Sage Journals: Discover world-class research

Abstract

In order to meet the practical requirement for Cognitive Wireless Sensor Networks applications, this paper proposes innovative fast channel selection algorithm to solve the shortcomings of original Experience-Weighted Attraction algorithm's complexity, higher energy consuming, and the nodes’ hardware restrictions of real-time data processing capabilities. Research is conducted by comparing channel selection differences and timeliness with traditional Experience-Weighted Attraction learning. Though not as stable as traditional Experience-Weighted Attraction learning, fast channel selection algorithm has effectively reduced the complexity of the original algorithm and has superior performance than Q learning.

1. Introduction

Traditionally, the licensed radio spectrum allocations are regulated by official authorities. The public and government use of radio spectrum is managed by the National Telecommunications and Information Administration (NTIA) and the Federal Communications Commission (FCC) is in charge of commercial radio resources, respectively, in the USA. With more and more applications of wireless devices, the rapid increasing requisition for radio spectrum licensing has led to current shortage of radio spectrum allocations and put their governing bodies into trouble. In fact, FCC's recent research has shown that these fixed static frequency channels are always idle or not occupied most of the time. Spectrum bands are not efficiently used or under utilization either at a temporal or on a geographical level. By seeking “spectrum holes” (unused frequency channels), Cognitive Radio (CR) can greatly improve the use efficiency of spectrum resources and solve these problems presented above in a “secondary utilization” (with lower priority than legacy users) way. First introduced by Mitola III [1], Cognitive Radio (CR) is often considered as an extension and expansion of Soft Radio (SR), which is equipped by general hardware and capable of programming to transmit and receive various radio waves.

There has already been lots of research in many aspects of CR. In sensing, Panahi and Ohtsuki [2] present a Fuzzy Q Learning (FQL) based scheme for channel sensing in CR networks. Zhang et al. [3] proposed a novel detection algorithm in which the fractal box dimension is used when the Signal to Noise Ratio (SNR) is high, while the improved TCC algorithm is used when the SNR is low, and Khalaf [4] formulated the detection problem based on the eigendecomposition technique. Hossain et al. [5] evaluated the performance of cooperative spectrum sensing with the hard combination OR, AND, and MAJORITY rules. Bkassiny et al. [6] presented an autonomous CR architecture, referred to as the Radiobot, to detect and identify the sensed signals. Lunden et al. [7] also distributed multiuser multiband spectrum sensing policies for CR networks based on multiagent reinforcement learning while Reinforcement Learning-Based Cooperative Sensing (RLCS) method was proposed to address the cooperation overhead problem and improve cooperative gain in CR ad hoc networks. [8] In channel allocation, Gállego et al. [9] presented a game theoretic solution for joint channel allocation and power control in CR networks analyzed under the physical interference model. In channel access, Teng et al. [10] demonstrated a reinforcement learning-based double auction algorithm aiming to improve the performance of dynamic spectrum access in CR networks. In security, Wang et al. [11] proposed a Four-Dimensional Continuous Time Markov Chain model to analyze the communication performance of normal Secondary Users under PUEAs, typically affected by SMUs, and compared several PUEA detection schemes.

As revolutionary development of Intelligent Radio (IR), CR implements Soft Radio by adding Knowledge Base, Reasoning Engine, and Learning Engine to be an independent Cognitive Engine (CE), which makes the radio capable of learning and adapting to the surrounding radio environment [12]. Knowledge Base which stores variety of cases, relations, and rules can be seen as memory in human's brain and is very common in Artificial Intelligence (AI) logic planning. Just like expert system in Artificial Intelligence, Reasoning Engine executes all kinds of state information for reference of Knowledge Base by logic thinking and then generates processed results or actions to drive Soft Radio changing setting parameters to adapt to changing environment. As the core component and key feature for CR implementation, Learning Engine is in charge of keeping Knowledge Base updated by accumulating new environmental experience into new knowledge extension, which is what differentiates CR from traditional preprogrammed ones.

