9.1. Introduction

In wireless networks, field resources in terms of frequency and bandwidth availability are increasingly scarce. Consequently, new solutions are required to minimize the energy consumption and optimize the allocation of radio resources.

For flexible access to the spectrum, Mitola and Maguire (1999) introduced cognitive radio (CR) relying on software-defined radio. Software-defined radio is a radio that can realize in software form the typical functions of the radio interface generally realized in a hardware form, such as the carrier frequency, signal bandwidth and modulation. Indeed, Mitola and Maguire (1999) combined their software-defined radio experiences, as well as their passions for machine learning (ML) and artificial intelligence (AI) to set up CR technology. According to Mitola (2000), a CR is able to know, perceive and learn from its environment, and then act in order to simplify the user’s life. In 2005, Haykin (2005) reviewed the concept of CR and dealt with it as brain-empowered wireless communication.

CR is a technology that detects the environment, analyzes its transmission parameters, and then makes the decisions related to resource allocation and management. However, the formulations of optimization for resource allocation offer optimal solutions sometimes at the expense of the computing time and the processing complexity. To reduce this complexity and obtain resource allocation within reasonable, the cognitive radio network (CRN) must be equipped with learning and reasoning capacities. The cognitive engine must coordinate the CR actions using ML techniques.

Consequently, a CR must be intelligent and able to learn from its experience by interacting with its radiofrequency environment. Therefore, learning is an essential element of CR, which can be provided using AI and ML techniques. Indeed, applying AI, and in particular ML to CRN, has recently attracted considerable interest in the literature.

Learning aims to enable machines to carry out similar tasks to those of an expert. The intelligent machine senses its environment and takes measures to maximize its own utility. AI focuses on deduction, reasoning, problem solving, knowledge and learning representation (Woods 1986).

The learning process in the CRN is illustrated in Figure 9.1 and can be presented as follows (Abbas et al. 2015):

1. – detection of radiofrequency parameters such as the channel quality;
2. – observation of the environment and analysis of its reactions;
3. – learning;
4. – conservation of decisions and observations for model update;
5. – decision on resource management problems and consequent transmission error adjustment (Bkassiny et al. 2013; Russell 2016).

There are several works in the literature that focus on the application of AI to CRN. In Zhao and Morales-Tirado (2012), the authors introduced the use of AI and ML techniques in CR. They also presented the possible applications and the fundamental ideas of CR. In Bkassiny et al. (2013), the authors presented a survey on the various AI techniques, such as reinforcement learning (RL), game theory (GT), neural networks (NNs), support vector machines (SVM) and Markov models (MM). The survey discusses the strengths and weaknesses of these techniques, as well as the difficulties encountered in their applications in the CR field. In Gavrilovska et al. (2013), the authors studied the GT, RL and reasoning approaches such as Bayesian networks (BNs), fuzzy logic (FL) and case-based reasoning (CBR) in CRN. The survey presented by Abbas et al. (2015) was dedicated to FL, genetic algorithms (GA), NNs, GT, RL, SVM, CBR, decision trees (DT), BNs, MMs, multiagent systems (MASs) and artificial bee colonies algorithms.

It can however be noted that the article by Abbas et al. (2015) did not discuss the application of particle swarm optimization (PSO), a commonly used metaheuristic in CR. Other more recent metaheuristics that are also used in CR were ignored by Abbas et al. (2015), such as the firefly algorithm (FA), cuckoo search (CS) and the gravitational search algorithm (GSA).

This chapter focuses on the AI techniques that were most commonly used during the last three years in CR (the state-of-the-art of work before 2015 was already done in Abbas et al. (2015)).

Our focus is on work that was not discussed in previous surveys, such as the FA, CS, GSA and PSO. We also present recent work related to the application of other AI techniques in CR, namely GA, bee colony algorithms, FL, GT, NNs, MMs, SVM, CBR, DT, BNs, MASs and RL. Altogether, 17 AI techniques will be studied in this chapter.

The essential points dealt with in this chapter relate to:

• – the presentation of a complete study on AI techniques, their definitions and their applications to CR. It is worth noting that some techniques have never been discussed before in this field;
• – the presentation of the main tasks of CR and their corresponding challenges;
• – the categorization of the presented techniques depending on the type of learning (supervised or unsupervised) and their applications depending on the CR tasks.

