of the main user is already known for a specific geographical region. For various FM channels, the likelihood of detection for both methods is evaluated and contrasted.
ANFIS, FIS, and ANFIS with different M2M couples provide different energy consumption results, as shown in Figures 2(a) and 2(b). When comparing ANN with FIS, it is easy to observe that ANN consumes much more energy. The following may be discussed in further detail: in cases including a variety of M2M pairings, the FIS prevents data from being transferred along the resource allocation boundaries, so limiting the radius of allocation and hence conserving energy and resources [49].
(a)
(b)
In Figures 3, due to the fact that these two roads are on different sides of the resource allocation gaps, it helps to minimize unintended road crashes [40]. However, it increases the energy consumption of each node on the paths in question.
Figures 4 represent using different M2M pairings, the results of the packet delivery ratio comparison between the current.
In Figure 5, according to the data, ANN and FIS have very different performance profiles, which may be attributed to two factors. First and foremost, we provide two node-disjoint anchor lists to direct data supply, therefore significantly shortening the route to resource allocation. For the second time, we are navigating two distinct but equally spaced pathways on each side of the M2M resource assignment holes.
Figure 6 shows the validation plots of the proposed work.
5. Conclusion
The approach based on energy detection is confronted with the issue of noise uncertainty, which may be mitigated by the application of a dynamic threshold. An alternative method is to use a matched filter-based algorithm, which is only feasible in specific cases. The conditional two-step spectrum sensing method satisfies the requirements of CRNs for speed and accuracy in a fast and accurate manner. When compared to individual techniques, it provides a 25% improvement in the chance of detection and a 70% reduction in the amount of time it takes to detect. Utilizing a reduced number of secondary users in collaboration increases detection probability by 10% compared to the typical cooperative spectrum sensing algorithm. It has been shown that the throughput of the multislot parallel cooperative sensing method grows as the number of sensing slots increases, which aids in the speeding up of the sensing process.
The standard spectrum sensing algorithms are incapable of dealing with signals that utilise spread spectrum, time or frequency hopping codes, or a combination of the two. As a consequence, these sorts of signals provide a significant challenge in the detection of spectral information [12, 22, 50]. As a result, the sensing algorithms may be improved even more in order to discover main users of the spread spectrum. This article describes how to create a dedicated CR network that has numerous CR nodes in order to illustrate the opportunistic usage of spectrum white space using the algorithms suggested in the work given here. Such types of experiments will aid in the deployment of CRN on a broad scale in the near future, which will assist to address the issue of spectrum scarcity while also increasing the usage of available spectrum. Today, there are several low-cost software-defined radio (SDR) hardware platforms available on the market, all of which run open source software. The creation and testing of prototypes with such low-cost choices will allow for the deployment of CRN to occur more quickly.
Data Availability
The data that support the findings of this study are available on request from the corresponding author.
Conflicts of Interest
Authors declare that there are no conflicts of interest.
Acknowledgments
The authors would like to acknowledge the Deanship of Scientific Research, Taif University, for funding this work.