Final Repo Page i
EVEREST IST-2002-001858
D23
Final Repo
Con ac ual Da e o Deli e y o he CEC: 31-12-2005
Ac ual Da e o Deli e y o he CEC: 15-01-2006
Au ho (s): Fe nando Casade all (UPC)
Pa icipan (s): UPC, KCL, PTIN, TI, TID, TEL
Wo kpackage: WP1
Es . pe son mon hs: 0.5
Secu i y: PU
Na u e: Repo
Ve sion: 0.1
To al numbe o pages: 49
Abs ac :
This deli e able cons i u es he inal epo o he p ojec IST-2002-001858 EVEREST.
A e i s success ul comple ion, he p ojec p esen s his documen ha i s ly
summa izes he con ex , goal and he app oach objec i e o he p ojec . Then i p esen s
a concise summa y o he majo goals and esul s, as well as highligh s he mos
aluable lessons de i ed o m he p ojec wo k. A lis o deli e ables and publica ions is
included in he annex.
Fo mo e de ailed echnical esul s please conside he public deli e ables, a ailable a
h p://www.e e es -is .upc.es
Keywo d lis : Co-ope a ion wi h o he P ojec s, Conce a ion Ac i i ies
Final Repo Page ii
DISCLAIMER
The wo k associa ed wi h his epo has been ca ied ou in acco dance wi h he highes
echnical s anda ds and he EVEREST pa ne s ha e endea ou ed o achie e he deg ee
o accu acy and eliabili y app op ia e o he wo k in ques ion. Howe e since he
pa ne s ha e no con ol o e he use o which he in o ma ion con ained wi hin he
epo is o be pu by any o he pa y, any o he such pa y shall be deemed o sa is ied
i sel as o he sui abili y and eliabili y o he in o ma ion in ela ion o any pa icula
use, pu pose o applica ion.
Unde no ci cums ances will any o he pa ne s, hei se an s, employees o agen s
accep any liabili y wha soe e a ising ou o any e o o inaccu acy con ained in his
epo (o any u he consolida ion, summa y, publica ion o dissemina ion o he
in o ma ion con ained wi hin his epo ) and/o he connec ed wo k and disclaim all
liabili y o any loss, damage, expenses, claims o in ingemen o hi d pa y igh s.
Final Repo Page iii
DOCUMENT HISTORY
Da e Ve sion S a us Commen s
9-01-2006 1.0 In
Final e sion o commen s and app o al by
he PCC membe s
12-01-2006 1.0 Ap Submission o he E.U.
Final Repo Page i
Table o Con en s
1. O e iew o Gene al P ojec Objec i es ............................................................................ 1
2.- Con ac o In ol ed ....................................................................................................... 4
3.- Wo k Pe o med ............................................................................................................ 5
4.- End Resul s ..................................................................................................................... 8
4.1 New E2E QoS a chi ec u e .......................................................................................... 8
4.2 E alua ion o RRM/CRRM algo i hms...................................................................... 11
4.3 EVEREST Tes bed ..................................................................................................... 36
5 Main conclusions eached ................................................................................................ 42
Annex 1.- Lis o Publica ions ................................................................................................. 45
Annex 2.- Rela ion wi h he S anda ds .................................................................................... 49
Final Repo Page
EXECUTIVE SUMMARY
This epo summa ises he main achie emen s o he EVEREST P ojec , an IST esea ch and
echnological de elopmen p ojec ca ied ou be ween Janua y 2004 and Decembe 2005 by
Uni e si a Poli écnica de Ca alunya (UPC); King’s College London (KCL); Po ugal Telecom Ino açao
(PTIN); Telecom I alia Lab (TILAB); Tele ónica In es igación y Desa ollo (TID), TeliaSone a (TEL).
The mos impo an echnical achie emen s o he p ojec co e many di e en aspec s ela ed o
Radio Resou ce and QoS Managemen and Common Radio Resou ce Managemen (CRRM).
Di e en algo i hms ela ed o Admission Con ol, Conges ion Con ol as well as on Packe Scheduling
p ocedu es ha e been p oposed and e alua ed o he en isaged Radio Access Technologies, namely
GERAN, UMTS, WLAN. Mo eo e issues ela ed o he end- o-end QoS a chi ec u e ha e been also
s udied and e alua ed.
The Pe o mance e alua ion o he p oposed a chi ec u e and RRM/CRRM echniques was comple ed
by means o a se o labo a o y es s ca ied ou using a Demons a o de eloped in he p ojec . This
Demons a o is a SW/HW lexible ool, which p o ides a ealis ic eal ime emula ion o an e ol ed
B3G adio access sys em able o manage mul imedia IP based applica ions. Finally, signi ican
dissemina ion policy, based on publica ions on high quali y magazines and con e ences, was ca ied
ou . Mo eo e , se e al s anda ds con ibu ions we e also gene a ed and p esen ed o he pe inen
3GPP echnical commi ees.

Final Repo Page 1
1. O e iew o Gene al P ojec Objec i es
The p o ision o beyond 3G (B3G) he e ogeneous ne wo k opologies is concep ually a e y a ac i e
no ion; howe e , i is a challenge o accomplish an e icien ne wo k design. In his con ex , Radio
Resou ce Managemen (RRM) s a egies a e esponsible o an u mos e icien u ilisa ion o he ai
in e ace esou ce in he a ailable Radio Access Ne wo ks (RANs). Any s and-alone wi eless sys ems o
he e ogeneous hyb ids he eo , ely on RRM s a egies o gua an ee a ce ain p io ag eed QoS.
The objec i e o EVEREST is o de ise and assess a se o speci ic s a egies and algo i hms o
access and co e ne wo k, leading o an op imal u ilisa ion o sca cely a ailable esou ces o he
suppo o mixed se ices wi h end- o-end QoS mechanisms wi hin he e ogeneous ne wo ks B3G.
In o de o achie e he o me main objec i e, he ollowing pa ial objec i es will be add essed in he
p ojec :
 To iden i y, p opose, simula e, assess and alida e ad anced RRM algo i hms o GERAN and
UMTS as well as no el adio concep s as HSDPA and WLAN ha ep esen s he i s e olu iona y
s eps o cu en 3G sys ems owa d beyond 3G.
 To de elop Common RRM (CRRM) algo i hms o he e ogeneous ne wo ks ocused on UTRAN and
GERAN. Bo h igh and e y igh coupling will be conside ed.
 To conside o he echnologies ha can be complemen a y o GPRS/UMTS, such as:
 WLAN o indoo ho spo s
 Di e en ypes o epea e s, ac ing as co e age ex ensions
 To in es iga e issues ela ed o he in e -wo king o he co e ne wo k (CN) Bandwid h B oke (BB)
and he UMTS, GERAN and WLAN RRM en i ies (e.g., o how o connec he ele an elemen s in he
wo en i ies; wha ype o signalling o use; wha mechanism and how o en o upda e he in o ma ion
held in he BB; how o ansla e he pa ame e s o RRM in o he BB pa ame e s; e c.).
 To demons a e he bene i s o he p oposed RRM and CRRM algo i hms by means o mul imedia IP-
based applica ions o e a eal- ime es bed. In o de o ela e consis en ly ne wo k pe o mance wi h
ad anced applica ion usage, some examples o highly demanding se ices will be conside ed and
ela ed h ough he use o QoS managemen en i ies.
EVEREST is p o iding angible con ibu ions owa ds a he e ogeneous ealisa ion o 2G/2.5/3G (e.g.,
GERAN and UTRAN) and 3.5G ne wo ks (e.g., HSDPA) wi h he inclusion o newly eme ging RANs (e.g.,
WLAN o ho spo co e age) o he 2007-2010 ime- ame. P oposed solu ions will be complian wi h
and aligned o s anda disa ion ac i i ies in he ield e.g. 3GPP, IETF, IEEE.
In o de o accomplish he abo e objec i es, he p ojec e ol es a ound o main ac i i ies: (1) algo i hmic
de elopmen and simula ion by means o ad anced simula ion ools, and (2) demons a ion o he
echnology by means o implemen ing eal- ime es beds o p o e concep s.
Resea ch Challenges
Compa ably li le wo k has been de o ed o da e in p o iding solid and publicly a ailable RRM s a egies.
This is due o he ac ha RRM algo i hms a e no subjec o s anda disa ion, leading o an inc ease in
compe i i eness among he manu ac u e s. The aim o he p ojec is o p o ide solid ad ances in his
ield, allowing o a uly op imised he e ogeneous ne wo k deploymen . The esea ch challenges, o be
ackled by EVEREST, can hus be summa ised as ollows:
• The RRM s a egies o legacy ne wo ks (i.e. ea ly eleases o GERAN and UTRAN al eady being
deployed by ope a o s) a e in es iga ed. The GSM/GPRS pa is o a he low dimensionali y, i.e. only
a ew pa ame e s a e needed o une hei op imali y. Joining hese ne wo ks in o a he e ogeneous
ne wo k, le e ages he RRM s a egies in o op imisa ion p oblems wi h many mo e deg ees o
eedom.
Final Repo Page 2
• In he case o UMTS and UTRAN (UMTS Te es ial RAN), a huge amoun o human and ma e ial
esou ces has been de o ed o da e ha led o a as ple ho a o 3GPP speci ica ions, i s
manu ac u e p oduc s and ope a o oll-ou s. The ocus o da e has mainly been in adio ne wo k
planning, since he i s s ep was an ini ial ne wo k oll-ou . Some e y basic RRM s a egies a e
being in oduced on he p esump ion ha he a ic would be low. In he coming yea s, when mo e
and mo e adio enginee s become amilia wi h 3G and beyond echnologies, i would be necessa y
inc ease and ha monise he gene al knowledge on W-CDMA (i.e. UTRAN-FDD) RRM s a egies.
Only his will acili a e he u u e success o 3G and beyond ne wo ks wi h an e e -inc easing
sophis ica ion in RRM s a egies. Clea ly, op imised RRM algo i hms a e he only key o success ully
handle highly loaded ne wo ks, which a e being en isaged o domina e he wi eless a ena in i e
yea s om now. Addi ionally, eme ging concep s, such as RAN-sha ing wi hin UMTS will lead o
mixed domains wi h equi alen public land mobile ne wo ks (ePLMN) solu ions, complica ing RRM
s a egies o speci ic se ices.
• WLANs a e also expec ed o play an impo an ole in he p o ision o high da a a e se ices. He e,
wi eless indoo co e age could be be e p o ided by WLAN as high powe esou ces would ha e o
be al e na i ely spen in UMTS. Again, a new dimension in he RRM p oblem is in oduced, he
in luence o which needs o be assessed.
• Fo each echnology, he e will be a need o es ablish a mapping be ween equi alen bea e s. This
mapping will ease he de ini ion o he in e -wo king om he se ice poin o iew.
• Fo each echnology, he de elopmen o CRRM algo i hms wi hin he RAN is i al o a p ope
unc ioning o he he e ogeneous ne wo k. De eloped, assessed and compa ed ha e o be CRRM
algo i hms o igh and e y igh coupling.
• Fu he , o suppo end- o-end QoS in a he e ogeneous wi eless (and wi ed) mobile en i onmen , he
in e ac ion be ween he QoS managemen en i ies o he CN and he indi idual RRM o each RAN o
he CRRM o a ple ho a o RANs, in an adminis a i e domain is o p ime impo ance. The CN o
UMTS e ol es owa ds an All-IP a chi ec u e wi h QoS mechanisms and p o ocols sugges ed by he
IETF. Mos o hese mechanisms and p o ocols will be based on he BB concep . The BB is he main
a chi ec u e elemen o he con ol plane o he Di se model p oposed by IETF o suppo ing end-
o-end QoS in IP-based ne wo ks. The BB is a logical en i y esponsible o esou ce alloca ion in an
adminis a i e domain and co-o dina es in e -domain and in a-domain esou ce alloca ion. Fo he
in e -domain esou ce alloca ion, he BB o he CN should be awa e o he esou ce equi emen s and
esou ce a ailabili ies o he pee CNs. Fo he in a-domain esou ce alloca ion, he BB o he CN
should be awa e o he esou ce a ailabili ies o i s unde lying RANs.
The a ional esea ch app oach endea ou s o encompass he Radio Access Ne wo k (e.g. RRM o
single RAT, CRRM algo i hms), he Co e Ne wo k (e.g. BB concep s o he en isaged a chi ec u es,
In a-CN-domain Di se Signalling P o ocols, e c.) and a coupled a chi ec u e be ween bo h (e.g.,
compa a i e s udy on a chi ec u e and signalling issues; compa ison be ween a ious CRRM + BB
algo i hms depending on loose, igh o e y igh coupling).
Rele ance o he objec i es
EVEREST is amed wi hin he “Mobile and Wi eless Sys ems B3G” s a egic objec i e (SO) o he IST
p io i y. As s a ed in he objec i es o his IST p io i y, “ea ly p epa a o y wo k has cha ac e ised sys ems
B3G as a ho izon al communica ion model, whe e di e en e es ial access le els and echnologies a e
combined o complemen each o he in an op imum way o di e en se ice equi emen s and adio
en i onmen s”.
In his espec , he EVEREST esea ch scope is ully aligned wi h wha is known as sys ems B3G adio
en i onmen s, since a di e si y o echnologies a e conside ed: GERAN, UMTS and WLAN. EVEREST
u he s esses ha ea ly p epa a o y wo k has conside ed hese combined and complemen a y
scena ios o a e y limi ed ex en . Mos o hese echnologies ha e hus no been su icien ly analysed o
yield hei ull capabili ies, e.g. QoS in GPRS; o , he s udies ha e mainly ocused on he i s s age
deploymen capabili ies, e.g. UMTS phase 1 capabili ies and hei limi ed associa ed ea u es in e ms o
Final Repo Page 3
RRM algo i hms, implemen ed RABs, e c.; o , hey ha e solely ocused on he a ailable p oduc
capabili ies, e.g. IEEE 802.11b wi hou QoS o WLAN.
