Cognitive Radio Network Technology for IoT-Enabled Devices ^†

Abstract

The exponential development of wireless applications has increased problems in the spectrum. The unlicensed frequency spectrum is becoming highly saturated, in order to support the conditions of new radio devices with increasing data rates. The spectrum that has previously been allocated is also unused. As a consequence of these developments, scientists have been putting a lot of effort into developing an approach to the issue of the limited spectrum that could make it possible to create a more effective utilization of it. As cognitive radio permits opportunistic use of the licensed spectrum in less crowded areas, it has been proposed as a solution to this challenge. This paper provides an overview of the cognitive radio environment, including a dynamic spectrum access strategy, as well as additional information on the cognitive capabilities operating in combination IoT communication technologies. We investigate the utilization of cognitive radio in the Internet of Things along with the significant role that cognitive radio plays in making the Internet of Things possible. Cognitive radio will provide a comprehensive examination of spectrum sensing, which will cover the many types of sensing, sensing that is based on machine learning, as well as open topics that still need to be addressed further in this sector. This research paper is written in a way that provides detailed instructions for the purpose of assisting new researchers in the area of Cognitive Radio Networks.

1. Introduction

Considering the predetermined frequency allocation of typical wireless networks, there is only a limited amount of the wireless spectrum that is appropriate for broadband communications. As a result of this scarcity, the concept of Cognitive Radio (CR) communication was developed. CR communication comprises a number of mechanisms that enable licensed and unlicensed user systems to coexist on the same spectrum. The utilization of intelligent transceivers has become possible via developments in Software-Defined Radio (SDR), and digital signal processors, which allow for the dynamic and adaptive utilization of the available spectrum. Licensed and unlicensed wireless devices may share the same spectrum with the CR method, which was inspired by the concept of software-defined radio [1]. Figure 1 shows that numerous companies are making significant efforts to realize CR technology because of the wide variety of potential uses it could have, from TV White Spaces (TVWSs) [2] to satellite communications [3,4,5]. According to the CR network, principal (licensed) users (PUs) have spectrum priority. In contrast, Secondary (unlicensed) Users (SUs), also known as cognitive users, make use of this spectrum in a way that does not disrupt the typical operation of Licensed Primary Users (PUs). SUs are also known as cognitive users. In order for the SU to be able to take advantage of the unused sector of the spectrum, it is necessary for it to have CR capabilities. These CR capabilities allow the SU to collect information about its working conditions and to make autonomous adjustments to its radio parameters. Numerous studies have sought to define the “CR” and its various cognitive cycles since it started [6,7]. Existing wireless systems can improve their spectral efficiency in two main ways:

• Implementing a strategy for opportunistic spectrum access, commonly known as DSA.
• Spectrum sharing, which refers to the process of allowing primary user (PU) and secondary user (SU) systems to share the spectrum that is currently available.

In addition, the most essential activities that are available in any of the hypothesized cognitive cycles are spectrum awareness, spectrum analysis, decision, and spectrum adaptation. These tasks need to be completed repeatedly until there is a full adaptation to the changed environmental state.

The structure of this paper is as follows: Section 2 illustrates the spectrum sensing strategy. Section 3 highlights the main practical imperfections that may arise in a practical CR system cognitive radio network based on IoT. Then, Section 4 presents the cognitive capabilities operating in combination with IoT communication technologies. Section 5 concludes the paper.

2. Spectrum Sensing Strategy

2.1. Spectrum Sensing Technique

Spectrum sensing’s fundamental purpose is to choose between two hypotheses:

• H₀: The status of the channel is idle, which means it is available for opportunistic usage secondary users (SUs) and means no licensed user signal is present (PU):

$${ H}_0:Y\left(n\right)=N\left(n\right);$$

(1)

• H₁: The status of channel is busy, meaning a licensed user (PU) signal is active, leading to preventing SUs from occupying the spectrum:

$$H_1: Y\left(n\right)=h X\left(n\right)+N\left(n\right).$$

(2)

The SU received signal is denoted by Y(n), the PU broadcast signal by X(n), and the AWGN is indicated by N(n), where AWGN has a mean of zero and a variance of ${\sigma }^2$. The channel gain is denoted by h, samples are indexed from 1 to N where n is an integer, and N is the total number of samples. In order to choose between the two hypotheses, the deciding threshold value is often compared to a test statistic of Y(n). H₁ is presumed to be true and PU’s signal is considered active if the test statistic is larger than the threshold. On the other hand, the test statistic is below the threshold, H₀ is considered valid, and the PU signal can be nonexistent. The overarching paradigm of spectrum detection is shown in Figure 2.