There are varieties of learning algorithms available for CR, including neural networks, genetic models, and hidden Markov algorithms [13]. Bkassiny et al. characterized the learning problem in CR and state the importance of Artificial Intelligence in achieving real cognitive communications systems [14] and proposed a Bayesian nonparametric signal classification approach for spectrum sensing in CR [15]. Bizhani and Ghasemi [16] used Multiresponse Learning Automata (MRLA) to control how Secondary Users should access the licensed primary channels in CR networks. Tsagkaris et al. [17] used neural network-based learning to predict data bit rate of CR. Galindo-Serrano and Giupponi [18] proposed a form of real-time decentralized Q learning to manage the aggregated interference generated by multiple WRAN systems. Li [19] applied Multiagent Reinforcement Leaning (MARL) for the Secondary Users to learn good strategies of channel selection. Chen et al. [20] presented an intelligent policy based on reinforcement learning to acquire the stochastic behavior of Primary Users (PUs). Zhang and Liu [21] obtained the capability of iteratively online learning environment performance by using Reinforcement Learning (RL) algorithm after observing the variability and uncertainty of the heterogeneous wireless networks. Gállego et al. [9] provided no-regret learning algorithms to perform the joint channel and power allocation and overcome the convergence limitations of the local game. Zhu et al. [22] employed Reinforcement Learning (RL) approach to find a near-optimal policy under undiscovered environment. Torkestani and Meybodi [23] proposed the learning automata-based CR to address the spectrum scarcity challenges in wireless ad hoc networks. Yang and Grace [24] improved channel assignment in multicast terrestrial communication systems with distributed channel occupancy detection by using intelligence based on reinforcement learning and transmitter power adjustment. Zhou et al. [25] designed a robust distributed power control algorithm with low implementation complexity for CR networks through reinforcement learning, which does not require the interference channel and power strategy information among Secondary Users (SUs) and from SUs users to PUs.

However, as known with our best effort till now, little focus has been placed on implementing Learning Engine of CR with Experience-Weighted Attraction (EWA) algorithms. The innovative proposed channel selection algorithm based on EWA learning [26, 27] allows cognition to learn radio environment communication channel characteristics online. By accumulating the history channel experience, it can predict, select, and change the current optimal communication channel, dynamically ensure the quality of communication links, and finally reduce system communication outage probability. The effectiveness of this algorithm has been validated by simple probability method [26] and with handoff scheme [27] in our preliminary studies. However, it is not applicable for processing capability and power-restricted nodes of Wireless Sensors Networks (WSNs) due to original EWA algorithm's high complexity and energy consuming. Based on our lots of earlier research, the study focus has been shifted to fast channel selection algorithm EWAS with low complexity and green energy. The rest of this paper is presented as follows. In Section 2, EWAS algorithms will be introduced in full detail; then the simulation results comparison and analysis are presented in Section 3. In the end, the conclusion comes in Section 4.

2. Fast Cognitive Channel Selection Model

In the problem of radio communication channel selection, different wireless channels should have different channel availabilities; that is, the idle probabilities α of difference channel should not be the same for CR. Assuming radio propagation environment can be divided into n channels, then the idle probability of channel $i (1 \leq i \leq n)$ can be expressed as $α_{i}$ , or $Α = [α_{1}, α_{2}, \dots, α_{n - 1}, α_{n}]$ in vector form. Let $β_{i}$ be the successful transmission probability of channel $i (1 \leq i \leq n)$ ; then $Β = {β_{1}, β_{2}, \dots, β_{n - 1}, β_{n}}$ . Think of the radio channel characteristics change over time; the channel idle probability and successful transmission probability of channel $i (1 \leq i \leq n)$ should not be the same at different time t; then the forms of probabilities after introducing time parameter t are $Α (t) = {α_{1} (t), α_{2} (t), \dots, α_{n - 1} (t), α_{n} (t)}$ and $Β (t) = {β_{1} (t), β_{2} (t), \dots, β_{n - 1} (t), β_{n} (t)}$ , respectively.