Consequen ly, he goals o EVEREST include:
 Fu he p og ess on he de ini ion o ad anced RRM mechanisms (e.g. including o example
loca ion aided RRM algo i hms) leading o an op imised usage o he di e en echnologies
acco ding o a echnology oadmap d i en by he e olu ion o he wi eless scena io;
 Acknowledge and con ibu e o a CRRM, whe e a pool o esou ces belonging o di e en
echnologies a e commonly conside ed and commonly op imised.
 Use and es he CRRM, o p o iding end- o-end QoS in an IP mobile access ne wo k. De ine
he in e ac ions be ween a BB and he adio en i ies, in o de o p o ide he adap ed QoS o he
se ice and o use in an op imal way he he e ogenei y o he IP access ne wo k.
This p ojec is clea ly ne wo k ope a o -d i en. They ha e iden i ied he scena ios o in e es o be
conside ed a di e en ime scales and o di e en ne wo k oll-ou phases. The close in ol emen o
Mobile Ope a o s gua an ees ha ealis ic ne wo k da a (e.g. a ic dis ibu ions, p opaga ion losses, da a
aces, e c.) can be p ocessed and u ilised o he bene i o op imised RRM algo i hms o he e ogeneous
ne wo ks. F om he ope a o side, i is well unde s ood ha sui able app oaches o RRM all well beyond
a me e ne wo k deploymen ; hey cons i u e an inno a i e esea ch ield wi h clea indica ions on how o
manage adio esou ces in he e ogeneous ne wo ks accommoda ing adi ional and no el se ices. In
his espec , he ope a o -o ien ed app oach adop ed in EVEREST ensu es ha he analysed scena ios
a e ma ke - ele an and use -cen ic.
Mo eo e , esul s coming om he p ojec will p o ide a manu ac u e -independen and complemen a y
analysis o he RRM s a egies. This will allow he mobile ope a o s o e alua e and compa e solu ions
coming om he ma ke wi h an a ailable e e ence o he sys em pe o mance. I is wo h no ing ha he
open na u e o he algo i hms de eloped wi hin he p ojec and i s a ailabili y o he en i e wi eless
communi y is expec ed o con ibu e o a be e ansi ion om he di e en e olu iona y scena ios
conside ed. Ou pu s om EVEREST will cons i u e a aluable e e ence o ope a o s, manu ac u e s and
academia, acili a ing u he p og ess in his ield o many yea s o come.
Final Repo Page 4
2.- Con ac o In ol ed
Pa icipan Name Sho Name Coun y URL Add ess
Uni e si a Poli ecnica de
Ca alunya UPC Spain h p://www.upc.edu
King's College London KCL Uni ed Kingdom h p://www.kcl.ac.uk
Po ugal Telecom Ino açao PTIN Po ugal h p://www.p ino acao.p
Telecom. I alia TI I aly h p://www. elecomi alialab.com
TELEFONICA I+D TID Spain h p://www. id.es
TELIASONERA TEL Sweden h p://www. eliasone a.se
Final Repo Page 11
In summa y, a new end– o-end (E2E) QoS a chi ec u e, including message sequence cha o he End-
o-End QoS signalling, has been p oposed. The p oposal ollows a policy-based app oach. A new
concep , Wi eless QoS B oke (WQB), has been in oduced o managing he esou ces o he adio
access pa wi hin he amewo k o B3G ne wo ks. The a chi ec u e p esen ed encompasses CRRM
unc ionali y suppo in he adio pa and p oposes a QoS con ol plane o join ly manage IP esou ces in
he co e ne wo k and adio bea e s in he RANs. The p oposed a chi ec u e was also de eloped in e ms
o p ocedu es and signalling and QoS class mapping be ween UMTS and Di Se classes.
4.2 E alua ion o RRM/CRRM algo i hms
RRM is a complex p oblem wi h many ac o s in luencing in he achie ed pe o mance and wi h many
mixing e ec s. Fu he mo e, he RRM p oblem has mul iple dimensions and mul iple unc ionali ies ha ,
ei he in a mo e di ec o indi ec way, impac on he ai in e ace. Then, in he case o EVEREST, he
RRM mechanisms conside ed a e: GERAN, UTRAN and WLAN
RRM issues o UMTS
WCDMA access ne wo ks, such as he conside ed in UTRA-FDD p oposal, p o ide an inhe en lexibili y
o handle he p o ision o u u e 3G mobile mul imedia se ices. UMTS will o e an op imiza ion o
capaci y in he ai in e ace by means o e icien Radio Resou ce Managemen (RRM) algo i hms. RRM
is a complex p oblem wi h many ac o s in luencing in he achie ed pe o mance and wi h many mixing
e ec s. Then, a c ucial aspec is o iden i y ele an issues and undamen al elemen s in luencing on he
o e all RRM p ocess, hen achie ing a wide-scope and open-minded pe spec i e. In his con ex , he
subsec ions de ailed below desc ibe a a ie y o s udies ela ed o RRM o UMTS.
I- A new amewo k o cap u ing coupling among cells
An inno a i e ma hema ical amewo k cap u ing he ai in e ace coupling among he di e en cells in
he scena io has been de eloped based on he de i a i es o he cell uplink load ac o and he
downlink ansmi ed powe [1][2]. This amewo k is p esen ed in a compac o mula ion o bo h uplink
and downlink, and i allows implemen ing mechanisms suppo ing sma load con ol ac ions including
admission con ol, conges ion con ol algo i hms. Fo ins ance, igu es 6 and 7 show how he algo i hm
based on de i a i es achie es he load educ ion wi h he lowes NRT h oughpu educ ion. The
di e ence wi h espec o an algo i hm ha selec s in a andom way (“alea ” algo i hm) he NRT use
whose h oughpu should be dec eased is e y high. On he con a y, bo h algo i hms based on
de i a i es o e en in in e ac i e load pe o m much be e .
0
10
20
30
40
50
60
70
80
90
100
050 100 150 200 250 300 350
Use s
UL NRT Th Reduc ion (%)
De i
Alea
In e ac
Figu e 6.- UL NRT Th oughpu educ ion in he neighbou ing cells
Figu e 7.- DL NRT Th oughpu educ ion in he neighbou ing cells
II- Indoo T a ic
The implica ions o indoo a ic in 3G W-CDMA based sys ems may signi ican ly di e om 2G TDMA-
based solu ions because ansmi ed powe le els a e he key adio esou ces in W-CDMA. The highe
powe le els needed o indoo se ice will lowe cell capaci y bo h o uplink and downlink a ic. Table
1 p esen s he capaci y loss o bo h uplink and downlink as he ac ion o indoo a ic inc eases.
Capaci y is de ined as he maximum numbe o use s in he scena io ha gua an ees a BLER ≤ 2%. I
can be obse ed ha he deg ada ion due o indoo use s is much mo e signi ican in he uplink han in
he downlink di ec ion.
0
20
40
60
80
100
120
0
50
100
150
200
250
300
350
Use s
DL NRT Th Reduc ion (%)
De i
Alea
In e ac

Final Repo Page 12
Table 1 .- Capaci y loss (%) ela i e o he
case wi h no indoo a ic (p=0) o 64/64
kb/s and 64/384 kb/s adio bea e
UL
64 Kb/s
DL
64 Kb/s
DL
384 Kb/s
P=0.1
19.2%
9.3%
12.5%
P=0.2
37.7%
11.6%
18.8%
P=0.5
88.4%
15.3%
25.0%
III- T a ic Ho -Spo s
In a eal mobile ne wo k, he e a e ce ain geog aphical a eas wi h high densi y o use s. In hese
a eas, a high demand o adio esou ces may appea . In o de o assu e he use QoS (Quali y o
Se ice) equi emen s in a ho spo , no only ne wo k planning bu also RRM algo i hms mus be
conside ed.
Figu e 8 shows he impac o a ho spo (HS) loca ion on he ansmission powe o wo base s a ions,
named 5 and 10 espec i ely. As shown, when he ho spo is loca ed a om Base S a ion 5 (HS a
200m), he powe inc ease is highe han when he ho spo is nea e (HS a 50m). This is because
ho spo use s loca ed a om he base s a ion demand highe le el o powe o gua an ee he (Eb/No)
a ge . So, i he ho spo is loca ed ela i ely a om i s base s a ion, i can happen ha he Base
S a ion may no ha e enough powe o sa is y all use equi emen s, causing bad signal quali y. A
connec ion is d opped i he cu en (Eb/No) is 1dB below he a ge alue du ing he 90% o ime in
1second.
Figu e 8.- E ec o s a ic ho spo loca ion on BS powe
Figu e 9.- Impac o ho spo mo emen on d opping p obabili y
I is wo h no ing ha i he ho spo use s we e ini ially loca ed a 0me es om BS5, he o e load
si ua ion in BS5 would las longe because he ho spo use s connec ed o BS5 would need much ime
o each he co e age a ea o BS10 and make hando e s. When ho spo use s a i e nea he cell
edge, hey will begin o hando e o BS 10, and hen, he BS 5 ansmission powe will be educed.
Ob iously, his ac will happen i s i he ho spo is ini ially loca ed a 200m om BS. Mo eo e , when
he ho spo is ini ially loca ed nea BS5 he o e load si ua ion and he d oppings occu la e , and hen
he a ailable ime o p e en his o e load is highe such as i is shown in igu e 9.
In he uplink case, he load ac o o di e en base s a ions and he sys em pe o mance in e ms o
d opping p obabili y ha e been analysed and simila esul s o hose ob ained o he downlink a e
eached.
Ano he echnique ha is commonly used is o adjus he ansmission pilo powe o he ho spo cells
and i s adjacen cells [3],[4]. Figu e 10 compa es he ansmission powe o BS5 and BS10 wi h he
p oposed pilo adjus men algo i hm (wi h load balancing) and wi hou he p oposed pilo adjus men
algo i hm. The pilo powe is changed once in a second (Tp=1second) and α=1. The pa ame e α
de e mines he sensi i i y in he changes o he pilo powe . As shown, when no load balancing is
conside ed, an o e load o BS5 and BS10 can be obse ed as use s mo e om BS5 o BS10. Wi h he
load balancing echnique, he ansmission powe o BS5 and BS10 is equalised (see igu e 10)
adjus ing he pilo in an adequa e way as shown in igu e 11, whe e he a e age alue o he pilo powe
o di e en BS is plo ed. The load balancing echnique educes he powe limi a ion p obabili y o he
di e en base s a ions, and his will educe he d opping p obabili y as i is shown in Table 2.
Final Repo Page 13
Figu e 10.- Impac o load balancing in BS ansmission powe
Figu e 11 .- E olu ion o he pilo powe o di e en BS when he
p oposed pilo adjus men algo i hm is applied
On he o he hand Table 3 also shows he d opping p obabili y and he a e age ime be ween
hando e s pe use , o di e en alues o α and when no load balancing is ca ied ou . Once again, he
p oposed algo i hm educes he d opping p obabili y wi h espec o he non-load balancing case, bu i
inc eases he a e o hando e s causing an inc ease in ne wo k signalling. Mo eo e , he mos
app op ia e alue o α is be ween 0.1 and 0.25.
Table 2.- Impac o Tp on d opping p obabili y and
hando e s a is ics Table 3.- Impac o α in d opping and hando e s a is ics.
No
Balance
Tp =1 Tp =10
No
Balance
α =0.1 α =0.25 α =1
D opping p ob.
2.38
0.51
0.73
D opping p ob.
2.38
0.16
0.14
0.51
A g. Time be ween
HO pe use (sec.)
42.93 12.24 18.81
A g. Time be ween
HO pe use (sec.)
42.93 34.71 23.47 12.24
IV- S a ic T a ic
The lexibili y in he p o ision o mul iple bi a e se ices in 3G communica ion sys ems will allow use s
o bene i om se ices be e adjus ed o hei speci ic equi emen s. Highe bi a es, QoS (Quali y o
Se ice) will also be c ucial o 3G success om he use poin o iew. In his con ex , i becomes p ime
impo an o iden i y he key elemen s cha ac e ising he di e en se ices and an icipa e he equi ed
mechanisms o suppo hese se ices h ough he ai in e ace in a sui able and op imised manne . In
pa icula , use s ha ecei e da a a ic se ices, ypically wi h lap ops in scena ios like o ices, ai po s,
e c., use o be s a ic o , a leas , wi h a e y limi ed mobili y. This ac gi es oom o p opose mo e
sophis ica ed admission con ol s a egies, which may p o ide signi ican pe o mance imp o emen s.
Unde his amewo k, an algo i hm, deno ed as PLEBAC (Pa h Loss Es ima ion Based Admission
Con ol) has been p oposed aking in o accoun he bene i s om he easie p edic abili y in e ms o
powe consump ion o s a ic da a use s. I makes use o he measu emen epo s p o ided by he
e minal du ing he call se -up p ocess in o de o ha e a mo e accu a e es ima ion o he equi ed
powe along he connec ion ime. The p oposed algo i hm has been e alua ed unde di e en
condi ions o se ice bi a e and cell adii and compa ed agains a e e ence algo i hm and compa ed
agains an algo i hm ha simply conside s he a e age powe consump ion, deno ed as PABAC (Powe
A e aged Based Admission Con ol).