Receiver Operating Characteristic (ROC) curves are frequently utilized to evaluate sensing accuracy. These curves represent the relationship between the possibility of a false alarm and the possibility of a missed detection [8,9]. To describe these probabilities, we assume that P_d is the possibility that a report made by the unlicensed SU about the presence of a licensed PU signal is present.

$$P_d=Prob\left[H_1\right|H_1].$$

(3)

It is possible to protect primary receivers from interference if the probability of detection is highly sufficient. As a result, having a high P_d is very desirable. False alarm probability (P_fa) is the chance that the SU will report a PU signal when none actually exists.

$$P_{fa}=Prob\left[H_1|H_0\right].$$

(4)

False alarms reduce spectral efficiency because of the chances lost when the spectrum is not used. Moreover, this decline may have an unfavorable impact on QoS. As a result, P_fa should be kept low to prevent the wasted potential of the spectrum. Miss detection probability (P_m) is the chance that a PU signal will be missed even when it is actually present.

$$P_m=Prob\left[H_0\right|H_1].$$

(5)

In the event of miss detection, SUs may transmit in the same band as main users (PUs), which results in severe interference for PUs. Thus, P_m should be reduced so that license users are not interrupted.

2.2. Machine Learning with Spectrum Sensing

CR is an intelligent radio system with perception, learning, and reasoning [10]. Spectrum sensing lets the system perceive its radio surroundings. Classification and generalization algorithms help the system learn from data. The system uses knowledge and reasoning to achieve its goals [10]. Spectrum sensing that is based on machine learning (ML) is able to identify channels occupied in cognitive radio networks due to the intelligent architecture of the system [10,11]. This is accomplished by addressing categorization and assessment. Figure 3 provides an illustration of the supervised and unsupervised machine learning methods that have been developed. Energy statistics, probability vectors, and temporal occupancy are just a couple of the features employed in machine learning models for spectral sensing. The selected features may obviously influence the efficiency with which primary users are identified.

• Learning without supervision is an option that could be practically viable for cognitive radio networks that operate in unknown RF environments.
• In the case that cognitive radio already possesses previous data regarding the surroundings, it may make use of this information by employing the supervised learning technique.

3. Cognitive Radio Network Based on IoT

The amount of connected hardware has increased because of the IoT’s many different connectivity methods. This category includes a wide variety of electronic equipment, some examples of which are smartphones, smart home appliances, and other forms of smart equipment. These high-tech gadgets can communicate with one another, collect information, process it, and take appropriate action. This identifies the internet of things at the center of both present and future economics. In addition, the exponential growth of internet-of-things devices has raised competition for a limited spectrum. In terms of spectrum availability, cognitive radio looks like an achievable solution that might support both existing and emerging internet-of-things devices. This constitutes an extensive understanding of how cognitive radio and the internet of things can work together. The Internet of Things (IoT) is able to be combined with cognitive radio technology, and the resulting system is referred to as the Cognitive Radio Internet of Things (CRIoT). CRIoT will be used in many applications, including time-sensitive ones like technology and intelligent transportation. The combined technologies must satisfy many network parameters, including channel assignment delay, end-to-end latency, dependability, energy efficiency, and high throughput. Adapting to expanding connectivity requires several communication standards and technologies. CR applications for Bluetooth, ZigBee, and Wi-Fi interior smart environments are among the many IoT applications and services. IoT networks must accomplish various tasks. In the Primary User’s presence, interference-free channels will be collected, and providing any IoT operating channel changes with continuous PU monitoring. For many Internets of things gadgets, licensed spectrum access is regulated to reduce interference with core users [12].