To reduce the complexity of channel selection strategy based on EWA learning algorithm, exponential operation should be firstly avoided. Next the fast algorithm should simplify the calculation procedure and optimize and update the objective function directly in ideal. This paper calculates and carries iterative operation directly on channel selection probabilities and innovatively proposes a fast simplified cognitive channel selection algorithm EWAS.

Define the probability of selecting channel j in channel preferable selection policy $s_{i}^{j}$ at time t as $P_{i}^{j} (t)$ ; then the mathematical expression of $P_{i}^{j} (t)$ is

\begin{array}{l} P_{i}^{j} (t + 1) = \frac{1 - σ}{1 - σ \cdot \{1 - I [s_{i}^{j}, s_{i} (t)]\}} \\ \cdot 〈 \frac{(1 - τ) \cdot P_{i}^{j} (t)}{1 - σ \cdot \{1 - π_{i} [s_{i}^{j}, s_{- i} (t)]\}} + π_{i} [s_{i}^{j}, s_{- i} (t)] \\ \cdot τ 〉 + I [1, x (j)] \cdot I [s_{i}^{j}, s_{i} (t)] \cdot π_{i} [s_{i}^{j}, s_{- i} (t)] \cdot σ, \end{array}

(1)

where

\begin{matrix} x (j) = \{\begin{cases} 0, & Transmission failure on channel j, \\ 1, & Successful transmission on channel j, \end{cases} \\ π_{i} [s_{i}^{j}, s_{- i} (t)] = \{\begin{cases} 0, & channel j is sensed busy, \\ 1, & channel j is sensed idle \end{cases} \end{matrix}

(2)

and

I [\cdot]

is the indicator function, which is defined as follows:

\begin{matrix} I (x, y) = \{\begin{cases} 1, & x = y, \\ 0, & x \neq y . \end{cases} \end{matrix}

(3)

Parameters σ and τ are attenuation coefficients of probability and $σ < τ \in (0,1)$ . As can be seen through in-depth analysis of (1), in the period of radio environment sensing of CR, when perceiving the current state of the channel j being busy (strong electromagnetic noise over interference threshold for transmission), the state flag status is set to 0 (unavailable), and the strategy of selecting channel j for transmission channel will get no payoff, or the award function value of $π_{i} [s_{i}^{j}, s_{- i} (t)]$ is 0 and channel selection probability declines to $(1 - τ) \cdot P_{i}^{j} (t)$ ; while perceiving the current state of the channel j being idle (electromagnetic noise below interference threshold for transmission), the state flag status is set to 1 (available), and the strategy of selecting channel j for transmission channel will get the payoff of $π_{i} [s_{i}^{j}, s_{- i} (t)]$ , respectively. In addition, the value of $π_{i} [s_{i}^{j}, s_{- i} (t)]$ is assumed to equal 1 and channel selection probability is updated to $(1 - τ) \cdot P_{i}^{j} (t) + τ$ .

These available channels are candidate channels for channel selection of CR, and the candidate channel with the highest probability (if more than one channel reaches the highest selection probability, then one of these channels will be selected randomly) of channel selection will be chosen for transmission. After successful transmission, this channel selection probability will go up to $(1 - σ) \cdot [(1 - τ) \cdot P_{i}^{j} (t) + τ] + σ$ . But if the transmission is unfortunately failure, the channel selection probability will be $(1 - σ) \cdot [(1 - τ) \cdot P_{i}^{j} (t) + τ]$ .

At this point, it can be seen that the complexity of EWAS fast channel selection algorithm is $O (n)$ . Due to exponentiation operation, EWA's complexity is $O (n^{2})$ eventually.