In o de o assess he po en ial o he p oposed PLEBAC algo i hm, a se o sys em le el simula ions
ha e been ca ied ou . Di e en scena ios ha e been conside ed wi h di e en cell adii, anging om
500 m up o 2 Km. Figu e 12 p esen s he ansmi ed powe and he powe inc ease es ima ion o bo h
algo i hms as a unc ion o he pa h loss. No ice ha PLEBAC adjus s mo e accu a ely he powe
inc ease es ima ion o he eal ansmi ed powe , which is highe when he pa h loss inc eases.
Final Repo Page 14
22
24
26
28
30
32
34
36
38
40
105 115 125 135 145
Pa h Loss PL (dB)
Powe (dBm)
T ansmi ed Powe
∆PT PLEBAC
∆PT PABAC
Figu e 12 .- Ac ual ansmi ed powe and powe inc ease as a
unc ion o he pa h loss
0
5
10
15
20
25
30
35
40
500 1000 1500 2000
Cell Radius (m)
Gain (%)
384 kb/s
256 Kb/s
Figu e 13 .- Th oughpu gain as a unc ion o he cell adius
Figu e 13 illus a es he h oughpu gain o he PLEBAC admission con ol wi h espec o PABAC o
di e en cell adii, when he o e ed bi a e is 256 kb/s and 384 kb/s. No ice ha he gain is bigge o
la ge cell adii and highe se ice bi a es. In hese cases, he highe powe demand o use s wi h
la ge pa h loss leads o pe o mance deg ada ion no only o hese use s bu also o o he use s ha
a e loca ed close o he base s a ion. As a esul , an algo i hm like PLEBAC, which akes in o accoun
use ’s pa h loss, imp o es he o e all pe o mance while a he same ime i gua an ees he quali y o
he accep ed use s. Fu he mo e, he highe he se ice bi a e, he highe he con ibu ion o a use o
he o al sys em h oughpu , and consequen ly, a bad admission o a bad ejec ion u ns in o la ge
h oughpu educ ions. As a esul , i can be concluded ha he PLEBAC s a egy is be e adap ed o
use ’s dis ibu ion in he ne wo k.
V- Repea e s
The epea e s a e equipmen s be ween he adio base s a ion and he e minals, able o ampli y he
ecei ed signals bo h on he uplink and on he downlink. Typically, a epea e is o med by wo
an ennas: one an enna (called dono an enna) is s ongly di ec i e and is di ec ed o he dono base
s a ion; he o he an enna (called co e age o se ice an enna) has he pu pose o co e he se ice
a ea. The possible usage si ua ions o epea e s in a WCDMA sys em a e he ollowing:
• co e age ex ension: in o de o co e he so called “dead spo ” (a eas no co e ed du ing he i s
deploymen o he ne wo k);
• capaci y ex ension: in o de o inc ease he capaci y o a base s a ion wi h an inc eased a ic
load;
• so -hando e egion educ ion: hanks o he epea e s i is possible o educe he so -hando e
a eas o al eady-connec ed use s.
When a epea e is in oduced in a WCDMA sys em, he i s e ec ha i is possible o obse e is he
inc ease o he noise igu e o he base s a ion. Conside ing he co e age a ea, wi h he inc ease o he
noise igu e o he base s a ion we ha e he ecei e sensi i i y making wo se, wi h he consequence
ha he base s a ion co e age dec eases, al hough he o al cell adius is inc eased due o he
in oduc ion o he epea e .
In e ms o capaci y, he in oduc ion o he epea e s leads o a modi ica ion o he in e e ence
cha ac e is ics o he scena io. Figu e 14 shows he pe cen age o se ed use s when a ying he
epea e gain. Two load si ua ions: 50 use s pe cell and 20 use s pe ho -spo ; 60 use s pe cell and 25
use s pe ho -spo ha e been conside ed and h ee di e en condi ions ha e been analyzed:
• Neglec ing he e ec s o he epea e s on he noise igu e and he blocking due o codes in
downlink (case III)
• Conside ing he e ec s o he epea e s on he noise igu e and neglec ing he blocking due o
codes in downlink (case II)
• Conside ing bo h he e ec s o he epea e s on he noise igu e and he blocking due o codes in
downlink (case I).
Final Repo Page 15
88%
89%
90%
91%
92%
93%
94%
95%
96%
020 40 60 80 100
G_ ep [dB]
50 (20) I 50 (20) II
60 (25) I 60 (25) II
50 (20) III 60 (25) III
Figu e14 – Pe cen age o se ed use s (I=noise igu e modi ied,
blocking on codes in downlink; II=noise igu e modi ied,
no blocking on codes; III=no noise igu e modi ied, no
blocking on codes).
Figu e 15 – Bes se e a ea a ia ion in cells wi h and wi hou
epea e s
The pe cen age o se ed use s in he sys em inc eases wi h epea e gain alues lowe han 70-80 dB,
when he maximum capaci y alue is ob ained. Ce ainly, inc easing he epea e gain, he in luence o
he epea e on he su ounding a ea inc eases (see igu e 15) and he e o e he bes se e a ea o i s
dono base s a ion inc eases also. Thus, in he ho -spo a ea he ou -o -se ice use s dec ease and, a
he same ime, he numbe o use s ansmi ing o he dono cell h ough he epea e inc eases.
VI- Mul iple RF Ca ie s
The scena io simula ed consis in an indoo a ic Ho Spo wi hin an u ban a ea The UMTS ope a o
uses wo equencies, whe e equency 2 only applies a a h ee sec o si e and equency 1 is applied
a all he mac o laye o cells. The s udy conside s a mix o CS (speech) and PS (HTTP) a ic. O he
o e ed load in By es 25% is speech and 75% is HTTP. Di e en me hods a ec ing on blocking a io,
d opping a io and bi a e ha e been s udied. The me hods conside ed a e unsymme ical alloca ion o
load o equency 1 and 2, edi ec ion o blocked se ice eques s on equency 1 o equency 2 (o
ice e sa), alloca ion o all speech o one equency and u ning o so hando e on equency 2. As
a e e ence me hod 50% o he speech eques s a e alloca ed o equency 1 and he es o he
speech eques s a e alloca ed o equency 2. The same alloca ion p obabili y is applied o HTTP.
Di e en load alloca ion
0
0,5
1
1,5
2
2,5
3
3,5
4
50 51 52 53 54 56
Load on 2 [%]
Resul ela i ely he
e e ence me hod
Speech block
Speech d op
PS block
PS bi a e
Figu e 16. - 50 o 56 % o he load, speech and HTTP, is
alloca ed o equency 2.
Redi ec
0
0,2
0,4
0,6
0,8
1
1,2
CS PS PS a e 10s
Resul ela i ely he e e ence
me hod
Speech block
Speech d op
PS block
PS bi a e
Figu e 17.- Resul ob ain when edi ec ing se ice eques s
blocked on equency 1 o 2, and ice e sa.
Figu e 16 below shows he impac o alloca ing mo e and mo e speech and HTTP load o equency 2.
We see ha i is he PS blocking ha is a ec ed mos ly as he load inc eases. The PS blocking is as
smalles when 52% o he load is alloca ed o equency 2, bu i s a s o inc ease i 2 is loaded wi h
mo e load. The speech blocking and d opping is ai ly insensi i e o shi ing mo e and mo e load om
equency 1 o 2 (a leas in he load ange es ed he e). Hence, he admission con ol me hod p o ec s
he speech use s om de e io a ing QoS when load inc eases, as i should. The PS bi a e does no
change much ei he as 2 ca ies mo e load. The esul s indica e ha PS blocking is sensi i e o he
load alloca ion on he equencies, whe eas speech blocking and d opping, and PS bi a e, is less
sensi i e o di e en load dis ibu ions. HTTP blocking is imp o ed and he speech QoS is wo sening
when all speech is alloca ed o equency 1. Speech QoS is imp o ed hen all speech is alloca ed o
equency 2 (assuming ha in e - equency hando e does no cause d opped calls). I so hando e is
Final Repo Page 16
uned o on equency 2, he PS bi a e is inc eased, bu , un o una ely, he PS blocking is also
inc eased.
The esul ob ained when edi ec ing se ice eques s blocked on equency 1 o 2, and ice e sa, is
shown in Figu e 17. The i s esul in Figu e 17, called “CS”, is he esul o edi ec ing speech se ice
eques s. The second esul , called “PS”, is he esul o edi ec ing HTTP se ice eques s when e e
hey a e no admi ed on he se ing equency. The hi d esul , “PS a e 10s”, is when HTTP se ice
eques s a e edi ec ed i s a e been queued 10 seconds on he se ing equency, i.e., a e he
HTTP se ice eques has egis e ed a PS block on he cu en equency. We see in he esul ha
bo h speech and HTTP gains when edi ec ing speech se ices. The speech blocking is howe e only
educed abou 6%, which is ai ly li le. On he o he hand, when edi ec ing PS se ices, he PS
blocking is educed 84%. The speech se ice does no imp o e i s QoS when HTTP eques s a e
edi ec ed. When edi ec ing a HTTP se ice i s a e 10 seconds, he PS blocking is also imp o ed a
lo . Fo his case he PS blocking is educed 68%.
VII- Hie a chical Cell S uc u es
In his sec ion some dynamic e alua ions on Hie a chical Cell S uc u e (HCS), aking in o accoun
mac o and mic o laye s, a e epo ed. The ac i i y’s goals we e o iden i y how o se he h esholds ha
con ol he algo i hm applied by he e minal o he iden i ica ion o he high/low mobili y s a e by means
o he numbe o cell eselec ions pe o med in a ixed pe iod o ime.
F om Figu es 18 and 19 i could be no ed ha , dec easing he Qo se 2 alues o mac o cells, he
pe cen age o ime du ing which he high-mobili y use s s ay on he mac ocellula laye inc eases while
he s aying o low-mobili y use s a e no a ec ed. Wi hou he se ing o he Qo se alue (i.e. wi h
Qo se se o 0), he p edominance o he mic ocellula laye is eally e iden : only 30% o high-mobili y
use s is on he mac o laye , and nea ly 95% o low-mobili y use s is on he mic o laye . Wi h he lowes
alue o Qo se , ins ead, he pe cen age o high-mobili y use s on he mac o laye pass 65%.Mo eo e ,
in Figu e 18 i has also o be no ed ha wi h Qo se alues lowe han -10 dB he pe cen age does no
inc ease mo e, as o Qo se alues highe han -4 dB he pe cen age does no dec ease signi ican ly.
I could be concluded ha in he simula ed layou , he mos a ec ing alues o Qo se a e be ween -4
dB and -10 dB.
Time pe cen age on Mac o laye
0.00%
10.00%
20.00%
30.00%
40.00%
50.00%
60.00%
70.00%
80.00%
0 -2 -4 -6 -8 -10 -12 -14
Qo se mac o cells
3 km/h use s
50 km/h use s
Figu e 18.- Pe cen age o ime o s ay on he Mac o laye – Mic o
wi h omni-di ec ional an ennas
Time pe cen age on Mic o laye
0.00%
20.00%
40.00%
60.00%
80.00%
100.00%
0 -2 -4 -6 -8 -10 -12 -14
Qo se mac o cells
3 km/h use s
50 km/h use s
Figu e 19.- Pe cen age o ime o s ay on he Mic o laye – Mic o wi h
omni-di ec ional an ennas
Simila esul s a e ob ained when i-sec o ial an ennas a e conside ed.
In ela ion o he capaci y analysis, Figu e 20 shows a ic esul s, conside ing:
• Inpu load: he a ic o e ed o he sys em
• Sys em load: he a ic ca ied by he sys em
• Blocked load: he blocked a ic (blocked when en e ing in he sys em)
• D opped load: he d opped a ic (du ing he call).
I has o be no ed ha , dec easing he alue o Qo se , he blocked and d opped loads inc ease and
hen he sys em load dec eases. The same phenomenon could be seen analyzing he blocking
2 This speci ies he o se be ween he se ing cell and neighbo cell. I is used o FDD cells in case he quali y measu e o cell selec ion and
e-selec ion is se o CPICH Ec/No.

Final Repo Page 17
pe cen age in case o new RAB se up, as depic ed in Figu e 21: bo h he uplink and he downlink
blocking pe cen ages inc ease wi h he inc ease o he seg ega ion o he use s on he laye s acco ding
o hei mobili y class. No ice ha , dec easing he alue o Qo se means o mo e use s om he mic o
laye o he mac o laye . The e o e all he high-speed use s a e mo ed om he mic o o he mac o
laye . The use s seg ega ion leads o inc ease he “nea - a ” e ec : in ac , conside ing a use camped
on a mac o cell, he could be close o a mic o cell an enna han a use camped on his mic o cell, hus
inc easing he uplink in e e ence. Fu he mo e, in downlink a simila si ua ion is p esen : a use
camped on a mac o cell could be close o a mic o cell an enna han he an enna o i s mac o cell, hus
inc easing he downlink in e e ence.
T a ic esul s
0.00
5.00
10.00
15.00
20.00
0 -6 -14
Qo se mac o cells
E l/squa e km
Inpu Load
Sys em Load
Blocked Load
D opped Load
Figu e 20.- Inpu load, Sys em load, Blocked load and D opped load
esul s
Pe cen age o RAB blocked
0.00%
1.00%
2.00%
3.00%
4.00%
5.00%
6.00%
7.00%
8.00%
0 -6 -14
Qo se mac o cells
UL
DL
Figu e 21.- RAB blocking pe cen ages due o uplink and downlink
conges ion
The esul s o his analysis highligh ha a ne wo k ope a o could manage he se ings o a ple ho a o
pa ame e s in o de o seg ega e he a ic acco ding o i s necessi ies. In he si ua ion desc ibed in his
s udy, he seg ega ion has been based on he mobili y class o he use s. In addic ion, i has been also
showed ha an ope a o may se he ade-o be ween he nega i e e ec s o he seg ega ion (e.g.
impac s on he in e e ence le el due o he inc ease o he “nea - a ” e ec ) and he bene i s o i ,
acco ding o i s objec i es. The decision o which solu ion has o be used is le o be assessed case by
case.