3.1. Internet of Things Definition

Internet of Things (IoT) is a network of things connected to the Internet that can communicate via various communication protocols. Sensors and communication modules allow these objects to interact with the surroundings [13,14,15]. Pressure, proximity, humidity, and temperature sensors are among those sensors. Nowadays, the IoT connects the real and digital worlds. Things will interact and learn with IoT. The Internet of Things is going to affect every aspect of our lives, connecting our televisions, automobiles, smartphones, utility meters, heart monitors, thermostats, and virtually everything else we can think of. Several technologies and protocols, such as Wireless Sensor Networks, IPv6, IPv6 Low power Wireless Personal Area Networks (6LoWPAN), Constrained Application Protocol (CoAP), and Cognitive Radio, have contributed to the development of the IoT since its conception at the end of the year of 1998 [16]. As a result of this shift, there are now more app-development environments available online, and more compact, low-power devices, such as sensors and actuators, on the market. Limitations in size, cost, power, battery life, range, storage, processing, and data transfer are common among IoT gadgets. Connectivity in IoT networks is typically high, and the underlying protocols are simple [17]. The five layers of Figure 4 make up the IoT architecture [18,19,20]:

• The radio frequency identification (RFID) tags, sensors, actuators, etc., that make up the physical world are represented by the Perception (or Recognition) layer. Data collection and transformation is the primary function. Some of them, like actuators, take a control signal and turn it into a predetermined motion;
• The primary function of the transmission (or network) layer is to send or receive control signals between the middleware layer and the perception layer via various networking technologies;
• The middleware is a software layer that processes the information it receives from the lower layers and makes decisions depending on the results;
• The IoT applications can be found in the application layer. It then uses that information to tailor its service offering to the end user;
• As well as the information gathered at the transport layer, the business layer gives system administrators command over the entire IoT infrastructure. It creates numerous types of company plans.

It is anticipated that, by the year 2025, Internet of Things networks would include more than 50 billion connected heterogeneous items. Some examples of these devices include sensors, surveillance cameras, kitchen appliances, mobile phones, thermostats, utility meters, and nearly anything else. The requirement for reliable wireless communication will increase dramatically as a result of this expansion. A few examples of how the internet of things is already being integrated into day-to-day life include the smart grid, smart energy management, smart security, smart farming, smart transportation, smart housing, and smart cities [21,22]. A significant barrier is presented by the scarcity of the available spectrum. In this study, we focus on this issue, which is a direct result of the proliferation of IoT devices and the consequent convergence of the cognitive radio paradigm.

3.2. Cognitive Radio Network Technology for IoT

In recent years, researchers’ attention has largely been focused on the three fields, communication, opportunistic, and sensing. However, IoT will not be capable of evaluating its capacity to deal with potentials of developing difficulties without extensive cognitive capabilities [23]. It is currently clear that both intra- and inter-cognitive communication (INTRACC and INTERCC, respectively) are crucial for the Internet of things. In INTRACC mode, cognitive objects share the same set of abilities, whereas in INTERCC mode, they have a unique sets of skills. Cognitive radio-based IoT frameworks are being researched. Future IoT items should think, learn, and recognize both social as well as physical worlds [24,25]. These objects should feature observation action cycles, intelligent decision making and knowledge discovery, enormous data analytics, and on-demand service provisioning. Hence, IoT will require cognitive radio network integration. The following causes necessitate it:

• Wireless methods will become increasingly important in the future for transmitting data between sensors and cloud servers. According of their limited range, wireless solutions like Bluetooth and Zigbee have made way for the CRNs-based IoT paradigm [23,25];
• The situation will increase because of the massive number of IoT objects and it will be tough to provide bandwidth to such a large number of devices. Concurrently, the growing population of licensed users will provide challenges for those who lack proper authorization for this purpose. Bandwidth acquisition costs will also be expensive. This encourages us to look outside the box for solutions, and CRNs might be that resolution;
• Interference issues will arise when the number of IoT objects increases and they are moved around. IoT objects utilizing a CRN can opportunistically seek out channels with low interference to improve upon communication;
• Spectrum sharing is not possible with current wireless communication technologies. Spectrum sharing problems will be an issue for future cellular networks as well. This advantage of cognitive radio technology bodes well for the future, when more machines will compete for limited spectrum. However, the implementation of spectrum-sharing areas should take into account regulatory, business, and technological frameworks.