The rest of this chapter is organized as follows. Section 9.2 presents the cognition cycle, the main CR tasks and their corresponding challenges. Section 9.3 offers a state-of-the-art on the application of AI methods to CR. Section 9.4 proposes a categorization of the techniques presented depending on the type of learning (supervised or unsupervised) and presents their applications depending on the CR tasks. Finally, section 9.5 is dedicated to the conclusion.

9.2. Cognitive radio

9.3. Application of AI in CR

9.3.1. Metaheuristics

In the literature, metaheuristics are classified into two subcategories: those with a single solution and those that are population based. Single solution metaheuristics are iterative approaches that start with a single initial solution and improve it with each iteration taking advantage of its neighborhood. Population-based metaheuristics explore the research space using a set of solutions known as “population”. This latter category is also divided into two subcategories: evolutionary algorithms and algorithms based on swarm intelligence.

Evolutionary algorithms draw their inspiration from the natural evolution of species, and more precisely from the natural selection principle stated in the theory of evolution, developed by Darwin (2009) in his work entitled On the Origin of Species: By Means of Natural Selection or the Preservation of Favored Races in the Struggle for Life. The algorithms based on swarm intelligence were introduced for the first time by Gerardo and Wang (1993) in their work entitled “Swarm intelligence in cellular robotics systems”, which describes the behavior of a group of robots cooperating to accomplish a task or solve a problem.

The following section presents the application of six metaheuristics most commonly used in CRN, namely FA, CS, bee colony algorithms, GA, GSA and PSO.

9.3.1.1. Firefly algorithm

The FA is an optimization approach based on swarm intelligence that was proposed by Xin-She Yang in 2008 (Surafel and Ngnotchouye 2017). Its principle is inspired by the lighting behavior of fireflies. Indeed, each potential solution is assimilated with a firefly, whose luminosity is proportionally related to its quality (quality of the solution). The following works dealt with the application of FA to Cognitive Radio Network (CRN).

The FA was adapted by the authors in a CRN based on orthogonal frequency-division multiplexing (OFDM) (Saoucha and Benmammar 2017). Multiobjective optimization was also used to optimize the quality of communication between secondary users (SU). The performances of the FA were validated through a comparison with the PSO algorithm and with the cross entropy in terms of convergence speed, quality of the solution and stability.

The FA was used in Tounsi and Babes (2017) to solve the problem of power control and channel allocation in CRN. A modified version of the FA, using the new attractiveness factor, was proposed to solve this problem. A theoretical analysis was presented in this article to prove the efficiency and the existence of Nash equilibrium concerning the proposed strategy. The results presented in this article show that the proposed method outperforms the approaches in the literature in terms of convergence speed.

To estimate the amplitudes of cancellation subcarriers, the authors in Elahi et al. (2017) proposed two search algorithms: the GA and the FA. According to the simulation results, the proposed algorithms enable a better reduction of the secondary lobes compared to the techniques currently mentioned by the literature.

In Ghanem et al. (2016), the problem of attack based on primary user emulation (PUE) is solved by means of a localization defense model based on the use of the FA. The CR users cooperate to detect and localize the attacker comparing its location with the position of the primary users (PU). The simulation results are compared with existing methods and prove that the FA reduces the localization error and requires fewer SUs to cooperate.

9.3.1.2. Cuckoo search

CS is an optimization approach based on swarm intelligence, which was proposed by Yang and Deb (2017). Its development draws inspiration from the parasitic behavior of certain bird species. In the real world, cuckoos lay their eggs in the nests of other bird species. In most cases, the host bird believes that the laid eggs are hers and, consequently, takes care of them. However, in certain cases the eggs laid are discovered and discarded by the host bird or the host bird leaves the nest.

The following works focus on the application of CS in the CRN.

A new method for the OFDM-based estimation of the channel state in a CRN was proposed in Manjith (2016). The method is hybridization between bacterial foraging optimization (BFO) and a modified CS algorithm.

In Kaur et al. (2018a), the authors proposed a new multiuser CR system, as well as its optimization by means of CS algorithm. The transmission parameters of several SUs are considered according to the IEEE 802.22 WRAN standard. The optimization results were compared to another efficient optimization technique based on biogeography and simulated annealing.

9.3.1.3. Bee colony algorithm

Various groups of researchers independently participated in the development of bee algorithms during the last 10 years. Tovey, at Georgia Tech, in collaboration with Nakrani, at the University of Oxford, proposed the honey bee algorithm for the first time in 2004 (Nakrani and Tovey 2004). The virtual bee algorithm was created in 2005 by Xin-She Yang at the University of Cambridge to solve numerical optimization problems (Yang 2005). Haddad et al. (2006) presented the honey bees mating optimization algorithm.