Ano he aspec ha has been s udied in HCS is he con enience o using o no di e en ca ie
equencies o he mac ocell and mic ocell laye depending on he a ic spa ial dis ibu ion. In
pa icula , le assume a scena io wi h di e en cells belonging o di e en HCS laye s in which he
spa ial a ic dis ibu ion is non-homogeneous because o he exis ence o ho spo s. Such ho spo s
may no in gene al be pe manen bu hey will change in he di e en pe iods o he day and
geog aphical loca ions, due o e.g. people concen a ions a a bus-s op, es au an s a eas du ing he
meal imes, e c. As a esul , he load obse ed by he di e en cells as well as he in e ac ions be ween
hem may expe ience signi ican a ia ions du ing hese pe iods. Al hough hese issues may be co e ed
a some ex en by s a ic adio ne wo k planning when de ising he numbe and loca ions o he di e en
cells and equency alloca ions, he dynamics and andomness associa ed o he human beha iou may
lead o si ua ions in which he equency alloca ion should be modi ied due o e.g. an excessi e load in
ce ain cells. Unde his amewo k, a p oposal based on he load g adien compu a ions has been
analysed o de ec he cells ha ing he highes in luence o e he es o he cells in he scena io, hus
being able o decide which cells should ope a e wi h a di e en ca ie equency, whene e a ca ie
equency should be changed in he scena io. The p oposed me hod has been compa ed wi h ano he
scheme in which he cell wi h he highes load ac o is selec ed o ope a e wi h ano he equency.
Resul s in e ms o ou age shown in Table 4 e eal o di e en case s udies ha he g adien
me hodology clea ly ou pe o ms he load-based algo i hm as well as he case in which a single ca ie
is used in he scena io.
Table 4.- Ou age p obabili y in he di e en scena ios
Equal F eq.
G adien
Alg.
Load.
Alg.
Case 1
2.52 %
1.06 %
1.62 %
Case 2
28.92 %
18.28 %
27.99 %
Case 3
6.89 %
3.45 %
4.87 %
Final Repo Page 18
VIII- T anspo Channel Type Swi ching
This sec ion summa izes he wo k ca ied ou wi h he aim o in es iga e he sys em-le el pe o mances
o UTRAN when a T anspo Channel Type Swi ching (TCTS) algo i hm is aken in o accoun . No ice
ha In he case o a discon inuous da a ans e , i is e y impo an o be able o op imize he usage o
dedica ed anspo channel (DCH), p e en ing channeliza ion code sho age in he downlink wi hou
deg ading he end- o-end quali y o se ice expe ienced by he use in an app eciable manne .
Figu e 22 epo s he o al amoun o a ic ca ied by he adio access ne wo k (“ a ic load”) in he wo
di e en cases, oge he wi h he co esponding amoun o blocked a ic (“blocked load”) 3.The
achie ed esul s show ha he UTRAN capaci y is highe when he anspo channel ype swi ching
algo i hm is ac i e. The capaci y gain achie ed by he swi ching algo i hm can be jus i ied aking in o
accoun he admission and conges ion con ol p ocedu es ope a ed by UTRAN. Wi h espec he
admission con ol, he simula ion esul s show ha 3.8% o adio access bea e se up eques s a e
blocked due o sho age o a ailable downlink codes in he DCH-only case, whe eas hese e en s do
no occu in he TCTS case, as ep esen ed in igu e 23.
2615
2413
201.7
2615
2432
183.1
0
500
1000
1500
2000
2500
3000
a ic (kby e/s)
DCH-only TCTS
T a ic load
Inpu load
T a ic load
Blocked load
Figu e 22.- T a ic esul s (DCH-only e sus TCTS)
3.82%
0
0.00%
1.00%
2.00%
3.00%
4.00%
DCH-only TCTS
Pe cen age o RAB se up eques s blocked
due o lack o DL codes
Figu e 23.- Pe cen age o RAB se up eques s blocked due o lack
o DL codes (DCH-only e sus TCTS)
The di e ence in he achie ed esul s be ween he wo cases can be unde s ood aking in o accoun
ha in DCHonly mode, each DCH has i s own speci ic OSVF code co esponding o an app op ia e
sp eading ac o assigned o he whole du a ion o he connec ion. I he numbe o use s equi ing
DCHs becomes high, he code sho age cause he block o some use s.
In summa y, and as a gene al ule, e en i he common channels a e used only when a e y small
amoun o da a (o no hing a all) ha e o be ans e ed, he QoS is be e o DCH-only han o TCTS
because o he swi ching ime. DCH o e s highe ans e speeds ( h oughpu ) bu equi es a signi ican
se up ime, whe eas sha ed channels ha e a low h oughpu bu also a low se up ime. Due o hese
conside a ions he usage o common channels is mo e e icien only when he a ic is spo adic o o
sho ime, o he wise o long and equen da a sessions, he usage o a dedica ed channel is highly
ecommended. Ano he e y impo an issue consis s in he p ope se ing o he pa ame e s con olling
he epo ing o he a ic measu emen s pe o med by he use s, in o de o p e en an o e head o
useless signalling a ic.
IX- Di Se Awa e Scheduling
This sec ion ocuses on he colou awa e RRM algo i hms o suppo Di Se o e CDMA adio
sys ems. The schemes a e designed o use he IP laye a ic in o ma ion o inc ease he u iliza ion o
adio ne wo k. The c i ical impe us o he p oposed app oach, beyond minimiza ion o he ansmi ed
powe , is mainly wo old. Fi s ly, o achie e he equi ed pe -class agg ega e da a a e while p io i izing
and ensu ing QoS o in-p o ile packe s, and secondly o inc ease powe gains by penalizing ou -o -
p o ile packe s in sense o powe consump ion. The seminal aspec o he p oposed scheme is ha
angible powe gains can be achie ed by di e en ia ing ansmission o con o man and non-con o man
3 Taking in o accoun ha he use s ha e been dis ibu ed uni o mly in all he 36 cells o he simula ed scena ios, he mean amoun o a ic in
inpu pe each cell is app oxima ely equal o 600 kbi /s in bo h cases, as epo ed in 3.4.
Final Repo Page 19
packe s while a he same ime he agg ega e powe gains o AF classes can be u ilized o enhance he
pe o mance o in-p o ile a ic.
Figu e 24.- T ansmission Ra e.
Figu e 25.- Delay Pe o mance o IF and MRR wi h colo awa e.
Resul s in Figu e 24 clea ly show he di e en ia ion in esou ce alloca ion in e ms o ansmission a e
wi h he p oposed scheme. In his igu e as expec ed, he ansmission a es dec ease o all ype o
use s as a ic load inc eases. And he ansmission a es o all ypes o packe a e almos he same
wi hou colo awa e shown by he solid line. Howe e in he case wi h colo awa e queue, he
ansmission a e o each colo is di e en o each o he . G een use s always ha e a be e a e han
o he use s because hey a e gi en high p io i y han o he use s. Also as a ic load inc eases, he
di e ence becomes la ge . The ed use s’ ansmission a e dec eases mo e apidly ( om mo e han
145by s d ops o jus o e 100by es) and in he con as , he g een use s a e a e only d op abou
10by es. Figu e 25 shows he delay pe o mance o MRR (Modi ied Round Robin) and IF (In e e ence
Fac o )4 queue bo h wi h colo awa e. IF o e s a be e pe o mance in he a e age delay. The di e si y
o delay pe o mance o di e en colo ed packe s shows he consis ence wi h he di e en ia ion in
ansmission a e alloca ion p esen ed in Figu e 24. Again o enhance he delay pe o mance o he in
p o ile packe s such as g een packe s, he ou p o ile packe s such as ed packe s a e penalized in high
a ic load si ua ion.
X- High Speed Downlink Packe Access(HSDPA)
The main bene i wi h he high-speed downlink packe access (HSDPA) e olu ion o he WCDMA is ha
i educes he use packe call delay and i inc eases he sys em capaci y in downlink. This inc eased
capaci y can be used o ei he inc ease he numbe o use s in each cell, and/o o p o ide he exis ing
use s wi h highe a e age da a a es. I is impo an o no e ha he ade-o be ween HSDPA sys em
h oughpu e sus QoS is c i ical, and se e al di e en pa ame e s needs o be aken in o accoun
when designing RRM s a egies o HSDPA. The scheduling me hod used o decide which use is
alloca ed o he channels is o g ea impo ance. The HSDPA scheduling is e y lexible. I allows
sha ing o he sp eading codes (he e deno ed as “channels”). This means ha one o se e al channels
a e alloca ed o one use and se e al use s may ecei e da a unde one ansmission ime in e al
(TTI). I is also possible o schedule all channels o one use unde one TTI, and his is conside ed as
he p ima ily means o sha ing.
Ad anced schedule s, ha combine bo h channel in o ma ion and se ice equi emen s, ha e been
e alua ed and he esul s a e p esen ed in e ms o sa is ied use s and cell h oughpu bo h o indoo
and ehicula en i onmen s. In pa icula , schedule algo i hms we e p oposed o he ideo s eaming
and a e based on weigh ing pa ame e s ha a ec he applica ion unc ionali y and he sys em (cell)
h oughpu . Complemen a y o he schedule , he RRM alloca ion policy ies o main ain lexibili y by
assigning se e al use s o a single TTI, scheduling indi idual packe s.
4 A each ame, he use s a e o de ed acco ding o hei p opaga ion condi ions. The use wi h he bes p opaga ion condi ion will be placed
on he op he queue
Final Repo Page 20
Call h oughpu (25 use s)
64.2
64.3
64.4
64.5
64.6
64.7
64.8
64.9
MaxCI
P io i yS d
P io i yA gDelay
P io i yQueueSize
Th oughpu (kbp
Call h oughpu (kbps)
QoS call h(kbps)
Figu e 26. Use a e age bi a e – Compa ing all schedule s
% o Sa is ied use s (25 use s)
0.86
0.88
0.90
0.92
0.94
0.96
0.98
MaxCI P io i yS d P io i yA gDelay P io i yQueueSize
Figu e 27. Pe cen age o sa is ied use s –
Compa ing all
schedule s
Compa ing he schedule s, conce ning QoS, be e esul s bo h o QoS bi a e and sa is ied use s,
we e ob ained wi h he P io i y S anda d schedule , P1. The Max. CI schedule is he one ha lead o
wo s QoS esul s, since i does no conside se ices cons ains.
In addi ion o ha , a easibili y s udy on suppo ing VoIP o e HSDPA by using biased AMC scheme
was also ca ied ou . Dynamic simula ion esul s showed ha a sys em capaci y o 70 E langs can be
achie ed a eloci ies up o 20 km/h, by combining a longes delay i s (LDF) discipline wi h a
p opo ionally ai (PF) code mul iplexing schedule , and by using a block size o 20 o 80 ms. Howe e ,
a 80 km/h he capaci y is educed o 12 E langs due o he inabili y o ack as ading wi h CQI
eedback delay; he AMR and schedule has o ely on ou da ed CQI epo s. Obse ing ha he
channel es ima e is less accu a e when he channel SIR is low, a biased AMC scheme was u he
in oduced. The ob ained simula ion esul s showed ha adap ing he o se by he a e age SIR
( ep esen ing loca ion) p oduces a ma ginal signi icance, and a ixed bias is su icien o imp o e he
VoIP QoS. This implies ha he CQI eedback delay is he main e il ha mus be mi iga ed. The esul s
also showed ha he biased AMC inc eases he sys em capaci y by o e a h ee- old a 80 km/h.
Mo eo e , HSDPA is capable o suppo ing a ious Qos le els. In ac , he HSDPA speci ica ion
suppo s up o 16 p io i y le els and 8 queues pe use . Then, when he e ogeneous a ic coexis s, an
inno a i e HSDPA packe schedule ha disc imina es QoS classes was p oposed o p o ide be e
QoS suppo han hose ha conside only he channel condi ion. Th ee se ices a e conside ed: oice
o e IP (VoIP), a iable bi a e (VBR) ideo s eaming (MPEG, he ea e ), and web-b owsing (HTTP,
he ea e ).The p oposed schedule was shown o suppo he he e ogeneous a ic e icien ly, wi hou
causing any pa icula se ice being a bo leneck in p o iding capaci y. The simula ion esul s u he
showed ha he p oposed schedule is obus o changes in he a ic cons i u ion, and suppo s call
a i al a es o up o 3.5 calls/s in gua an eeing 5% ou age o all he se ices in he de aul a ic
scena io. This is subs an ially la ge han some con en ional schedule s, as s ic p io i y queuing
schemes suppo only up o 2.8 calls/s, whe eas single queue schemes ail o secu e 1 call/s, acco ding
o he simula ion esul s.
XI- RAN Sha ing
Sha ing spec um can be e y a ac i e, o example, in u al a eas UMTS co e age can be o e ed
wi h much lowe in es men cos s, bu also in u ban a eas and ho spo a eas capaci y gains can be
achie ed. Ce ainly, he e is a capaci y gain due o he inc eased unking e iciency as channels a e
pooled oge he be ween he ope a o s. Howe e , Fai ness o esou ce alloca ion can become an issue
when ope a o s sha e spec um.