3. Results and Discussion

Assume the number of channels in simulation environment is 5, or $n = 5$ . For the coefficients, τ is set to the default value 0.1 according to general experience. Since the value of σ should be lower than parameter τ, we pick half value of $τ δ$ for coefficient σ in this paper; that is, $σ = τ / 2 = 0.05$ . While there shall be some differences between each channel, the idle probabilities of these channels will not be the same. To reflect the general channels’ available probabilities, uniform distribution vector in the range of 0 to 1 will be selected for the idle probability of each channel; that is, the initial channel idle probability vector $Α_{0} = {0.4,0.9,0.6,0.5,0.7}$ , while the initial channel successful transmission probability vector $Β_{0} = {3 / 4,8 / 9,5 / 6,4 / 5,6 / 7}$ . Then the initial channel available probability vector $Γ_{0} = Α_{0} \cdot Β_{0} = {0.3,0.8,0.5,0.4,0.6}$ . In order to verify that this intelligent algorithm is capable of deciding and guiding CR real-time switch to the new transmission channel with the highest available probability online accurately, the channel idle probability vector will change to $Α_{1} = {0.6,0.4,0.7,0.9,0.5}$ , and the channel successful transmission probability vector will change to $Β_{1} = {5 / 6,3 / 4,6 / 7,8 / 9,4 / 5}$ after 33 rounds during the simulation process. Therefore the channel available probability vector will be $Γ_{1} = Α_{1} \cdot Β_{1} = {0.5,0.3,0.6,0.8,0.4}$ after the simulation environment change. Taking suddenness and randomness of the above parameters under actual wireless environment into account, the value generated in each simulation round meets exponential distribution of the corresponding parameter above followed by the general rule.

In this paper, a simple repeated experimental method is applied to verify the effectiveness of probability of channel selection algorithm based on EWA learning. That is, Turn-Based Strategy (TBS), a single uniformly distributed random number within range $[0,1]$ , is generated in each round. If this number is less than the channel available probability $α_{i}$ , channel i is judged as idle available state; else it is busy unavailable state. Idle channel with the highest selection probability will be the preferable communication channel in the current round. If more than one channel reaches the highest probability of channel selection, then one of these channels will be selected randomly. After algorithm selects preferable channel j, a single uniformly distributed random number within range $[0,1]$ is also generated. Communication channel transmission is successful if this number is less than the probability of successful data transfer completion $β_{j}$ ; otherwise it fails.

After the parameters above are set, the track records of channel selection probability based on EWA learning are shown in Figure 1.

Figure 1

Channel selection probability based on EWA learning.

In Figure 1, EWAS learning algorithm randomly selects channel 5 as the access channel in the condition of the same initial channel selection probabilities. After short initialization process, EWAS learning algorithm can successfully track and lock channel 2 as its preferable channel and its selection probability fluctuates slightly around 0.87. For the reason of channel availability, probability changes after 36th round and the selection probability of channel 2 falls dramatically, while the selection probability of channel 4 increases, respectively, and steadily overtakes the selection probability of channel 2 after 40 rounds. Channel 4 eventually replaces channel 2 to become optimal access channel under new channel available probability states.

In order to highlight better performance of channel selection algorithm based on EWA learning than other traditional radio with fixed transmission channel, the times of availability to access the channel and successful completion of transmissions in 100 rounds are collected in 3 scenes: fixed channel 2 as transmission channel, fixed channel 4 as transmission channel, and channel selection algorithm based on Q learning channel selection algorithm based on EWA learning and channel selection algorithm based on EWAS learning. The statistical data is compared in Figure 2.

Figure 2

Comparison between EWA/EWAS learning and reference ploys.