The artificial bee colony (ABC) algorithm was developed in 2005 by Karaboga (2005) for the optimization of numerical functions.

The following works focused on the application of bee colony algorithms in CRN.

A new spectral handover algorithm based on the ABC algorithm in a CRN was proposed in Bayrakdar and Calhan (2018). In the algorithm proposed by the authors, the spectrum availability characteristic is observed on the basis of bee missions in order to minimize the spectral handover delay and maximize the probability of finding an inactive channel. The main advantage of this algorithm is that the delay of the spectral handover of SUs is considerably reduced for a different number of users, without reducing the probability of finding an available channel.

A hybrid algorithm between artificial bee colonies and GA was proposed in Elghamrawy (2018) to optimize the use of spectrum by detecting the PUE attacks and increasing the detection probability. The proposed algorithm integrates the genetic operators with the ABC algorithm in order to reach equilibrium between use and exploration to find the optimal solution. The simulation results indicate promising performances of the proposed algorithm for spectrum detection optimization, compared to recent detection algorithms.

The main objective of Zaheer et al. (2016) is to minimize the transmission powers and thus reduce the interferences in a CRN by using the ABC algorithm.

9.3.1.4. Genetic algorithms

The principle of GA was introduced by John Holland at the University of Michigan in the United States, in the 1960s (Holland 1992), and highlighted by the reference work by Goldberg (1989). In a genetic algorithm, a population is constituted by a set of individuals, each of which is identified by a set of genes known as “chromosomes”. Reproduction involves the recombination of chromosomes of two primary individuals, thus giving birth to child individuals whose genetic fingerprint is inherited from parents. However, the genetic code of the children may contain genes that their parents do not have, thus modeling the mutation genetic phenomenon. The latter enables changes in the morphology of species, always leading to a better adaptation to the natural environment.

The following works focus on the application of GA to CRN.

The problem formulation, the development and use of a genetic algorithm for channel assignment in a CRN was presented by Elhachmi and Guennoun (2016). This approach offers the PU and SU an efficient means to access the available spectrum. Compared to existing methods, simulation results prove that the proposed algorithm yields satisfactory results in terms of interferences and throughput.

In Jiao and Joe (2016), the authors considered a new model of CRN in which the networks of PU are constituted of heterogeneous PU. The authors consider the problem of energy-efficient resource allocation for the CR user having a coverage area in which heterogeneous PU operate simultaneously via multiradio access technology. The authors proposed a research diagram based on two level crossover GA to obtain an optimal solution in terms of power and throughput. Simulation results show that the algorithm proposed by the authors is stable and its convergence is more rapid.

9.3.1.5. Gravitational search algorithm

In 2009, Rashedi et al. (2009) developed a GSA, which is an optimization metaheuristics inspired by nature. GSA relies on Newton’s law of gravitation, according to which gravitation is an attraction between massive bodies. The masses of the bodies (solutions) are proportional to their values of objective functions (costs). With each iteration, masses are mutually attracted by gravitational forces. The heaviest mass exerts the strongest force of attraction. Consequently, the heaviest masses, which are probably close to the global optimum, attract other masses depending on their distances. Each object is determined by four specifications: position, inertia, active gravitational mass and passive gravitational mass. The position corresponds to a solution to the problem; inertia and gravitational masses are determined using the objective function.

The following works focused on the application of the GSA in CRN.

Guo et al. (2018) present a new method for solving the problem of spectrum waste in CRN. This method relies on graph coloring and GSA. The authors compared the performances of their algorithm with the PSO and GA.

In Kaur et al. (2018b), a hybrid algorithm of PSO and gravitational research is presented for CRN optimization. A new CR environment is proposed, enabling several SUs to access the spectrum while their channels undergo Nakagami-m fading. The transmission factors belonging to several SUs and relying on IEEE 802.22 WRAN standard are optimized to reach several objectives related to the expected QoS using PSO, GSA and the hybrid PSOGSA algorithm. Objective functions that are modified and influenced by fading are established for the optimization task. The optimization results indicate an improved performance of the hybrid algorithm compared to other basic techniques.