3.3. Reasons for Utilizing Cognitive Radio Network in the IoT

The deployment of IoT networks is difficult due to a number of challenges, including a limited communication range, resource scarcity, interference problems, and Reconfigurability. The application of cognitive radio has the potential to improve all of these challenges, as well as others, including heterogeneity, reconfigurability, and automaticity. The utilization of Cognitive Radio in the Internet of Things is probably driven by the requirement to overcome these issues. It is difficult to allot frequency bands to the ever-increasing number of Internet of Things devices because there is a limited amount of the spectrum available. By enabling more frequency reuse, CR provides a solution to this challenge. The vast majority of technologies that are a part of the internet of things, such as RFID and IEEE 802.15.4 (ZigBee), make use of the same oversubscribed ISM and UHF frequency bands. This results in interference. Therefore, it is to be anticipated that interference will occur when multiple devices use these frequencies.

CR solves this problem by allowing for interference-free spectrum access on demand. Because of the restrictions placed on them by the ISM’s unlicensed bands, wireless technologies can only communicate over short distances. Buying licensing to access a frequency spectrum that guarantees long-distance connectivity is expensive and unnecessary. With CR, you do not have to spend money on a license, and you may take advantage of empty frequencies to have long-distance conversations whenever you want. In order to store and analyze the data they produce, IoT items must establish connections with one another and send that data to a large number of servers (Cloud servers).

To fix this problem, CR is the best option. Many different approaches can be taken to solving heterogeneity issues in IoT applications. To handle this diversity, new forms of communication should be developed that allow for self-discovery, self-organization, and self-management of their respective environments [26]. Among these paradigms, CR is a viable option for dealing with heterogeneity problems. The flexibility to change configurations and operate autonomously is another factor. In fact, it is anticipated that smart things will be able to alter themselves automatically. This means that things should be able to scan their surroundings, identify nearby companions, and reorganize themselves accordingly. In this case, CR is shown to be an effective strategy. When everything is considered, it is clear that cognitive radio holds great potential as an IoT enabler.

3.4. IoT Applications and CR

As shown in Figure 5, the IoT will be implemented in every facet of society, from private residences to public infrastructures, from the academy to the hospital, from the factory to the local supermarket. The internet of things presents unprecedented possibilities; its full influence and potential will be seen when more and more things become online in the future. Applications in the medical field, the military, cognitive radio-vehicular ad hoc networks, emergency networks, smart grids, and smart metering are all possible uses for this technology. All of these are examples of probable applications for CR technology inside the Internet of Things; however, there are currently only a small number of studies in the literature that address these issues. The authors of [27,28,29,30,31,32,33] provide an overview of IoT applications and demonstrate how CR can be used to solve some of the problems inherent to these programs.

3.5. Problems and Challenges in Cognitive Radio Based on IoT

In order to fully benefit beyond what cognitive radio can provide IoT networks, some unsolved research issues have to be resolved. Many of these concerns are addressed in the next section.

• Spectrum efficiency in the context of a network built on cognitive-radio-based internet-of-things nodes requires the optimization of a number of resources, including energy efficiency, transmission power, latency, and data throughput. The authors of [34] formulated an optimization problem that considers transmission rate, transmission power, and transmission delay in order to identify the best possible solution given specific limitations. They used a polyhedral branch to obtain a solution. The results demonstrate that transmission latency, power rate, and interference all rise in tandem with network and packet size. There is much work to be carried out on the formulation and solution of multi-objective optimization problems that take into account a wide range of variables.
• Energy efficiency is a major issue that must be resolved in CR-based IoT networks. The power consumption problem is exacerbated for energy-constrained nodes and battery-powered devices when IoT items with cognitive radio capabilities perform additional functions, primarily spectrum sensing. Energy harvesting as well as Cooperative Wireless Networks are among the methods offered to address energy efficiency challenges. The term “energy harvesting” refers to the practice of collecting and storing energy from renewable resources like the sun and the wind for later consumption [35]. When it comes to Green Computing and the practical use of the internet of things, energy harvesting is often regarded as one of the most enabling and important technologies. Using energy harvesting, the authors of [36] present a differential game model to handle the problem of resource allocation in cognitive wireless sensor network (WSN).
• Security is of paramount importance in CR-Based IoT networks, but it is also a difficult undertaking as, because of the inherent heterogeneity of most IoT products, each one must conform to its own standardized security requirements. Not all heterogeneous networks can be successfully implemented using these guidelines. Authentication, security assurance, and intrusion software are just a few of the privacy and security considerations that should be made when developing IoT systems [23]. There have been commendable attempts to address security concerns in several cognitive-radio-enabled IoT applications, as illustrated in Figure 6.