The number of availabilities to access the channel with fixed channel 2 as transmission channel is 63, and the number of successful completion of transmissions with fixed channel 2 as transmission channel is 54; the number of availabilities to access the channel with fixed channel 4 as transmission channel is 74, and the number of successful completion of transmissions with fixed channel 4 as transmission channel is 65. The numbers of availabilities to access the channel with channel selection algorithm based on Q, EWA, and EWAS learning are the same as 100 with no block, but the numbers of successful completion of transmissions are 80 for Q learning, 81 for EWAS, and 82 for EWA learning, respectively. Finally, the probability of successful completion of transmission with channel selection algorithm based on EWA learning is 81%, much higher than that of channel 2 (54%) and channel 4 (65%). By evident statistical comparison, channel selection algorithm based on EWAS learning can greatly improve the probabilities of successful channel access and transmission completion, much the same as Q (80%) and EWA (82%) learning. But its advantage has more intuitive reflection on the comparison chart of real-time channel selection below.

In order to highlight better performance of channel selection algorithm based on EWAS learning than Q learning, the channel selection tracks based on both learning algorithms are recorded under the same initial states and radio environments. The results are illustrated in Figure 3.

Figure 3

The tracks of channel selection based on EWA learning and Q learning.

Note that channel number 0 indicates that full channel blocking occurs, which means all the channels are in busy states and are not available for communication which is the situation in the 21st round. The differences between EWAS learning and EWA learning algorithms are mainly in the transition period (36th–45th round) of switching channel from former selected channel 2 to new optimal channel 4, and this reflects the differences between these two different algorithms. However sudden channel change of fast algorithm in the 86th round arouses our big interest. In order to analyze the reason of this phenomenon, the channel selection probability records after each round are derived and shown in Tables 1 and 2.

Table 1

The probability table of channel selection based on EWAS learning.

Iteration number	85	86	87	88	89	90	91
Channel 1	0.5304	0.5774	0.6197	0.6577	0.6919	0.7227	0.6505
Channel 2	0.4355	0.3920	0.3528	0.3175	0.3858	0.4472	0.4025
Channel 3	0.7985	0.8187	0.8449	0.8674	0.8867	0.9031	0.8171
Channel 4	0.8341	0.8131	0.8318	0.8486	0.8638	0.8774	0.8897
Channel 5	0.4589	0.4130	0.3717	0.3345	0.4011	0.4610	0.4149

Table 2

The probability table of channel selection based on EWA learning.

Iteration number	85	86	87	88	89	90	91
Channel 1	0.1802	0.1862	0.1887	0.1909	0.1900	0.1892	0.1862
Channel 2	0.1731	0.1717	0.1677	0.1641	0.1659	0.1674	0.1668
Channel 3	0.2033	0.2075	0.2080	0.2084	0.2057	0.2032	0.1985
Channel 4	0.2690	0.2617	0.2669	0.2716	0.2718	0.2719	0.2811
Channel 5	0.1745	0.1729	0.1687	0.1650	0.1667	0.1682	0.1675

Table 1 records channel selection probabilities values calculated by EWAS algorithm after each round, while Table 2 presents channel selection probabilities values based on EWA learning. It can be seen from Table 1 that selection probabilities of channel 3 and channel 4 are very close to each other from the 85th round to the 91st round. After transmission failure of preferred channel 4 in the 85th round, selection probability of current channel 4 falls from 0.8341 to 0.8131, while selection probability of channel 3 increases from 0.7985 up to 0.8187 and weakly overtakes channel 4 to be new selected transmission channel by EWAS fast algorithm. Even if the same trend in the probability changes, the selection probability of channel 4 calculated by EWA learning is still the largest of all in the 86th round and channel 4 being the optimal transmission channel remains unchanged. In summary, channel selection based on EWAS fast algorithm has the same performance in fast tracking, locking, and switching to the current optimal channel from changing communication environment and is superior to Q learning algorithm even not as much stable as original EWA algorithm.

4. Conclusion

In this paper, an innovative fast channel selection algorithm EWAS is proposed to solve the shortcomings of original EWA algorithm's complexity, higher energy consuming, and the nodes’ hardware restrictions of real-time data processing capabilities in order to meet the practical requirement for Cognitive Wireless Sensor Networks (CWSNs) application. Research is conducted by comparing channel selection differences and timeliness with traditional Q learning and EWA algorithm. Though not as stable as EWA learning, fast channel selection algorithm EWAS has effectively reduced the complexity of the original algorithm and has superior performance than Q learning. However, EWAS algorithm is of passive channel detection and access; future research is lying on active channel state prediction and reallocation in application of Wireless Sensor Network (WSN).