9.3.1.6. Particle swarm optimization

PSO is an optimization metaheuristics in the family of evolutionary algorithms and proposed by Russel Eberhart (electrical engineer) and James Kennedy (sociopsychologist) (Kennedy and Eberhart 1995). PSO has its source in the observations made during computer simulations of flocks of birds and fish schools (Craig 1987; Heppner and Grenander 1990). Indeed, the PSO draws heavily on the observation of gregarious relations of migratory birds, which, in order to travel “long distances” (migration, search for food, aerial displays, etc.), must optimize their motions in terms of consumed energy and time (etc.), as for example in the V-shaped flight formation presented in Figure 9.2.

The following work focuses on the application of PSO in CRN.

A technique based on PSO and on the received signal strength indicator for the detection of the position of PU and the PUE–based attacker was proposed in Fihri et al. (2018). The authors aim to improve the detection accuracy and reduce the risk of false alarms.

The effect of eigenvalues of the covariance matrix on samples received through the SNR estimation method was analyzed in Manesh et al. (2017). The authors proposed the use of PSO in the SNR estimation technique based on eigenvalues in order to optimize these parameters. The results of the proposed method are compared to those of the original SNR estimation method and the results validate the improvement obtained by the proposed technique compared to the original technique.

The simultaneous wireless information and the multiple user power transfer for the CRN relying on PSO and on semidefinite relaxation were studied in Tuan and Koo (2017). A secondary emitter with an antenna array provides information and energy to several secondary receivers with a single antenna. The authors proved through simulations that their algorithm features rapid convergence and better performances compared to other existing systems.

In Zhai and Wang (2017), the authors used PSO to solve the crowdsourcing paradigm, according to which mobile users are assigned the task of spectrum detection. Simulation results show that the proposed algorithm reaches higher performances compared to those of other algorithms.

In Tang and Xin (2016), the authors used PSO to study the compromise between utility and energy consumption in a CRN based on OFDM. Considering the low convergence of the original PSO around local optima, an enhanced version combining chaos theory is proposed in this study in order to help PSO identify solutions around the best global results. Using simulations, the authors proved that their algorithm requires a smaller number of iterations and can reach a higher energy efficiency than the other algorithms.

Tuan and Koo (2016) proposed a hybrid method based on PSO and Brute-Force Search (BFS). This method is used for the maximization of the SU throughput in a full-duplex CRN, when it has two distinct antennas and a capacity to autoeliminate its interferences. The simulations show that for certain values of the parameters, the considered system provides a much higher throughput than the previously proposed systems.

Alhammadi et al. (2016) discussed the three spectral handover mechanisms (proactive, reactive and hybrid) used to reduce the handover delay. The article contains an implementation of the PSO algorithm to minimize the total service time of the spectral handover to optimal value. The numerical results show that PSO significantly reduces the total service time compared to otter spectral handover systems.

9.4. Categorization and use of techniques in CR

Data availability influences the choice of the learning technique to be used. Supervised learning is used when learning data are labeled and the CR has previous information on the environment. Unsupervised algorithms do not require labeled learning data. Unsupervised learning is used when certain RF components are not known by the CRN, which enables it to operate autonomously with no previous knowledge.

There are differences between learning techniques in terms of strengths, limitations, challenges and applications in CRN.

DTs, NNs, SVMs and CBR are considered supervised learning techniques. The GT, RL, MMs, metaheuristics, FL and BNs are considered unsupervised learning techniques.

It is worth noting that all the techniques studied in this chapter are used both for sensing and for decision in CR, except for the metaheuristics and for the CBR, which are used exclusively for decision. DTs, GT and FL should be used for sensing taking into account the capacity of the spectrum detection technique employed.

MAS can be used in combination with all the techniques presented in this chapter.

9.5. Conclusion

This chapter presents a complete study of the AI techniques used in CRN. The definitions of various techniques and their applications in CR are also discussed in this work. Indeed, a complete state-of-the-art on the application of the FA, CS, GSA, PSO, GA, bee colony algorithms, FL, GT, NNs, MMs, SVMs, CBR, DTs, BNs, MASs and RL in CRNs was presented in this chapter. It also presented the main CR tasks and their corresponding challenges, classified the presented techniques depending on the type of learning (supervised or unsupervised) and presented their applications depending on two CR tasks (sensing and decision).

9.6. References

Abbas, N., Nasser, Y., and El Ahmad, K. (2015). Recent advances on artificial intelligence and learning techniques in cognitive radio networks. EURASIP Journal on Wireless Communications and Networking (JWCN), 174 (2015), 1–20.

Abdalzaher, M.S., Seddik, K., and Muta, O. (2017). Using Stackelberg game to enhance cognitive radio sensor networks security. IET Communications, 11(9), 1503–1511.