4. Cognitive Capabilities Operating in Combination IoT Communication Technologies

It is predicted that additional forms of communication technologies will make it possible for IoT to incorporate cognitive elements. This study demonstrates why these reducing forms of communication are so crucial in the modern world.

• CR with Cloud Services: We predict that the “everything as a service” model, based on cloud-based resource sharing, will become the common practice in the world in the near future. With an internet connection, people will be able to use any service at any time and from any location. Unfortunately, gaining access to these tools is extremely difficult. Service-based infrastructure, service-based platforms, and service-based software are the three primary service models in cloud computing. Sensing as a service is also advocated in addition to these options. Due to cloud technology’s unique properties, communication, storage, setup, and management should be considered at the network edge, close to end users. Cloud computing and CRN’s cognitive skills, including dynamic spectrum access, can enable IoT services.
• CR with WSN: Recently, wireless sensor networks (WSNs) have shown their value in the real world. Because of their small size and low price, sensors have become an invaluable tool for collecting data. Their incorporation into IoT frameworks has been facilitated by the requirement for constant awareness everywhere. Spectrum availability and interference continue to hinder the gathering of information from these sensors in huge spatial deployments and domestic applications. In addition, sensors with limited resources generate and transmit unprocessed data instead of vast and valuable data streams. WSN data can be advantageous for localized applications, but may not be optimal for the internet of things. Consequently, WSNs and CRNs offer infrastructure-free, self-organized networks with intriguing IoT applications [37,38].
• CR with Wireless Sensor and Actuator Networks (WSANs): WSANs take action based on environmental sensing. Their ad hoc nature results in minimal physical exertion and, typically, a single task. The integration of WSANs in the IoT requires a broader perspective. A significant obstacle is the timely and efficient transmission of actions to actuators. Near-field communication (NFC) and RFID have been primarily adapted for Wireless Sensor and Actuator Networks; however, a cognitive info communication perspective is also being researched for inclusion in the IoT framework [39].
• CR with M2M and D2D Communication: The increasing number of devices has shifted the focus from human-to-human (H2H) communication to machine-to-machine (M2M) communication. Several devices, including mobile phones, laptops, and sensors, are capable of exchanging data without human intervention. Machine-to-machine (M2M) communication is crucial to the development of future internet of things frameworks because of the significance of machine data. Increases in the number of connected devices are becoming more common each year. Another difficulty is that the current state of the internet does not allow for the efficient transport of massive amounts of data from these devices. The vision centers on the incorporation of intelligence into these devices, with the expectation that they will act independently to set up self-configuring networks. Because of their vulnerability to interference, the conventional wireless methods proposed for M2M communication have been largely abandoned [40]. The integration of CRNs is also necessary for the efficient sharing of network resources between M2M and conventional H2H communications. D2D communication, on the other hand, can be utilized with CRNs to facilitate the internet of things because it enables devices to interact directly with one another without the need for a relay node.

5. Conclusions

There is a rising demand for IoT devices that are both intelligent and CRN-enabled in order to monitor surrounding radio environments, assess available frequencies, and determine when and where to transmit. As a result of this requirement, there has been a great deal of effort to create CR-based IoT systems as the standard paradigm for intelligent IoT systems by making use of CR’s cognitive capacities. This study has supplied the current frameworks of CR-based IoT systems, investigated the most recent SS and spectrum sharing methodologies, and emphasized the benefits and drawbacks of each. This research has also investigated the advantages of incorporating various new technologies, such ML methods, into CR-based IoT setups. Lastly, several obstacles, future research areas, and outstanding issues in development of CR-based IoT systems have been addressed and highlighted by this work.