Footnotes

Conflict of Interests

The authors declare that there is no conflict of interests regarding the publication of this paper.

Acknowledgment

This work was supported by the National High Technology Research and Development Program of China (863 Program) (no. 2012AA062103).

References

Mitola

III

Cognitive radio for flexible mobile multimedia communications

Mobile Networks and Applications 2001 6 5 435 441

10.1023/a:1011426600077

2-s2.0-0035444223

Panahi

F. H.

Ohtsuki

Optimal channel-sensing scheme for cognitive radio systems based on fuzzy q-learning

IEICE Transactions on Communications 2014 97 2 283 294

10.1587/transcom.e97.b.283

2-s2.0-84893294333

Zhang

Xiao

An improved cognitive radio spectrum sensing algorithm

TELKOMNIKA Indonesian Journal of Electrical Engineering 2013 11 2 583 590

10.11591/telkomnika.v11i2.1980

Khalaf

G. A.

An optimal sinsing algorithm for multiband cognitive radio network

International Journal of Information and Network Security 2013 2 1 60 67

10.11591/ijins.v2i1.1473

Hossain

M. S.

Abdullah

M. I.

Hossain

M. A.

Hard combination data fusion for cooperative spectrum sensing in cognitive radio

International Journal of Electrical and Computer Engineering 2012 2 6 811 818

10.11591/ijece.v2i6.1814

Bkassiny

Jayaweera

S. K.

Avery

K. A.

Wideband spectrum sensing and non-parametric signal classification for autonomous self-learning cognitive radios

IEEE Transactions on Wireless Communications 2012 11 7 2596 2605

10.1109/twc.2012.051512.111504

2-s2.0-84864131832

Lunden

Kulkarni

S. R.

Koivunen

Poor

H. V.

Multiagent reinforcement learning based spectrum sensing policies for cognitive radio networks

IEEE Journal on Selected Topics in Signal Processing 2013 7 5 858 868

10.1109/JSTSP.2013.2259797

2-s2.0-84884508436

B. F.

Akyildiz

I. F.

Reinforcement learning for cooperative sensing gain in cognitive radio ad hoc networks

Wireless Networks 2013 19 6 1237 1250

10.1007/s11276-012-0530-4

2-s2.0-84880306616

Gállego

J. R.

Canales

Ortín

Distributed resource allocation in cognitive radio networks with a game learning approach to improve aggregate system capacity

Ad Hoc Networks 2012 10 6 1076 1089

10.1016/j.adhoc.2012.02.002

2-s2.0-84859759867

10.

Teng

Y. L.

F. R.

Han

Wei

Y. F.

Zhang

Reinforcement-learning-based double auction design for dynamic spectrum access in cognitive radio networks

Wireless Personal Communications 2013 69 2 771 791

10.1007/s11277-012-0611-9

2-s2.0-84879688708

11.

Wang

S.-S.

Luo

X.-G.

B.-N.

Primary user emulation attacks analysis for cognitive radio networks communication

TELKOMNIKA Indonesian Journal of Electrical Engineering 2013 11 7 3905 3914

10.11591/telkomnika.v11i7.2840

12.

Bantouna

Stavroulaki

Kritikou

Tsagkaris

Demestichas

Moessner

An overview of learning mechanisms for cognitive systems

EURASIP Journal on Wireless Communications and Networking 2012 2012, article 22

10.1186/1687-1499-2012-22

2-s2.0-84872847606

13.

Gavrilovska

Atanasovski

Macaluso

DaSilva

L. A.

Learning and reasoning in cognitive radio networks

IEEE Communications Surveys and Tutorials 2013 15 4 1761 1777

10.1109/surv.2013.030713.00113

2-s2.0-84888387564

14.