Agarwal, A., Dubey, S., Khan, M. A., Gangopadhyay, R., and Debnath, S. (2016). Learning based primary user activity prediction in cognitive radio networks for efficient dynamic spectrum access. International Conference on Signal Processing and Communications. 12–15 June 2016, Bangalore, India, 1–5.

Ahmed, M.E., Kim, D.I., Kim, J.Y., and Shin, Y. (2017). Energy-arrival-aware detection threshold in wireless-powered cognitive radio networks. IEEE Transactions on Vehicular Technology, 66(10), 9201– 9213.

Alhammadi, A., Roslee, M., and Yusoff Alias, M. (2016). Analysis of spectrum handoff schemes in cognitive radio network using particle swarm optimization. 3rd International Symposium on Telecommunication Technologies. 28–30 November 2016, Kuala Lumpur, Malaysia, 103–107.

Banerjee, J.S., Chakraborty, A., and Chattopadhyay, A. (2017). Fuzzy based relay selection for secondary transmission in cooperative cognitive radio networks. Advances in Optical Science and Engineering. 279–287.

Bayrakdar, M.E. and Çalhan, A. (2018). Artificial bee colony–based spectrum handoff algorithm in wireless cognitive radio networks. International Journal of Communication Systems (IJCS), 31(5), 1–16.

Bellhouse, D. (2007). The problem of Waldegrave. Electronic Journal for the History of Probability and Statistics, 3(2), 1–12.

Beni, G., Wang, J. (1993). Swarm Intelligence in Cellular Robotic Systems. In Robots and Biological Systems: Towards a New Bionics?, Dario P., Sandini G., Aebischer P. (eds). Springer, Berlin, Heidelberg.

Bestaoui, A.A. (2015). Gestion de spectre dans un réseau de radio cognitive en utilisant l’algorithme d’optimisation par essaim de particules. Master’s thesis, University of Tlemcen, Algeria.

Biglieri, E., Goldsmith, A.J., Greenstein, L.J., Mandayam, N.B., and Poor, H.V. (2013). Principles of Cognitive Radio. Cambridge University Press, Cambridge.

Bkassiny, M., Li, Y., and Jayaweera, S.K. (2013). A survey on machine-learning techniques in cognitive radios. IEEE Communications Surveys & Tutorials, 15(3), 1136–1159.

Bolstad, W.M. (2004). Introduction to Bayesian Statistics. John Wiley & Sons, Hoboken.

Ching, W.-K. and Ng, M.K. (2006). Markov Chains. Models, Algorithms and Applications. Springer, Berlin.

Craig, W.R. (1987). Flocks, herds, and schools: A distributed behavioral model. Computer Graphics, 21(4), 25–34.

Darwin, C. (2009). The Origin of Species: By Means of Natural Selection, or the Preservation of Favoured Races in the Struggle for Life. Cambridge University Press, Cambridge.

Elahi, A., Qureshi, I.M., Atif, M., and Gul, N. (2017). Interference reduction in Cognitive radio networks using genetic and firefly algorithms. International Conference on Communication, Computing and Digital Systems.

Elderini, T., Kaabouch, N., and Reyes, H. (2017). Outage probability estimation technique based on a Bayesian model for cognitive radio networks. IEEE 7th Annual Computing and Communication Workshop and Conference.

Elghamrawy, S.M. (2018). Security in cognitive radio network: Defense against primary user emulation attacks using genetic artificial bee colony (GABC) algorithm. Future Generation Computer Systems (FGCS). 85(8), 1–19.

Elhachmi, J. and Guennoun, Z. (2016). Cognitive radio spectrum allocation using genetic algorithm. EURASIP Journal on Wireless Communications and Networking, 133 (2016), 1–11.

Fang, H., Xu, L., and Raymond Choo, K.-K. (2017). Stackelberg game based relay selection for physical layer security and energy efficiency enhancement in cognitive radio networks. Applied Mathematics and Computation, 296, 153–167.

Ferber, J. and Weiss, G. (1999). Multiagent Systems: An Introduction to Distributed Artificial Intelligence. Addison-Wesley, Boston.

Fihri, W.F., Arjoune, Y., El Ghazi, H., Kaabouch, N., and El Majd, B.A. (2017). Bayesian decision model with trilateration for primary user emulation attack localization in cognitive radio networks. International Symposium on Networks, Computers and Communications. 16–17 May 2017. Marrakech, Morocco, 1–6.

