Page 117 - First special issue on The impact of Artificial Intelligence on communication networks and services

,78 -2851$/ ,&7 'LVFRYHULHV 9RO 0DUFK

BAYESIAN ONLINE LEARNING-BASED SPECTRUM OCCUPANCY PREDICTION
IN COGNITIVE RADIO NETWORKS

$KPHG 0RKDPPHG 0LNDHLO
'HSDUWPHQW RI (OHFWURQLF (QJLQHHULQJ
6KDQJKDL -LDR 7RQJ 8QLYHUVLW\ 6KDQJKDL &KLQD
DKPHGBPLNDHLO #VMWX HGX FQ DKPHGBPLNDHLO#\DKRR FR XN

Abstract – Predicting the near future of primary user (PU) channel state availability (i.e., spectrum occupancy)
is quite important in cognitive radio networks in order to avoid interfering its transmission by a cognitive
spectrum user (i.e., secondary user (SU)). This paper introduces a new simple method for predicting PU channel
state based on energy detection. In this method, we model the PU channel state detection sequence (i.e., “PU
channel idle” and “PU channel occupied”) as a time series represented by two different random variable
distributions. We then introduce Bayesian online learning (BOL) to predict in advance the changes in time series
(i.e., PU channel state.), so that the secondary user can adjust its transmission strategies accordingly.
A simulation result proves the efficiency of the new approach in predicting PU channel state availability

.H\ZRUGV ± %D\HVLDQ RQOLQH OHDUQLQJ FRJQLWLYH UDGLR SULPDU\ XVHU VSHFWUXP RFFXSDQF\ SUHGLFWLRQ

1. INTRODUCTION WHFKQLTXHV FRQVLGHU RQO\ WLPH LQYDULDQW 38 PRGHO
EHKDYLRUV :KLOH LQ UHDO ZRUOG FRJQLWLYH UDGLR
,Q FRJQLWLYH UDGLR QHWZRUNV D VHFRQGDU\ XVHU LV V\VWHPV 38 WUDIILF SDWWHUQV FDQ DOVR H[KLELW
DOORZHG WR RSSRUWXQLVWLFDOO\ XWLOL]H WKH YDFDQW WLPH YDULDQW WUDIILF
VSHFWUXP FKDQQHOV OHIW E\ WKH SULPDU\ XVHU ZLWKRXW DERYH PHQWLRQHG PDFKLQH
LQWHUIHULQJ ZLWK WKHLU WUDQVPLVVLRQ 2QH RI WKH NH\ OHDUQLQJ DOJRULWKPV 2Q WKH RWKHU KDQG WKH %D\HVLDQ
FKDOOHQJHV IRU VHFRQGDU\ XVHUV LQ FRJQLWLYH UDGLR RQOLQH OHDUQLQJ DOJRULWKP %2/ > @ KDV D FDSDELOLW\
QHWZRUNV LV KRZ WR NQRZ ZKHQ WR RFFXS\ RU OHDYH WR WUDFN ERWK WLPH YDULDQW DQG WLPH LQYDULDQW
WKH VSHFWUXP L H WKH FKDQQHOV IRU SULPDU\ XVHUV¶ GXDO VWDWHV VZLWFKLQJ WLPH VHULHV EHKDYLRUV
WUDQVPLVVLRQ 7R WDFNOH WKLV SUREOHP WKH VHFRQGDU\ 0RWLYDWHG E\ WKH IDFW WKDW WKH QDWXUH RI WKH 38V
XVHU PXVW EH FDSDEOH RI SUHGLFWLQJ LQ DGYDQFH WKH FKDQQHO VWDWH DYDLODELOLW\ FDQ EH DOVR PRGHOHG DV
FKDQQHO DYDLODELOLW\ RI WKH SULPDU\ XVHU L H ZKHWKHU GXDO VWDWHV VZLWFKLQJ WLPH VHULHV ZH SURSRVH D QHZ
WKH 38 FKDQQHOV¶ VWDWXV DUH ³LGOH´ RU ³EXV\´ VR WKDW VSHFWUXP RFFXSDQF\ 38V FKDQQHO VWDWH SUHGLFWLRQ
LW FDQ RFFXS\ RU OHDYH WKH FKDQQHOV IRU 38 WHFKQLTXH WKDW XWLOL]HV %2/ WR SHUIRUP 38 FKDQQHO
WUDQVPLVVLRQ DYDLODELOLW\ SUHGLFWLRQ LQ FRJQLWLYH UDGLR QHWZRUN ,Q
PRUH GHWDLOV ZH FDSWXUHG WKH 38 FKDQQHO VWDWH
7KH VSHFWUXP RFFXSDQF\ SUHGLFWLRQ SUREOHP KDV HQHUJ\ GHWHFWLRQ VHTXHQFH XVLQJ D WLPH VHULHV WKDW
EHHQ ZLGHO\ LQYHVWLJDWHG IRU H[DPSOH WKH LGHD RI VZLWFKHV RYHU WKH WLPH EHWZHHQ WZR GLIIHUHQW UDQGRP
SUHGLFWLYH VSHFWUXP DFFHVV ZDV ILUVW LQWURGXFHG LQ GLVWULEXWLRQV UHSUHVHQWLQJ WKH 38 FKDQQHO VWDWH
> @ LQ ZKLFK WKH DXWKRUV XWLOL]H +LGGHQ 0DUNRY L H 38 LGOH RU 38 RFFXSLHG :H WKHQ IHG WKLV WLPH
0RGHO +00 WR VROYH WKH VSHFWUXP RFFXSDQF\ VHULHV DV DQ REVHUYDWLRQV VHTXHQFH LQWR D %2/
SUHGLFWLRQ SUREOHP /DWHU RQ WKH +00 EDVHG SUHGLFWLRQ DOJRULWKP WR HVWLPDWH RU SUHGLFW LQ
VSHFWUXP SUHGLFWLRQ PRGHO UHFHLYHG JUHDW DWWHQWLRQ DGYDQFH WKH SRLQW RI WKH WLPH ZKHQ WKH FKDQJH ZLOO
LQ WKH OLWHUDWXUH > @ $QG GXH WR WKH IDFW WKDW RFFXU EHWZHHQ WKH WZR VWDWHV RI WKH WLPH VHULHV VR
+00 EDVHG DSSURDFKHV UHTXLUH D SULRUL NQRZOHGJH WKDW 68V FDQ DGMXVW WKHLU WUDQVPLVVLRQ VWUDWHJLHV
RI WKH 38V¶ WUDIILF SDWWHUQ RWKHU PDFKLQH OHDUQLQJ DFFRUGLQJO\ 7KH H[SHULPHQWDO UHVXOWV VKRZ WKH
DSSURDFKHV VXFK DV QHXUDO QHWZRUN > @ %D\HVLDQ HIIHFWLYHQHVV RI WKH %2/ DOJRULWKP LQ SUHGLFWLQJ WKH
LQIHUHQFH > @ DQG RQOLQH VXSSRUW YHFWRU UHJUHVVLRQ FKDQJLQJ SRLQWV RI WKH WLPH VHULHV WKDW ZHUH
695 > @ KDYH EHHQ DGRSWHG IRU WKH SUHGLFWLRQ RI JHQHUDWHG WR FDSWXUH 38 FKDQQHO DYDLODELOLW\
38 FKDQQHO DYDLODELOLW\ +RZHYHU WKHVH SUHGLFWLRQ

‹ ,QWHUQDWLRQDO 7HOHFRPPXQLFDWLRQ 8QLRQ