Bkassiny

Jayaweera

S. K.

A survey on machine-learning techniques in cognitive radios

IEEE Communications Surveys and Tutorials 2013 15 3 1136 1159

10.1109/SURV.2012.100412.00017

2-s2.0-84881317360

15.

Bkassiny

Jayaweera

S. K.

Multidimensional dirichlet process-based non-parametric signal classification for autonomous self-learning cognitive radios

IEEE Transactions on Wireless Communications 2013 12 11 5413 5423

10.1109/twc.2013.092013.120688

2-s2.0-84895062080

16.

Bizhani

Ghasemi

Joint admission control and channel selection based on multi response learning automata (MRLA) in cognitive radio networks

Wireless Personal Communications 2013 71 1 629 649

10.1007/s11277-012-0834-9

2-s2.0-84879420116

17.

Tsagkaris

Katidiotis

Demestichas

Neural network-based learning schemes for cognitive radio systems

Computer Communications 2008 31 14 3394 3404

10.1016/j.comcom.2008.05.040

2-s2.0-49649129002

18.

Galindo-Serrano

Giupponi

Distributed Q-learning for aggregated interference control in cognitive radio networks

IEEE Transactions on Vehicular Technology 2010 59 4 1823 1834

10.1109/tvt.2010.2043124

2-s2.0-77952245702

19.

H. S.

Multiagent Q-learning for aloha-like spectrum access in cognitive radio systems

Eurasip Journal on Wireless Communications and Networking 2010 2010 15

876216

10.1155/2010/876216

2-s2.0-77955287167

20.

Chen

X. F.

Zhao

Zhang

Chen

Reinforcement learning enhanced iterative power allocation in stochastic cognitive wireless mesh networks

Wireless Personal Communications 2011 57 1 89 104

10.1007/s11277-010-0008-6

2-s2.0-79951944633

21.

Zhang

W. Z.

Liu

X. C.

Centralized dynamic spectrum allocation in cognitive radio networks based on fuzzy logic and q-learning

China Communications 2011 8 7 46 54

2-s2.0-83455259991

22.

Zhu

Wang

Luo

Adaptive transmission scheduling over fading channels for energy-efficient cognitive radio networks by reinforcement learning

Telecommunication Systems 2009 42 1-2 123 138

10.1007/s11235-009-9174-9

2-s2.0-69549108237

23.

Torkestani

J. A.

Meybodi

M. R.

A learning automata-based cognitive radio for clustered wireless ad-hoc networks

Journal of Network and Systems Management 2011 19 2 278 297

10.1007/s10922-010-9178-5

2-s2.0-79952191041

24.

Yang

M. F.

Grace

Cognitive radio with reinforcement learning applied to multicast downlink transmission with power adjustment

Wireless Personal Communications 2011 57 1 73 87

10.1007/s11277-010-0007-7

2-s2.0-79951941826

25.

Zhou

Chang

Y. S.

Copeland

J. A.

Reinforcement learning for repeated power control game in cognitive radio networks

IEEE Journal on Selected Areas in Communications 2012 30 1 54 69

10.1109/jsac.2012.120106

2-s2.0-84855426943

26.

Sun

Qian

J.-S.

Cognitive radio channel selection strategy based on experience-weighted attraction learning

TELKOMNIKA Indonesian Journal of Electrical Engineering 2014 12 1 149 156

10.11591/telkomnika.v12i1.3900

27.

Sun

Qian

J. S.

EWA selection strategy with channel handoff scheme in cognitive radio

Sensors & Transducers 2014 173 6 68 74

Fast Channel Selection Strategy in Cognitive Wireless Sensor Networks

Abstract

1. Introduction

2. Fast Cognitive Channel Selection Model

3. Results and Discussion

4. Conclusion

Footnotes

Conflict of Interests

Acknowledgment

References