Fihri, W.F., Arjoune, Y., El Ghazi, H., Kaabouch, N., and El Majd, B.A. (2018). A particle swarm optimization based algorithm for primary user emulation attack detection. IEEE 8th Annual Computing and Communication Workshop and Conference. 8–10 January 2018, Las Vegas, USA, 823–827.

François-Lavet, V., Henderson, P., Islam, R., Bellemare, M.G., and Pineau, J. An introduction to deep reinforcement learning. Foundations and Trends® in Machine Learning, 11(3/4), 219–354.

Fraser, A.M. (2008). Hidden Markov Models and Dynamical Systems. Siam, Philadelphia.

Gavrilovska, L., Atanasovski, V., Macaluso, I., and Da Silva, L. (2013). Learning and reasoning in cognitive radio networks. IEEE Communications Surveys Tutor, 15(4), 1761–1777.

Ghanem, W.R., Shokair, M., and Desouky, M.I. (2016). An improved primary user emulation attack detection in cognitive radio networks based on firefly optimization algorithm. 33rd National Radio Science Conference. 22–25 February 2016, Aswan, Egypt, 178–187.

Ghasemi, A. and Sousa, E.S. (2008). Spectrum sensing in cognitive radio networks: Requirements, challenges and design trade-offs. IEEE Communications Magazine, 46(4), 32–39.

Goldberg, D.E. (1989). Genetic Algorithms in Search, Optimization and Machine Learning. Addison-Wesley Longman Publishing, Boston.

Guo, L., Chen, Z., and Huang, L. (2018). A novel cognitive radio spectrum allocation scheme with chaotic gravitational search algorithm. International Journal of Embedded Systems, 10(2), 161–167.

Haddad, O.B., Afshar, A., and Mariño, M.A. (2006). Honey-bees mating optimization (HBMO) algorithm: A new heuristic approach for water resources optimization. Water Resources Management, 20(5), 661–680.

Haykin, S. (2005). Cognitive radio: Brain-empowered wireless communications. IEEE Journal on Selected Areas in Communications, 23(2), 201–220.

Haykin, S. (2008). Neural Networks and Learning Machines. Pearson, London.

Heppner, F. and Grenander, U. (1990). A Stochastic Nonlinear Model for Coordinated Bird Flocks. AAAS Publication, Washington.

Holland, J.H. (1992). Adaptation in Natural and Artificial Systems: An Introductory Analysis with Applications to Biology, Control, and Artificial Intelligence. MIT Press, Cambridge.

Ichikawa, K. and Fujii, T. (2017). Radio environment map construction using hidden Markov model in multiple primary user environment. International Conference on Computing, Networking and Communications. 26–29 January 2017. Santa Clara, CA, USA, 272–276.

Jiao, Y. and Joe, I. (2016). Energy-efficient resource allocation for heterogeneous cognitive radio network based on two-tier crossover genetic algorithm. Journal of Communications and Networks, 18(1), 112–122.

Karaboga, D. (2005). An idea based on honey bee swarm for numerical optimization, volume 200. Technical report, Erciyes University.

Kaur, K., Rattan, M., and Singh Patterh, M. (2018a). Cuckoo search based optimization of multiuser cognitive radio system under the effect of shadowing. Wireless Personal Communications, 99(3), 1217–1230.

Kaur, K., Rattan, M., and Singh Patterh, M. (2018b). Cognitive radio design optimization over fading channels using PSO, GSA and hybrid PSOGSA. Second International Conference on Intelligent Computing and Control Systems. 14–15 June 2018, Madurai, India, 1700–1706.

Kennedy, J. and Eberhart, R.C. (1995). Particle swarm optimization. Proceedings of the IEEE International Conference on Neural Networks IV. 27 November–1 December 1995, University of Western Australia, Perth, 1942–1948.

Kolodner, J. (2014). Case-Based Reasoning. Morgan Kaufmann, Burlington.

Lall, S., Sadhu, A.K., Konar, A., Mallik, K.K., and Ghosh, S. (2016). Multiagent reinfocement learning for stochastic power management in cognitive radio network. International Conference on Microelectronics, Computing and Communications. 23 January 2016, Durgapur, India, 1–6.

Lin, Y., Wang, C., Wang, J., and Dou, Z. (2016). A novel dynamic spectrum access framework based on reinforcement learning for cognitive radio sensor networks. Sensors, 16(10), 1675.