Table 5. 50% o use s om Ope a o A and 50% om Ope a o B
No admission
con ol
To al powe wi h
d opping
Hal powe
Mul i-cell
Admission (%)
100
91.20
83.89
89.35
D opping (%)
6.63
1.53
0.16
0.67
GoS
66.39
24.08
17.7
17.39 [G=1.75%]
Final Repo Page 27
Figu e 37.- Cumula i e dis ibu ion unc ion o end- o-end delay a
link laye le el o agg ega ed oice s eam. Wi hou
TXOP
Figu e 38.- Cumula i e dis ibu ion unc ion o end- o-end delay a link
laye le el o he agg ega ed oice s eams. Wi h TXOP
Compa ing he accumula i e dis ibu ion unc ion o he delay o he o agg ega ed oice s eams wi h
(Figu e 38) and wi hou (Figu e 37) TXOP mechanism i could be ealized ha he TXOP mechanism has
signi ican ly enhanced he sys em pe o mance. Ce ainly, Figu e 38 shows how now, wi h a TXOP limi ,
he expe ienced link laye delay by oice s eams is lowe han he ob ained wi hou TXOP limi (see
Figu e 37). Fo ins ance in he case o 3@19 he ob ained bound delay o he 95% o he cases is
0.0298 sec. wi h TXOP whe eas wi hou TXOP his limi eaches 0.3520 sec.). This imp o emen is due
o he ac ha when using packe bu s ing he e is lowe numbe o packe s con ending o a channel
access. Mo eo e , also no ice ha he ob ained delay alues o oice a ic a e wi hin equi ed limi s
(less ha 100 msec.) e en in deep sa u a ion scena io 7@110 when TXOP mechanism is used.
Mo eo e , simula ion esul s show ha he e exis op imum TXOP limi alues in e ms o sys em
pe o mance. Howe e , as hese alues change wi h se ice mix dis ibu ion a no el dynamic TXOP
con igu a ion algo i hm based on he numbe o packe s in AP queues has been p oposed. The
de eloped algo i hm p o ides TXOP limi s ha a e wi hin minimum mos ad an ageous TXOP limi s and
hence quasi-op imal sys em pe o mance is eached independen ly o sys em se ice mix.
Finally, by using he TXOP limi o pe o m agmen a ion o low p io i y packe s, u he enhancemen o
p io i iza ion mechanism is ob ain.
E alua ion o CRRM algo i hms
Common Radio Resou ce Managemen (CRRM) e e s o he se o unc ions ha a e de o ed o ensu e
an e icien use o he a ailable adio esou ces in he e ogeneous ne wo ks scena ios by means o a
p ope coo dina ion be ween he di e en adio access ne wo ks.
Wi h espec o Common Radio Resou ce Managemen , a i s con ibu ion o he p ojec has been he
ou line o a unc ional model o ha ing a common managemen o he pool o adio esou ces in
he e ogeneous scena ios. In he p esen ed model, he Common Radio Resou ce Managemen (CRRM)
e e s o he se o unc ions ha a e de o ed o ensu e a p ope coo dina ion be ween he di e en adio
access ne wo ks o achie e he mos e icien use o he a ailable adio esou ces. Di e en app oaches
o he CRRM and RRM in e ac ion ha e been p esen ed, ou lining he po en ial le els o coo dina ion in
he adio esou ce managemen decisions in he iden i ied unc ionali ies. Finally, he equi emen s in
e ms o in e wo king capabili ies and conside a ions abou he physical CRRM implemen a ion ha e
been de ailed. Acco ding o his amewo k, and aking in o accoun he scope and ime- ame o he
EVEREST p ojec , he CRRM s udies epo ed he e assume he unc ionali y spli in which he CRRM
akes cha ge o he RAT selec ion p ocedu es, including bo h ini ial RAT selec ion and e ical hando e ,
while he RAT-speci ic RRM algo i hms, like admission con ol, conges ion con ol o packe scheduling
a e execu ed locally a he RRM en i ies.
9 3@1, 3 web s a ions ansmi a 1 Mbps and he es a 11Mbps
10 7@1, 3 web, 2 ideo and 2 oice s a ions ansmi a 1 Mbps. and he es a 11 Mbps;

Final Repo Page 28
I- Se ice-Based RAT Selec ion Policies
A gene al policy-based amewo k o he speci ica ion o CRRM algo i hms has been de ined and
di e en policies conside ing he se ice ype as well as he ac ha he use s may be indoo o
ou doo ha e been e alua ed h ough simula ions. In o de o s ess he in luence o he ini ial RAT
selec ion p ocedu e, wi hou aking in o conside a ion hando e issues, no e ical hando e s a egy is
ini ially conside ed in hese simula ions, Table 6 p esen s he agg ega e h oughpu (in Mbi /s) achie ed
wi h he sum o bo h RATs (GERAN and UTRAN) and wi h he sum o bo h se ices ( oice and www)
o he wo conside ed basic se ice-based policies: Voice UTRAN (VU) and Voice GERAN (VG).
Table 6.- Agg ega e h oughpu o he di e en policies
VU
VG
VG (no T CH swi ch)
UL
DL
UL
DL
UL
DL
www
use s
0.5
Km
1
Km
0.5
Km
1
Km
0.5
Km
1
Km
0.5
Km
1
Km
0.5
Km
1
Km
0.5
Km
1
Km
200
2.18
2.08
2.22
2.17
2.14
2.14
2.20
2.22
2.03
2.01
2.08
2.07
600
3.01
2.88
3.15
3.09
2.96
2.95
3.16
3.15
2.06
2.05
2.11
2.11
1000
3.80
3.64
4.05
3.96
3.77
3.76
4.08
4.08
2.08
2.05
2.14
2.13
I has been ob ained ha , in ou doo scena ios, VG basic policy u ns in o a highe h oughpu han VU,
ensu ing lowe in e ac i e packe delay. This is because o he highe e iciency o non- eal ime a ic
ansmission in UTRAN achie ed in he VG case, since web b owsing a ic is suppo ed by means o
dedica ed channels whe eas in VU a packe scheduling algo i hm mus be implemen ed in GERAN.
Ano he ype o basic ini ial RAT selec ion policies a e hose ha ake in o accoun adio ne wo k
conside a ions o each speci ic RAT. Fo ins ance, WCDMA capaci y is highly deg aded by indoo a ic
use s, which leads o de ining a new basic policy, deno ed as IN (indoo ) in which indoo use s a e
alloca ed in GERAN and ou doo use s in UTRAN. By combining he basic policies VG,VU and IN, he
ollowing complex policies ha e also been s udied:
• VG*INI he se ice is oice and he use is ou doo , he i s choice will be o alloca e i in
GERAN (i.e. acco ding o he VG policy). I no capaci y is a ailable in GERAN, he second choice
will be o alloca e i in UTRAN. I he se ice is www and he use is ou doo , a blocking will occu
i he e is no capaci y in UTRAN.
• IN*VG In his case, he i s choice akes in o accoun whe he he use is indoo o ou doo
and, i no capaci y is a ailable in he selec ed RAT, he second choice conside s he se ice ype.
• VG*VU Acco ding o his policy, oice use s will i s ill he capaci y a ailable in GERAN and
hen hey will be di ec ed o UTRAN. In u n, www use s will i s ill he capaci y in UTRAN and
hen hey will be di ec ed o GERAN. In his case, no eques is blocked p o ided ha he e is
capaci y a ailable in ei he UTRAN o GERAN
In scena ios wi h a mix o indoo and ou doo use s wi h di e en se ices, up o medium oice loads
(i.e. 200 use s) no ele an di e ences be ween he policies a e obse ed, al hough in gene al he
pe o mance o IN*VG is somewha poo e , mainly when he numbe o www use s inc eases. The
eason is ha , wi h IN*VG, he e is a highe numbe o www use s ha a e se ed h ough GERAN (i.e.
hose ha a e indoo ), which p o ides highe delays and lowe www h oughpu han UTRAN (see
Figu e 39 and Figu e 40). Fu he , when he a io o indoo use s inc eases, he numbe o in e ac i e
use s alloca ed in GERAN also inc eases and, consequen ly, IN*VG pe o mance is mo e deg aded.
Final Repo Page 29
200 oice
0
0.5
1
1.5
2
2.5
0200 400 600 800 1000 1200
www use s
DL Th oughpu www
VG*VU
VG*IN
IN*VG
Figu e 39- To al DL www h oughpu wi h 50% indoo a ic
200 oice, IN*VG
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
0200 400 600 800 1000 1200
www use s
DL a e age page delay (
UTRAN
GERAN
Figu e 40- DL a e age page delay wi h IN*VG policy and 50% indoo
a ic
In summa y, he pe o mance o IN*VG policy imp o es when he oice load inc eases, he www load
dec eases and he e is a high ac ion o indoo use s. On he con a y, o low oice loads and high
www loads VG*IN achie es a be e h oughpu . This sugges s ha he sui able con igu a ion o he
RRM and CRRM en i ies acco ding o speci ic policies depends on he exis ing a ic condi ions and
he e o e i may be modi ied a e.g. di e en pe iods o he day.
Wi h espec o e ical hando e , he in e wo king be ween ho izon al and e ical hando e s has been
s udied, wi h wo conside ed app oaches, namely he igh app oach T-VHO, in which he e ical
hando e algo i hm is execu ed a e e y ime ha a ho izon al hando e algo i hm should be ca ied
ou , so ha bo h possibili ies a e conside ed p io o aking a decision, and he loose app oach L-VHO,
in which he e ical hando e algo i hm is execu ed only when a ho izon al hando e ails o when a
call d opping is abou o occu due o bad p opaga ion condi ions.
When compa ing T-VHO wi h L-VHO, he i s aspec o conside is how he a ic dis ibu ion changes
in he di e en RATs. In pa icula , when 400 oice use s a e p esen in he scena io, Figu e 41 shows
he pe cen age o oice a ic se ed h ough UTRAN as a unc ion o he numbe o www use s. In his
case, e en when no e ical hando e is a ailable, a ce ain po ion o he oice a ic is se ed h ough
UTRAN, co esponding o he new oice calls ha a e o igina ed in a blocked GERAN cell ( o his
oice a ic load, he p obabili y ha his occu s is ound o be a ound 0.5%). In u n, when e ical
hando e is used, a ce ain numbe o oice use s a e also ans e ed om GERAN o UTRAN when
hey each blocked cells du ing ho izon al hando e s, hus inc easing he ac ion o oice a ic se ed
h ough UTRAN. No ice ha wi h he igh app oach he e a e less oice use s in UTRAN han wi h he
loose app oach, because in he la e use s ha a e handed o e o UTRAN will end o emain he e,
while wi h he igh app oach a oice use in UTRAN will y o e u n o GERAN in he i s UTRAN
ho izon al hando e .
400 oice
0
2
4
6
8
10
12
0200 400 600 800 1000 1200
www use s
UL Th oughpu oice UTRAN (%)
L-VHO
T-VHO
No VHO
Figu e 41 Pe cen age o oice h oughpu se ed h ough UTRAN o
he 400 oice use s case
400 oice
0
0.05
0.1
0.15
0.2
0.25
0200 400 600 800 1000 1200
www use s
UL Weigh ed Delay www (s)
L-VHO
T-VHO
No VHO
Figu e 42 Uplink weigh ed delay o www a ic wi h 400 oice use s
No ice ha he igh e ical hando e app oach acili a es he ul ilmen o he se ice policy, se ing he
www a ic as much as possible h ough UTRAN. This has an impo an impac in e ms o weigh ed
a e age delay o www a ic, as e lec ed in Figu e 42
In summa y, i has been shown ha he a ic dis ibu ion among he conside ed RATs can be qui e
di e en wi h he wo app oaches. In ha sense, he igh app oach allows a be e ul ilmen o he ini ial
Final Repo Page 30
RAT selec ion policy, due o ha ing mo e chances o execu e a e ical hando e han in he loose
case. Consequen ly, and o a se ice-based RAT selec ion policy, i has been obse ed ha he igh
app oach o e s a be e pe o mance in e ms o lowe delay o in e ac i e www use s han he loose
app oach because mos o he www a ic is se ed h ough UTRAN. On he o he hand, a highe
numbe o e ical hando e p ocedu es a e also equi ed wi h he igh app oach, which inc eases he
signalling o e head.
II.- Load Balancing -Based RAT Selec ion Policies
Load balancing (LB) is a possible guiding p inciple o esou ce alloca ion in which he RAT selec ion
policy will dis ibu e he load among all esou ces as e enly as possible. Taking his in o accoun , he
pe o mance o load balancing p inciples in he RAT selec ion p ocedu e has been co e ed and
compa ed agains a se ice-based policy.
Figu e 43 shows he a e age cell load o he cen al base s a ion in bo h RATs o SM111 when policies
VGVU and LB a e used. No e ha o VGVU policy, oice use s a e di ec ed o GERAN while no ully-
loaded; o he wise, eques s a e ans e ed o UTRAN. Load balancing policy LB beha es as expec ed,
main aining cell load le els in bo h RATs a app oxima ely he same le el. Figu e 44 shows he load
le els o policies VGVU and LB o SM2. Fo his se ice mixing, policy LB does no show a isible
imp o emen wi h espec o he se ice class policy VGVU. In ac , he load le els a e kep low,
compa ed o SM1, and he e o e no se e e d opping occu s.
