Research Article
Is Bayesian Multi-armed Bandit Algorithm Superior?: Proof-of-Concept for Opportunistic Spectrum Access in Decentralized Networks
@INPROCEEDINGS{10.1007/978-3-319-40352-6_9, author={Sumit Darak and Amor Nafkha and Christophe Moy and Jacques Palicot}, title={Is Bayesian Multi-armed Bandit Algorithm Superior?: Proof-of-Concept for Opportunistic Spectrum Access in Decentralized Networks}, proceedings={Cognitive Radio Oriented Wireless Networks. 11th International Conference, CROWNCOM 2016, Grenoble, France, May 30 - June 1, 2016, Proceedings}, proceedings_a={CROWNCOM}, year={2016}, month={6}, keywords={Cognitive radio Multi-armed bandit algorithm Opportunistic spectrum access Usrp testbed}, doi={10.1007/978-3-319-40352-6_9} }- Sumit Darak
Amor Nafkha
Christophe Moy
Jacques Palicot
Year: 2016
Is Bayesian Multi-armed Bandit Algorithm Superior?: Proof-of-Concept for Opportunistic Spectrum Access in Decentralized Networks
CROWNCOM
Springer
DOI: 10.1007/978-3-319-40352-6_9
Abstract
Poor utilization of an electromagnetic spectrum has led to surge of interest in paradigms such as cognitive radio, unlicensed LTE etc. Such paradigms allow opportunistic spectrum access in the vacant frequency bands of the licensed spectrum. Though various spectrum detectors to check the status of frequency bands (i.e., vacant or occupied) have been studied, the selection of the frequency band from wideband spectrum is a challenging problem especially in the decentralized network. In this paper, a testbed for analyzing the performance of decision making policies (DMPs) for identifying optimum frequency band in the decentralized network is presented. Furthermore, experimental results using real radio signals show that the proposed DMP using Bayesian multi-armed bandit algorithm leads to 7–12 % improvement in an average spectrum utilization over existing DMPs. Added advantages of 6–20 % lower switching cost and 30–46 % fewer collisions make the proposed DMP energy-efficient.