Liu, Y., Liang, J., Xiao, N., Yuan, X., Zhang, Z., Hu, M., and Hu, Y. (2017). Adaptive double threshold energy detection based on Markov model for cognitive radio. PLOS ONE, 12(5), 1–18.

Liu, M., Song, T., Hu, J., Yang, J., and Gui, G. (2019). Deep learning-inspired message passing algorithm for efficient resource allocation in cognitive radio networks. IEEE Transactions on Vehicular Technology, 68(1), 641–665.

Lu, Y., Zhu, P., Wang, D., and Fattouche, M. (2016). Machine learning techniques with probability vector for cooperative spectrum sensing in cognitive radio networks. IEEE Wireless Communications and Networking Conference. April 3-6 2016, Doha, Qatar.

Manesh, M.R., Quadri, A., Subramaniam, S., and Kaabouch, N. (2017). An optimized SNR estimation technique using particle swarm optimization algorithm. IEEE 7th Annual Computing and Communication Workshop and Conference. 9–11 January 2017, Las Vegas, USA, 1–6.

Manjith, R. (2016). A hybrid of BFO and MCS algorithms for channel estimation of cognitive radio system. Arabian Journal for Science and Engineering, 41(3), 841–852.

Mitola, J. (2000). Cognitive radio: an integrated agent architecture for software defined radio. PhD Thesis, Royal Institute of Technology, Stockholm.

Mitola, J. and Maguire, G.Q. (1999). Cognitive radio: Making software radios more personal. IEEE Personal Communications, 6(4), 13–18.

Morozs, N., Clarke, T., and Grace, D. (2016). Cognitive spectrum management in dynamic cellular environments: A case-based Q-learning approach. Engineering Applications of Artificial Intelligence, 55, 239–249.

Nakrani, S. and Tovey, C. (2004). On honey bees and dynamic server allocation in internet hosting centers. Adaptive Behavior, 12(3–4), 223–240.

Norris, J.R. (1998). Markov Chains. Cambridge University Press, Cambridge.

Pourpeighambar, B., Dehghan, M., and Sabaei, M. (2017). Non-cooperative reinforcement learning based routing in cognitive radio networks. Computer Communications, 106, 11–23.

Qiu, R. C., Hu, Z., Li, H., and Wicks, M. C. (2012) Cognitive Radio Communication and Networking: Principles and Practice. John Wiley & Sons, Hoboken.

Rahman, M.A., Lee, Y.-D., and Koo, I. (2016). An efficient transmission mode selection based on reinforcement learning for cooperative cognitive radio networks. Human-centric Computing and Information Sciences, 6(1), 2.

Raj, V., Dias, I., Tholeti, T., and Kalyani, S. (2018). Spectrum access in cognitive radio using a two-stage reinforcement learning approach. IEEE Journal of Selected Topics in Signal Processing, 12(1), 20–34.

Rashedi, E., Nezamabadi-Pour, H., and Saryazdi, S. (2009). GSA: A gravitational search algorithm. Information Sciences, 179(13), 2232–2248.

Rojas, R. (2013). Neural Networks: A Systematic Introduction. Springer Science & Business Media, Berlin.

Roy, A., Midya, S., Majumder, K., Phadikar, S., and Dasgupta, A. (2017). Optimized secondary user selection for quality of service enhancement of two-tier multi-user cognitive radio network: A game theoretic approach. Computer Networks, 123, 1–18.

Russell, S.J. and Norvig, P. (2016). Artificial Intelligence: A Modern Approach. Pearson Malaysia, Kuala Lumpur.

Salahdine, F., Kaabouch, N., and El Ghazi, H. (2017a). Techniques for dealing with uncertainty in cognitive radio networks. IEEE 7th Annual Computing and Communication Workshop and Conference. 9–11 January 2017, Las Vegas, USA, 1–6.

Salahdine, F., Kaabouch, N., and El Ghazi, H. (2017b). A Bayesian recovery technique with Toeplitz matrix for compressive spectrum sensing in cognitive radio networks. International Journal of Communication Systems (IJCS), 30(15), 1–9.

Saoucha, N.A. and Benmammar, B. (2017). Adapting radio resources in multicarrier cognitive radio using discrete firefly approach. International Journal of Wireless and Mobile Computing, 13(1), 39–44.

Shu, Z., Qian, Y., Yang, Y., and Sharif, H. (2016). A game theoretic approach for energy-efficient communications in multi-hop cognitive radio networks. Wireless Communications and Mobile Computing, 16(14), 2131–2143.