600 in e ac i e
0
0.2
0.4
0.6
0.8
1
1.2
0200 400 600 800 1000
oice use s
UL Load GERAN
VGVU
LB
(a)
600 in e ac i e
0
0.2
0.4
0.6
0.8
1
1.2
0200 400 600 800 1000
oice use s
UL Load UTRAN
VGVU
LB
(b)
Figu e 43- A e age UL load in (a) GERAN and (b) UTRAN o policies VGVU and LB wi h SM1
200 oice
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0200 400 600 800 1000 1200 1400
in e ac i e use s
UL Load GERAN
VGVU
LB
(a)
200 oice
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0200 400 600 800 1000 1200 1400
in e ac i e use s
UL Load UTRAN
VGVU
LB
(b)
Figu e 44- A e age UL load in (a) GERAN and (b) UTRAN o policies VGVU and LB wi h SM2.
In summa y, wi h espec o he ini ial RAT selec ion based on load balancing wi hou including VHO,
esul s e ealed a igh dependency be ween he sui abili y o load balancing RAT selec ion and
11 In SM1 he numbe o in e ac i e use s is ixed while oice use s inc ease. On he con a y, in SM2 he numbe o oice use s is ixed while
he numbe o in e ac i e use s inc eases
Final Repo Page 31
se ice-class mixing. I has been shown ha e en hough he o e all h oughpu may inc ease wi h load
balancing policies, his a he expense o in e ac i e a ic pe o mance deg ada ion. Ne e heless,
o he se ice ype mixings showed no ype o h oughpu imp o emen a all
In u n, he in oduc ion o VHO capabili ies allows highe lexibili y in he alloca ion o mul i-se ice
use s in a mul i-access scena io. We ha e compa ed wo ini ial RAT selec ion policies along wi h a igh
app oach o VHO p ocedu es. Resul s indica e ha no ema kable imp o emen is no ed on he o al
agg ega e h oughpu when using he LB policy as opposed o he VGVU policy. Mo eo e , wi h LB,
in e ac i e use s unde go highe a e age packe delays which impac he use ’s pe cei ed QoS.
Howe e , we ha e seen ha load balancing p ocedu es may imp o e he call d opping p obabili y due
o a mo e lexible alloca ion o use s on o bo h RATs, which is also a key pe o mance indica o o
conside .
III- Pa h Loss - Based RAT Selec ion
In FDMA/TDMA-based access sys ems (e.g. GSM/GPRS) he e is no in a-cell in e e ence, whe eas in
con as , in CDMA-based sys ems (e.g. UMTS) he in a-cell in e e ence is caused by e e y single use
ansmi ing in he cell. Fu he mo e, in e -cell in e e ence is also o igina ed by all simul aneous use s
in all neighbou ing cells, since a comple e equency euse is conside ed. CDMA sys ems a e much
mo e sensi i e o mul i-use in e e ence han FDMA/TDMA ones. The unde lying idea o he CRRM
app oach de eloped he e is o ake ad an age o he co e age o e lap ha se e al RANs using
di e en access echnologies may p o ide in a ce ain se ice a ea in o de o imp o e he o e all
in e e ence pa e n gene a ed in he scena io o he CDMA-based sys ems and, consequen ly, o
imp o e he capaci y o he o e all he e ogeneous scena io. This can be achie ed by con olling he
e ec i e cell adius o CDMA-based sys ems (i.e. a con olled cell-b ea hing e ec ) h ough app op ia e
RAT selec ion app oaches ha ake in o accoun he measu ed pa h loss. In his way, he in e e ence
le el in CDMA-based RATs is educed while a he same ime he a ge co e age a ea is assu ed by
means o he coope a ion o he FDMA/TDMA-based RATs. The abo e concep , deno ed he e as
Ne wo k-Con olled Cell-B ea hing (NCCB) can be e ec i ely complemen ed wi h load balancing
conside a ions.
Figu e 45 p esen s he compa ison in e ms o he o al agg ega ed h oughpu in he scena io (i.e.
including UTRAN and GERAN) agains a load balancing (LB) app oach when only oice se ice is
conside ed. The bene i s o he p oposed s a egy can be clea ly obse ed wi h h oughpu inc eases o
a ound 13% in he uplink and 24% in he downlink. Simila ly, Figu e 46 illus a es he imp o emen s
ob ained in e ms o d opping and blocking p obabili ies, espec i ely, showing he be e e iciency o
he NCCB s a egy o alloca e he a ailable esou ces in he wo RATs among he di e en use s.
1.5
2
2.5
3
3.5
4
4.5
5
400 500 600 700 800 900 1000 1100 1200
Use s
Agg ega ed h oughpu (Mb/s)
NCCB (Uplink) LB (Uplink)
NCCB (Downlink) LB (Downlink)
Figu e 45.- To al h oughpu in uplink and downlink
0
5
10
15
20
25
30
400 500 600 700 800 900 1000 1100 1200
Use s
(%)
Blocking (NCCB) Blocking (LB)
D opping (NCCB) D opping (LB)
Figu e 46.- Blocking and d opping p obabili ies
In u n, Figu e 47 and Figu e 48 show he e ec o a ying he main pa ame e o he NCCB algo i hm,
which is he h eshold PL h (i.e. use s wi h pa h loss below PL h a e alloca ed in UTRAN while use s
wi h pa h loss abo e PL h a e alloca ed in GERAN). I can be obse ed ha he alue PL h=120 dB,
co esponding o he 60- h pe cen ile o he pa h loss dis ibu ion (and he e o e achie es a ce ain
deg ee o “load balancing” be ween he wo RATs) ob ains he bes pe o mance. No ice in any case
ha he imp o emen wi h espec o LB comes om he sma e way o dis ibu ing use s p o ided by
NCCB.
Final Repo Page 32
0
2
4
6
8
10
12
14
16
18
400 500 600 700 800 900 1000 1100 1200
Use s
UL BLER UTRAN (
PL h=115 dB
PL h=120 dB
PL h=125 dB
LB s a egy
Figu e 47.- UL BLER in UTRAN o di e en alues o he h eshold
PL h
1.5
2
2.5
3
3.5
4
4.5
5
400 500 600 700 800 900 1000 1100 1200
Use s
UL Th oughpu (M
PL h=115 dB
PL h=120 dB
PL h=125 dB
LB s a egy
Figu e 48.- Uplink h oughpu o di e en alues o PL h
In summa y, when conside ing a single oice se ice, he NCCB s a egy has been compa ed agains a
classical load balancing s a egy ha ies o keep he same load le el in bo h RATs. Resul s e eal ha
a signi ican imp o emen in e ms o capaci y o bo h uplink and downlink is achie ed wi h he
p oposed s a egy. I has been also shown ha , by a p ope se ing o he maximum pa h loss PL h,
load balancing p inciples can also be achie ed in NCCB, hus ob aining he bene i s in e ms o
lexibili y o load balancing while a he same ime exhibi ing a highe capaci y han a pu e load
balancing.
When conside ing a mix be ween oice and www use s, di e en combina ions o he NCCB s a egy
wi h se ice-based policies ha e been es ed. I has been obse ed ha he bes pe o mance is
achie ed wi h he so-called NCCB_ oice, co esponding o applying he NCCB s a egy only o oice
use s and alloca ing www use s in UTRAN acco ding o he se ice-based policies. I has been also
obse ed ha he adequa e se ing o he h eshold PL h depends on he exis ing a ic mix and he
ade-o among www delay and oice BLER.
IV- RAT P io i y Lis -Based RAT Selec ion
The conside ed algo i hm in ends o dis ibu e he a ic be ween he di e en RATs ollowing se e al
a ge s and makes use o se ice-based policies in he o m o a RAT p io i y lis o each speci ic
se ice, while a he same ime i in oduces load-balancing concep s.
The objec i e o he s udy was he analysis o he possible impac in ne wo k capaci y o a join
admission con ol 2G/3G which akes in o accoun he kind o se ice. The s a egies simula ed a e
lis ed below:
• S a egy A: No con ol o e admission con ol is applied. When use s may be se ed by bo h
echnologies, GERAN and UTRAN, he decision is aken in a andom way, wi hou no p ede ined
decision om he he ope a o . Tha is, bo h RATs ha e he same p io i y o all he se ices.
• S a egy B: Voice se ice use s a e only se ed by GERAN. Da a se ices use s a e se ed by
UTRAN i is possible; i no , hey a e se ed by GERAN.
• S a egy C: Voice se ice use s a e se ed by GERAN i is possible; i no , hey a e se ed by
UTRAN. Da a se ices use s a e se ed by UTRAN i is possible; i no , hey a e se ed by
GERAN.
Figu es 49 and 50 show he pe cen age o a ended use s o bo h oice and ideo s eaming se ices.
Simila esul s a e ob ained o ideo- elephony and b owsing se ices

Final Repo Page 33
Figu e 49.- Pe cen age o a ended use s o oice se ice
Figu e 50.- Pe cen age o a ended use s o ideo s eaming se ice
F om he esul s o he simula ions he ollowing conclusions may be in e ed:
• Fo low load si ua ions, he s a egy B ( oice se ice only p o ided by GERAN and da a se ices
o UTRAN i is possible and, i no , o GERAN) shows he be e pe o mance o all he
se ices.
• Fo high load si ua ions, he s a egy C shows he be e beha iou in gene al because he
ne wo k capaci y is maximized in he wo k condi ions (s ong p esence o oice se ice in he
se ice mix). In his si ua ion, he s a egy B is which shows he wo s pe o mance wi h a high
numbe o oice use s no se ed. Howe e , i is possible ha s a egy B would pe o m be e in
a scena io wi h a s ong p esence o da a se ices
V- Pe cei ed TCP Th oughpu in CRRM F amewo k
The impac o CRRM s a egies on TCP h oughpu has also been s udied in a scena io including 3G
wi h HSDPA, 2G and WLAN sys ems. Fou di e en CRRM algo i hms, desc ibed in able 7, ha e been
e alua ed and compa ed agains he manual RAT selec ion algo i hm.
Table 7.- Fou di e en CRRM algo i hms used o CRRM pe o mance e alua ions
Type o algo i hm
Algo i hm desc ip ion
Algo i hm 1
Long e m CRRM
op imisa ion c i e ia
A each e alua ion he CRRM decisions a e made ying o minimise he ime o
emp y all bu e s assuming he cu en load.
Algo i hm 2
Sho e m CRRM
op imisa ion c i e ia
A each e alua ion he CRRM decisions a e made ying o minimise he ime un il
he i s da a session is inalised.
Algo i hm 3 RAT p io isa ion o
CRRM
Acco ding o se ice p io isa ion due o SLA o o he ope a o speci ic QoS
s a egies, he se ice wi h highes p io i y is alloca ed o he RAT wi h highes
sys em capaci y, i.e. WLAN in his scena io. I he high p io i y da a would o e load
he high capaci y RAT, he nex RAT (e.g. UTRAN) is also illed wi h high p io i y
da a e c. I he sys em can handle mo e a ic, he payload wi h he second highes
p io i y is alloca ed in he CRRM using he same p inciple, i.e. wi h WLAN usage as
bes choise.
Algo i hm 4 Manual RAT selec ion
The manual RAT selec ion used is ha all payload wi h highes p io i y is alloca ed
o WLAN, while he lowes p io i y payload is alloca ed o GERAN. The payload
om o he “middle p io i y se ices” is hen
alloca ed o UTRAN.
This CRRM algo i hm is used as he benchma k o he o he algo i hms.
The ob ained esul s show ha wi hou HSDPA, h oughpu inc eases up o 40% ha e been obse ed
hanks o CRRM. When in oducing HSDPA capabili ies, he imp o emen in 3G h oughpu leads o
inc eases o up o 60%.
VI- Impac o Mul i-Mode Te minals on CRRM Pe o mance
Today’s wi eless communica ions comp ise a b oad a ie y o Radio Access Technology (RAT)
s anda ds. In o de o ake ull ad an age o B3G ne wo ks, exis ing mobile e minal capabili ies need o
be ex ended. Pa icula ly, o p o ide connec i i y o a a ie y o unde lying access echnologies is a
Final Repo Page 34
mus . The e o e, he impac o mul i-mode e minals in he amewo k o he e ogeneous ne wo ks wi h
an ini ial RAT selec ion policy has also been assessed.
Figu e 51 shows he h oughpu deg ada ion de ined as he h oughpu dec emen wi h espec o he
h oughpu achie ed when 100% o use s ha e mul i-mode e minals in a scena io whe e RAT selec ion
is done acco ding o he se ice-based policy VG*VU. No ice ha , o di e en mul i-mode e minal
a ailabili ies and di e en numbe o use s eques ing se ice, h oughpu deg ada ion exhibi s di e en
ends. In pa icula , o 200 oice and 200 www use s, no big di e ences a e no iced, meaning ha , in
his case, GERAN is able o manage oice and single-mode e minals wi h ease. Ce ainly, he a e age
imeslo u ilisa ion ac o in GERAN e eals an occupa ion o esou ces below 70%. The inc ease o
use s eques ing o be se ed is ansla ed in o a bigge deg ada ion in e ms o h oughpu as mul i-
mode e minal a ailabili y dec eases. While o 400 oice and 400 www use s deg ada ion s a s o ge
no iceable o mul i-mode a ailabili ies o 25% and 50%, o 600 oice and 600 www use s his
deg ada ion is al eady pe cep ible a 75% o mul i-mode a ailabili y.
0
5
10
15
20
25
30
35
100 75 50 25
Mul i-mode Te minal A ailabili y (%)
Th oughpu Deg ada ion (%)
VU=600;WU=600
VU=400;WU=400
VU=200;WU=200
Figu e 51.- Uplink h oughpu deg ada ion due o mul i-mode
e minals
0
5
10
15
20
25
75 50 25
Mul i-mode Te minal A ailabili y (%)
A e age Packe Delay (s)
VU=400;WU=400
VU=200;WU=200
Figu e 52.- Uplink a e age packe delay o in e ac i e use s.
