This paper considers the problem of opportunistic spectrum access in a slow flat fading stationary wireless environment, in which a cognitive agent with a single transceiver uses existing licensed spectrum to find the best possible channel to use. Each wireless channel is modeled as a k-state first order Markov chain, with one vacant (good) state and k-1 faded (bad) states. In addition, each channel is assumed as irreducible, aperiodic, and non-reactive to cognitive agent actions. An algorithm is proposed based on posterior sampling that learns best channel to use without any prior knowledge about wireless environment. The idea is to assign a Dirichlet distribution to each state in every channel, and then greedily select a channel with the highest limiting probability of being in a vacant state. An error correction scheme was also proposed to mitigate misclassification of deeply faded states, which occurs when transmitter unintentionally interferes with licensed users. Simulations of the proposed algorithm were presented at the end this paper, which show superior performance compared to other algorithms in case of perfect detection. Also, simulations of the proposed error correction scheme show an improved performance as cooperation with other agents increases. To the best our knowledge, this is the first paper to use learning algorithms for fading channel that is modeled as K-state Markov chain.
[1]
E. Gilbert.
Capacity of a burst-noise channel
,
1960
.
[2]
P. Sadeghi,et al.
Finite-state Markov modeling of fading channels - a survey of principles and applications
,
2008,
IEEE Signal Processing Magazine.
[3]
W. R. Thompson.
ON THE LIKELIHOOD THAT ONE UNKNOWN PROBABILITY EXCEEDS ANOTHER IN VIEW OF THE EVIDENCE OF TWO SAMPLES
,
1933
.
[4]
Hong Shen Wang,et al.
Finite-state Markov channel-a useful model for radio communication channels
,
1995
.
[5]
Amir Ghasemi,et al.
Opportunistic Spectrum Access in Fading Channels Through Collaborative Sensing
,
2007,
J. Commun..
[6]
L. Devroye.
Non-Uniform Random Variate Generation
,
1986
.
[7]
E. O. Elliott.
Estimates of error rates for codes on burst-noise channels
,
1963
.
[8]
Peter Auer,et al.
Finite-time Analysis of the Multiarmed Bandit Problem
,
2002,
Machine Learning.
[9]
Bhaskar Krishnamachari,et al.
On myopic sensing for multi-channel opportunistic access: structure, optimality, and performance
,
2007,
IEEE Transactions on Wireless Communications.
[10]
Mingyan Liu,et al.
Online Learning of Rested and Restless Bandits
,
2011,
IEEE Transactions on Information Theory.