Supraja, P. and Pitchai, R. (2019). Spectrum prediction in cognitive radio with hybrid optimized neural network. Mobile Networks and Applications, 24(2), 357–364.

Surafel, L.T. and Ngnotchouye, J.M.T. (2017). Firefly algorithm for discrete optimization problems: A survey. KSCE Journal of Civil Engineering, 21(2), 535–545.

Tang, M. and Xin, Y. (2016). Energy efficient power allocation in cognitive radio network using coevolution chaotic particle swarm optimization. Computer Networks, 100, 1–11.

Tounsi, A. and Babes, M. (2017). An efficient joint spectrum and power allocation in cognitive radio networks using a modified firefly algorithm. International Journal of Communication Networks and Distributed Systems, 19(2), 214–236.

Tuan, P. and Koo, I. (2016). Throughput maximisation by optimising detection thresholds in full-duplex cognitive radio networks. IET Communications, 10(11), 1355–1364.

Tuan, P. and Koo, I. (2017). Robust weighted sum harvested energy maximization for SWIPT cognitive radio networks based on particle swarm optimization. Sensors, 17(10), 2275.

Umar, R. and Sheikh, A.U.H. (2012). Cognitive radio oriented wireless networks: Challenges and solutions. Proceedings of the International Conference on Multimedia Computing and Systems. 10–12 May 2012. Tangier, Morocco, 992–997.

Wang, B. and Liu, K.J.R. (2011). Advances in cognitive radio networks: A survey. IEEE J. Selected Topics Signal Process, 5(1), 5–23.

Wang, D. and Yang, Z. (2016a). A novel spectrum sensing scheme combined with machine learning. 9th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics. 15–17 October 2016, Datong, China, 1293–1297.

Wang, D. and Yang, Z. (2016b). An advanced scheme with decision tree for the improvement of spectrum sensing efficiency in dynamic network. 9th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics. 15–17 October 2016, Datong, China, 1288–1292.

Wiering, M. and Van Otterlo, M. (2012). Reinforcement learning. Adaptation, Learning, and Optimization, 12, 3.

Woods, W.A. (1986). Important issues in knowledge representation. Proceedings of the IEEE, 74(1), 1322–1334.

Wyglinski, A.M., Nekovee, M. and Hou, T. (eds) (2009). Cognitive Radio Communications and Networks: Principles and Practice. Academic Press, Cambridge.

Xingjian, L. et al. (2018). Intelligent power control for spectrum sharing in cognitive radios: A deep reinforcement learning approach. IEEE Access, 6, 25463–25473.

Yang, X.-S. (2005). Engineering optimizations via nature-inspired virtual bee algorithms. International Work-Conference on the Interplay Between Natural and Artificial Computation. Las Palmas, Canary Islands, Spain, 317–323.

Yang, X.-S. and Deb, S. (2017). Cuckoo search state-of-the-art and opportunities. IEEE 4th International Conference on Soft Computing & Machine Intelligence. 23–24 November 2017, Mauritius, 55–59.

Yin, Z., Wang, Y., and Wu, C. (2017). A multiagent collaborative model for bayesian opportunistic channel accessibility in railway cognitive radio. International Journal of Performability Engineering (IJPE), 13(4), 479–489.

Zadeh, L.A. (1965). Fuzzy sets. Information and Control, 8(3), 338–353.

Zaheer, M., Uzma, M., Asif, A., and Qureshi, I. M. (2016). Interference control in cognitive radio using joint beamforming and power optimization by applying artificial bee colony. 19th International Multi-Topic Conference. 5–6 December 2016. Islamabad, Pakistan, 1–6.

Zhai, L. and Wang. H. (2017). Crowdsensing task assignment based on particle swarm optimization in cognitive radio networks. Wireless Communications and Mobile Computing (WCMC). 4687974, 1–9.

Zhang, M., Diao, M., and Guo, L. (2017). Convolutional neural networks for automatic cognitive radio waveform recognition. IEEE Access, 5, 11074–11082.

Zhao, Y. and Morales-Tirado, L. (2012). Cognitive radio technology: Principles and practice. International Conference on Computing, Networking and Communications. 30 January–2 February 2012. Maui, USA, 650–654.

Zhu, J., Jiang, D., Ba, S., and Zhang, Y. (2017). A game-theoretic power control mechanism based on hidden Markov model in cognitive wireless sensor network with imperfect information. Neurocomputing, 220, 76–83.