Packe delay s a is ics o in e ac i e use s also e eal deg ada ion when conside ing a scena io wi h
mixed mul i-mode and single-mode e minals. This deg ada ion impac s di ec ly in he pe cei ed QoS
by he da a use and i is he e o e impo an o keep i s alue as low as possible. Figu e 50 shows he
uplink a e age packe delay o in e ac i e use s being se ed h ough GERAN o di e en mixings o
mul i-mode e minals and se ice-classes. No ice he inc easing packe delay when mul i-mode e minal
a ailabili y dec eases o 400 oice and 400 www use s. As o 200 oice and 200 www use s, he
a e age packe delay emains almos cons an wi h mul i-mode e minal a ailabili y and also a an
accep able le el, he e o e exhibi ing no deg ada ion in his sense.
The sugges ed app oach o o e come he delay deg ada ion esides in he in oduc ion o dedica ed
EGPRS imeslo s o in e ac i e use s. In his way, some esou ces may be ese ed o in e ac i e
use s and he e o e packe delay pe o mance can be imp o ed wi h espec o no ha ing any
ese a ion scheme. As o h oughpu pe o mance, i can be o eseen ha con ibu ion o oice use s
o he o al agg ega e h oughpu migh diminish. This educ ion o oice h oughpu may be
compensa ed, gi en ce ain condi ions, by in e ac i e use s alloca ed o he ese ed slo s. This en ails
a ade-o be ween he numbe and he applicabili y o ese ed esou ces (slo s) o in e ac i e use s
in GERAN and o he pa ame e s, such as o e ed load and mul i-mode e minal a ailabili y, as i is
obse ed in he ollowing able o di e en numbe s o oice use s (VU) and web use s (WU).
Final Repo Page 35
Table 8.- EGPRS Slo Rese a ion Gain (%).
jj
u = (VU , WU )
i

Mul i-mode Te minal A ailabili y (%)
VUj
WUj
25
50
75
100
200
200
0.00
0.00
0.72
0.00
400
1.14
0.56
1.13
0.55
600
0.00
0.46
-0.46
-0.91
400
200
-3.88
-4.76
-6.07
-7.83
400
3.49
-0.42
-4.40
-5.10
600
8.71
2.21
-1.39
-4.71
600
200
1.38
-2.10
-2.75
-1.48
400
13.22
7.00
-0.35
-4.23
600
13.87
12.23
3.45
-1.16
In summa y, he ob ained esul s indica e deg ada ion in e ms o h oughpu in oduced by he limi ed
ope a ion o single-mode e minals. By conside ing a ese a ion scheme in GERAN o in e ac i e
use s, we can imp o e he a e age packe delay o such use s. While o a high mul i-mode e minal
a ailabili y esul s indica ed ha he ese a ion scheme was no necessa y, o lowe mul i-mode
e minal a ailabili ies his scheme imp o ed bo h agg ega ed h oughpu and packe delay igu es,
pa icula ly o high numbe o in e ac i e use s.
Re e ences
[1] O. Sallen (edi o ) e al. “Fi s epo on he e alua ion o RRM/CRRM algo i hms”, Deli e able D11 o
he EVEREST IST-2002-001858 p ojec , No embe , 2004.
A ailable a h p://www.e e es -is .upc.es/.
[2] EVEREST IST-2002-001858, D15, “Repo on he e alua ion o RRM/CRRM algo i hms”
[3] Kim, D., e al, “Pilo Powe Con ol and Se ice Co e age Suppo in CDMA Mobile Sys ems”,
P oceedings o he 49 h IEEE Vehicula Technology Con e ence, pp 1464-1468, 1999.
[4] Valkealah i K., Höglund A., Pa kkinen J., Flanagan A., “WCDMA common pilo powe con ol wi h cos
unc ion minimiza ion”, P oc. IEEE Vehicula Technology Con e ence, Vancou e , Canada, Oc obe
2002.
[5] BIANCHI, Giuseppe. “Pe o mance Analysis o he IEEE 802.11 Dis ibu ed coo dina ion Func ion”,
IEEE Jou nal on selec ed a eas in communica ions, ol. 18, No. 3, Ma ch 2000.
Final Repo Page 36
4.3 EVEREST Tes bed
The EVEREST demons a o ( es bed) has been used o alida e in he labo a o y he bene i s o he
p oposed RRM/CRRM algo i hms and QoS Managemen echniques.
The EVEREST demons a o is a eal- ime ope a ion HW/SW pla o m including mul imedia e minals,
GERAN/UMTS/WLAN elemen s and IP based Co e Ne wo k, able o suppo eal- ime mul imedia calls.
The pu pose o he GERAN/UMTS/WLAN emula o s is o ep oduce he eal beha iou o a use unde
es wi h mo e accu acy han wha would be ob ained om he simula o s (mo e sui ed o global sys ems
pe o mance analysis), bu wi h less implemen a ion complexi y han in a eal sys em. The philosophy
used is o implemen a “subse ” o unc ionali y, app op ia e o emula ion o he c i ical aspec s ela ed o
RRM/CRRM and QoS issues, a he han o ealise a "one by one" ep esen a ion o hei specs. This
app oach leads o a ligh e implemen a ion, sui able o assess new RRM/CRRM algo i hms easily, no
conside ing pa ame e s ha do no ha e a ele an impac on hem. Tha is, he EVEREST demons a o
is unde s ood as a lexible SW/HW pla o m ha allows he expe imen al e alua ion o new Radio
Resou ce Managemen algo i hms unde con olled bu ealis ic condi ions
Tes bed A chi ec u e O e iew
F om a concep ual iew poin he es bed a chi ec u e can be di ided in wo main blocs: he
He e ogeneous Radio Access pa and he Co e Ne wo k.
Radio Access Domain
The unc ional a chi ec u e o he p oposed app oach o emula e a he e ogeneous RAN is shown in he
ollowing igu e.
Figu e 51.- He e ogeneous RAN Emula ion app oach
Final Repo Page 43
• Hie a chical Cell S uc u es (HCS).- The load g adien amewo k has been p oposed as a candida e
o s a e whe he o no is possible o ope a e di e en HCS laye s wi h he same ca ie equency
depending on spa ial a ic dis ibu ions.
• T anspo Channel Type Swi ching.- No ice ha In he case o a discon inuous da a ans e , i is e y
impo an o be able o op imize he usage o dedica ed anspo channel (DCH), p e en ing
channeliza ion code sho age in he downlink wi hou deg ading he end- o-end quali y o se ice
expe ienced by he use in an app eciable manne . The achie ed esul s show ha he UTRAN
capaci y is highe when he anspo channel ype swi ching algo i hm is ac i e.
• High Speed Downlink Packe Access (HSDPA).- Ad anced schedule s, ha combine bo h channel
in o ma ion and se ice equi emen s, ha e been e alua ed and he esul s we e p esen ed in e ms
o sa is ied use s and cell h oughpu bo h o indoo and ehicula en i onmen s. Compa ing he
di e en schedule s p oposed in he li e a u e, i was ob ained ha P io i y S anda d schedule
p o ides he be e esul s bo h o QoS bi a e and sa is ied use s.
Mo eo e , HSDPA is capable o suppo ing a ious Qos le els. In ac , he HSDPA speci ica ion
suppo s up o 16 p io i y le els and 8 queues pe use . Then, when he e ogeneous a ic coexis s,
an inno a i e HSDPA packe schedule ha disc imina es QoS classes was p oposed o p o ide
be e QoS suppo han hose ha conside only he channel condi ion The simula ion esul s u he
showed ha he p oposed schedule is obus o changes in he a ic cons i u ion, and suppo s call
a i al a es highe han he ob ained wi h con en ional schemes.
• RAN Sha ing.- Sha ing spec um can be e y a ac i e, o example, in u al a eas UMTS co e age
can be o e ed wi h much lowe in es men cos s, bu also in u ban a eas and ho spo a eas capaci y
gains can be achie ed. Ce ainly, he e is a capaci y gain due o he inc eased unking e iciency as
channels a e pooled oge he be ween he ope a o s. When analysing di e en app oaches, he
ob ained esul s show ha he so called Mul i-cell algo i hm p o ides he bes G ade o Se ice,
because i p o ides highe admission p obabili y han he Hal powe algo i hm and main ains he
d opping p obabili y in a low alue.
• Loca ion Awa e Resou ce Rese a ion.- Recen de elopmen s in he posi ioning echnology in he
con ex o WCDMA sys ems p o ide s ong assu ance ha accu a e posi ion measu emen s. This can
be exploi ed o de elop mo e ad anced RRM (Radio Resou ce Managemen ) s a egies ha inc ease
he sys em e iciency. In pa icula an inno a i e ese a ion algo i hm, whose main objec i e is o
assu e se ice o business use s mo ing along a main oad while a he same keeping he se ice o
consume use s a a sa is ac o y le el, has been p oposed. The ese a ion o ce ain esou ces o
hando e use s educes he numbe o d opped connec ions a expense o ce ain inc ease in he
blocking p obabili y o new connec ion eques s. The p oposed algo i hm akes ad an age o he
p edic abili y in he mo emen o use s along a main oad in o de o de e mine he mos adequa e
ins an o ime when he esou ce ese a ion o hando e use s should be made
CRRM issues
Common Radio Resou ce Managemen (CRRM) e e s o he se o unc ions ha a e de o ed o ensu e
an e icien use o he a ailable adio esou ces in he e ogeneous ne wo ks scena ios by means o a
p ope coo dina ion be ween he di e en adio access ne wo ks. Then:
• Wi h espec o Common Radio Resou ce Managemen , a i s con ibu ion o he p ojec has been
he ou line o a unc ional model o ha ing a common managemen o he pool o adio esou ces in
he e ogeneous scena ios
• CRRM pe o mance e alua ion epo ed he e assume a unc ionali y spli in which he CRRM akes
cha ge o he RAT selec ion p ocedu es, including bo h ini ial RAT selec ion and e ical hando e ,
while he RAT-speci ic RRM algo i hms, like admission con ol, conges ion con ol o packe
scheduling a e execu ed locally a he RRM en i ies.
• Wi h espec o RAT selec ion a amewo k o de ising policy based s a egies has been p oposed.
They conside bo h se ice-based c i e ia (e.g. alloca e oice in GERAN and in e ac i e se ices in

Final Repo Page 44
UTRAN) and adio ne wo k-based c i e ia (e.g. alloca e indoo use s in GERAN and in e ac i e
se ices in UTRAN) in he o m o he so-called complex policies.
• Fo se ice-based RAT selec ion policies and conside ing scena ios wi h a mix o indoo and ou doo
use s wi h di e en se ices, up o medium oice loads (i.e. 200 use s) no ele an di e ences
be ween he policies a e obse ed. Howe e , when he load inc eases as i s op ion is be e o
alloca e he oice use s in GERAN and he in e ac i e (non- eal ime) use s in UTRAN. I he e is no
mo e esou ces a ailable in GERAN hen apply a policy ha alloca es he indoo use in GERAN and
he ou doo use in UTRAN, mo ing oice ou doo use om GERAN o UTRAN.
• Wi h espec o e ical Hando e (VHO), wo di e en app oaches, deno ed as loose and igh in e -
wo king wi h ho izon al HO, ha e been analysed. I has been shown ha he igh app oach allows a
be e ul ilmen o he RAT selec ion policy a expense o an inc emen in he signalling.
• Load balancing (LB) is ano he possible guiding p inciple o esou ce alloca ion in which he RAT
selec ion policy will dis ibu e he load among all esou ces as e enly as possible. When conside ing
LB wi hou VHO p ocedu es, he ob ained esul s e ealed a igh dependency be ween he sui abili y
o load balancing RAT selec ion and se ice-class mixing. Howe e , he in oduc ion o VHO
capabili ies allows imp o ing he lexibili y achie ing lowe d opping a ios han wi h he se ice-based
policies.
• Fo Pa h Loss - Based RAT Selec ion, he unde lying idea o he CRRM app oach de eloped is o
ake ad an age o he co e age o e lap ha se e al RANs using di e en access echnologies may
p o ide in a ce ain se ice a ea in o de o imp o e he o e all in e e ence pa e n gene a ed in he
scena io o he CDMA-based sys ems and, consequen ly, o imp o e he capaci y o he o e all
he e ogeneous scena io. The abo e concep has been deno ed as Ne wo k-Con olled Cell-B ea hing
(NCCB).
When conside ing a single oice se ice he ob ained esul s e eal ha a signi ican imp o emen in
e ms o capaci y o bo h uplink and downlink is achie ed. When conside ing a mix be ween oice
and www use s, he bes pe o mance is achie ed wi h he so-called NCCB_ oice, co esponding o
applying he NCCB s a egy only o oice use s and alloca ing www use s in UTRAN acco ding o he
se ice-based c i e ion.
• When he impac o Mul i-Mode e minals on CRRM pe o mance is analysed he ob ained esul s
show ha deg ada ion in e ms o h oughpu appea s bu , conside ing a ese a ion scheme in
GERAN o in e ac i e single-mode use s, he a e age packe delay o such use s can be
signi ican ly imp o ed.
Final Repo Page 45
Annex 1.- Lis o Publica ions
The scope o his sub-sec ion is o p esen he dissemina ion ac i i ies ca ied ou by he EVEREST
p ojec in scien i ic o a. In o al he pa ne s ha e published 32 scien i ic pape s and con ibu ed wi h 10
p esen a ions in wo kshops. The publica ions a e o de ed acco ding he media used: scien i ic jou nals
and magazines, in e na ional con e ences, wo kshops.
Scien i ic Jou nals/Magazines
• Nima Na isi, A.Hamid Agha ami, “IP hando e delay o hos -based mic omobili y p o ocols” IEE
Elec onics le e s, Elec onics Le e s; Volume 41, Issue 3, 3 Feb 2005 Page(s):160 - 161.
• Vasilis F ide ikos, Lin Wang, Mikio Iwamu a, A. Hamid Agh ami, “Colo -awa e Powe and Ra e
Adap a ion in IP-based CDMA Radio Access Ne wo ks”, Jou nal o Wi eless Communica ions and
Mobile Compu ing, Volume 5 , Issue 4, June 2005 pp 407-420
• L. Wang, A.H. Agh ami, W. G. Chambe s, J. Pe ez Rome o, O. Sallen , “Pe o mance Analysis o an
In eg a ed CS/PS Se ices CDMA Sys em”, IEEE T ansac ions on Vehicula Technology, Volume
54, Issue 4, July 2005 Page(s):1488 – 1499
Con e ences:
The ollowing pape s we e p esen ed in In e na ional Con e ences:
• F. Adelan ado, O. Sallen , J. Pé ez-Rome o, R. Agus í ; “Impac o T a ic Ho spo s in 3G W-CDMA
Ne wo ks ” he IEEE Vehicula Technology Con e ence 2004 Sp ing. Milan (I aly), 17-19 May, 2004
• F. Casade all; P. Ka lsson; O. Sallen ; H. Gonzalez; A. Ba ba esi; M.Dohle ; M.Dinis; “ EVEREST
O e iew ”, IST Summi 2004. Lyon (F ance), 28-30 June, 2004.
• N. Na isi; L. Wang, H.Agh ami, R. Fe us, A. Gelonch, J. Pe ez-Rome o, O. Sallen , R. Agus i;
“Ex ending QoS Policy-based mechanisms o B3G Mobile Access Ne wo ks ”; IST Summi 2004.
Lyon (F ance), 28-30 June, 2004.
• J. Pe ez-Rome o, O. Sallen , D. Ruiz, R. Agus i; “Admission Con ol Algo i hm o Manage High Bi
Ra e S a ic Use s in W-CDMA”; IST Summi 2004. Lyon (F ance), 28-30 June 2004.
• Lin Wang, Vasilis F ide ikos, A.Hamid Agh ami, Misha Dohle ; “Colo -awa e Link Adap a ion o
Di Se o e CDMA Sys ems”; IST Summi 2004. Lyon (F ance), 28-30 June 2004.
• J. Pé ez-Rome o, O. Sallen , R. Agus í; “Impac o Indoo T a ic on W-CDMA Capaci y "; PMIRC04.
Ba celona (Spain), 5-8 Sep embe 2004.
• J. Pe ez-Rome o, O. Sallen , R. Agus i, “On he Capaci y Deg ada ion in Uplink/Downlink W-CDMA
Due To Indoo T a ic”; IEEE Vehicula Technology Con e ence 2004 F all. Los Angeles (USA), 26-29
Sep embe 2004.
• J.L. Valenzuela, A. Monleon, I. San Es eban, M. Po oles, O. Sallen ; “A Hie a chical Token Bucke
Algo i hm o Enhance QoS in IEEE 802.11: P oposal, Implemen a ion and E alua ion”, IEEE
Vehicula Technology Con e ence 2004 Fall. Los Angeles (USA), 26-29 Sep embe 2004.
• J. Pé ez-Rome o, O. Sallen , R. Agus í , “ A No el App oach o Mul icell Load Con ol in W-CDMA ”;
5 h In e na ional Con e ence on 3G Mobile Communica ion Technologies (3G 2004), London (UK),
18-20 Oc obe , 2004.
• J. Pé ez-Rome o, O. Sallen , R. Agus í, “On E alua ing Beyond 3G Radio Access Ne wo ks:
A chi ec u es, App oaches and Tools", IEEE Vehicula Technology Con e ence 2005 –Sp ing,
S ockholm, Sweden, May 30 - June 1, 2005
Final Repo Page 46
• J. Sánchez-González, O. Sallen , J. Pé ez-Rome o, R. Agus í. “An Admission Con ol Algo i hm o
WCDMA Conside ing Mobile Speed and Se ice Cha ac e is ics", IEEE Vehicula Technology
Con e ence 2005 –Sp ing, S ockholm, Sweden, May 30 - June 1, 2005
• R. Fe ús, A. Gelonch, J. Pé ez, O. Sallen , N. Na isi, M. Dohle , “A easible app oach o E2E QoS
managemen in coo dina ed he e ogeneous adio access ne wo ks”, 24 h IEEE In e na ional
Pe o mance, Compu ing, and Communica ions Con e ence, 2005. (IPCCC 2005); 7-9 Ap il 2005.
• M. Iwamu a, V. F ide ikos, L. Wang, H. Agh ami; “Biased Adap i e Modula ion/Coding o P o ide
VoIP QoS o e HSDPA”, 2005 IST Mobile Summi D esden (Ge many) 19-22 June, 2005.
• Ba ba esi, S. Ba be is, P. Go ia; “Admission Con ol Policy o WLAN Sys ems based on he
"Capaci y Region", 2005 IST Mobile Summi D esden (Ge many) 19-22 June, 2005.
• R. Fe ús, A. Gelonch, F. Casade all, X. Re és, N. Na isi, M. Dohle ; “Tes bed o CRRM and E2E
QoS Managemen E alua ion in B3G Ne wo ks”; 2005 IST Mobile Summi D esden (Ge many) 19-22
June, 2005.
• R. Fe ús, A. Gelonch, O. Sallen , J. Pé ez-Rome o, N. Na isi, M. Dohle ; “Ve ical Hando e Suppo
in Coo dina ed He e ogeneous Radio Access Ne wo ks”; 2005 IST Mobile Summi D esden
(Ge many) 19-22 June, 2005.
• X. Gelabe , J. Pé ez-Rome o, O. Sallen , R. Agus í, F. Casade all; “Radio Resou ce Managemen in
He e ogeneous Ne wo ks”; 3 d In e na ional Wo king Con e ence on Pe o mance Modeling and
E alua ion o He e ogeneous Ne wo ks (He -Ne s 05), Ilkley, Wes Yo kshi e (UK), 18 h-20 h July
2005.
• Lin Wang, Mikio Iwamu a, Vasilis F ide ikos, Nima Na isi and A. Hamid Agh ami; “Di Se Awa e
Link Adap a ion o CDMA Radio Sys em”; In . Con . on Quali y o Se ice in He e ogeneous
Wi eless/Wi ed Ne wo ks; O lando-USA; 22-24 Augus 2005
• X. Gelabe , J.Pe ez-Rome o,O iol Sallen , R. Agus í; “On he Impac o Mul imode Te minals in
He e ogeneous Wi eless Access Ne wo ks”; 2nd In e na ional Symposium on Wi eless
Communica ion Sys ems 2005; Siena- I aly; 5-7 Sep embe 2005.
• N. Na isi, R. Fe ús, A. Gelonch, O. Sallen , J. Pé ez-Rome o, L. Wang, M. Dohle , H. Agh ami; “QoS
Awa e Pa h Selec ion in a B3G Sys em”, PIMRC’05 Be lin (Ge many) ; 11-14- Sep embe 2005
• J. Pé ez-Rome o, O. Sallen , R. Agus í, P. Ka lsson, A. Ba ba esi, L. Wang, F. Casade all, M.
Dohle , H. González, F. Cab al-Pin o; “Common Radio Resou ce Managemen : Func ional Models
and Implemen a ion Requi emen s”, PIMRC’05 Be lin (Ge many) 11-14- Sep embe , 2005
• J. Sánchez-González, O. Sallen , J. Pe ez-Rome o, R. Agus i; “On Managing Dynamic T a ic
Ho spo s in WCDMA Ne wo ks”; PIMRC’05 Be lin- (Ge many) 11-14- Sep embe
• P. Emanuelsson, e al e s; “Admission Con ol a UMTS Spec um Sha ing” ; In . Con . on So wa e ,
Telecommunica ion and Compu e Ne wo k; Spli (C o ia), 13-17 Sep embe 205 .
• J. Pé ez-Rome o, O. Sallen , R. Agus í , “Policy-based Ini ial RAT Selec ion algo i hms in
He e ogeneous Ne wo ks” 7 h In . Con . on Mobile and Wi eless Communica ions Ne wo ks 05;
Ma akesh – (Mo oco) ; 19-21 Sep embe 2005;
• X. Gelabe , J. Pé ez-Rome o, O. Sallen , R. Agus í, “ On he Sui abili y o Load Balancing P inciples
in He e ogeneous Wi eless Access Ne wo ks” ; 1s In . In e na ional Wi eless Summi ; IWS 05.
Aalbo g (Denma k); 17-22 Sep embe 2005
• J. Majkowski, F. Casade all; “Admission Con ol in IEEE 802.11e EDCA” ; 1s In . In e na ional
Wi eless Summi ; IWS 05. Aalbo g (Denma k); 17-22 Sep embe 2005)
Final Repo Page 47
• F. Casade all, P. Emmanuelsson, R. Fe us, A. Gelonch, P. C. Ka lsson, N. Na isi , J. Pe ez-
Rome o, O. Sallen , L. Wang, “ EVEREST – A Policy Based QoS A chi ec u e o He e ogeneous
Wi eless Sys ems”, IST4Bal Con e ence “E ol ing Mobile Eu ope “ Vilnius (Li huania) 24 h and 25 h
Oc obe 2005.
• Lin Wang, Vasilis F ide ikos, Mikio Iwamu a, Nima Na isi, A.Hamid Agh ami, Misha Dohle , “A Link
Adap a ion Scheme o Suppo Rela i e Di Se QoS o e CDMA Radio Sys ems”, IEE 3G and
Beyond Con e ence 2005; London (UK) 7-9 No embe 2005.
• F. Adelan ado, F. Casade all, “Dynamic Common Pilo Powe Managemen in a Real Ho Spo ”;
IEEE GLOBECOM 2005 S . Louis-Missou i (USA), No embe 28 h –Decembe 2 h 2005.
Wo kshops
The ollowing p esen a ions we e done in he Wo kshops
• R. Fe us, A. Gelonch, F. Casade all, J. Pé ez-Rome o, O. Sallen , R. Agus í, Nima Na isi , Lin
Wang, M. Dohle , Hamid Agh ami, Pe e Ka lsson; “ End- o-End QoS A chi ec u e o Mul i-Domain
and Wi eless He e ogeneous Access Ne wo ks: The EVEREST app oach”. WWRF wo kshop in
Beijing (China), Feb ua y 2004.
• F. Casade all “EVEREST: En isaged Scena ios”. Clus e SB3G scena ios wo kshop, B ussels Ma ch
2004.
• J. Pé ez-Rome o; “En isaged RRM: Loca ion Aided App oaches in EVEREST”. Clus e LOBSTER
wo kshop, B ussels Ma ch 2004.
• J. Pé ez-Rome o, O. Sallen , D. Ruiz, R. Agus í, “PLEBAC: A New Downlink Admission Con ol o
UTRAN-FDD”, Mobile Venue'04 Wo kshop. A hens (G eece) 27-28 May 2004.
• R. Fe ús, A. Gelonch, F. Casade all, X. Re és, N. Na isi, “ An E2E QoS Tes bed o B3G ne wo ks ”,
E2R wo kshop, Ba celona (Spain), 5 Sep embe 2004.
• N. Na isi, L. Wang, M. Dohle , H. Agh ami, R. Fe ús, A. Gelonch, O. Sallen , J. Pé ez-Rome o, F.
Casade all, R. Agus í , “End- o-end QoS a chi ec u e and es bed o B3G sys ems: he E e es
app oach”, Wo kshop WRRF – Canada, No embe 2004.
• J. Pé ez-Rome o; "Radio Resou ce Managemen S a egies in he amewo k o he EVEREST
p ojec ", Clus e SB3G: Wo kshop on Radio Resou ce Managemen , B ussels 9 h Ma ch 2005.
• R. Fe ús, N. Na isi , A. Gelonch, O. Sallen , J. Pé ez-Rome o; “EVEREST QoS A chi ec u e”,
Wo kshop T ends in Radio Resou ce Managemen , Ba celona, 16 No embe 2005.
• O. Sallen , F. Casade all, P. Ka lsson, A. Ba ba esi, L. Wang, H. González, F. Cab al-Pin o
“EVEREST’S Ac i i ies on RRM/CRRM: an o e iew”, Wo kshop T ends in Radio Resou ce
Managemen , Ba celona, 16 No embe 2005.
• R. Fe ús, A. Gelonch, X. Re és, N. Na isi , “EVEREST Tes bed”, Wo kshop T ends in Radio
Resou ce Managemen , Ba celona, 16No embe 2005.
Final Repo Page 48
Accep ed bu no al eady published Pape s
Six pape s ha e been accep ed o p esen a ion a he ollowing in e na ional con e ences.
― R. Fe ús, A. Gelonch, F. Casade all, X. Re és, Nima Na isi, “EVEREST Tes bed: QoS Managemen
E alua ion in B3G Ne wo ks”, 2nd In e na ional IEEE/C ea e-Ne Con e ence on Tes beds and
Resea ch In as uc u es o he De elopmen o Ne wo ks and Communi ies (TRIDENCOM 2006) o
be held in Ba celona (Spain); 1-3 Ma ch 2006.
― J. Pé ez-Rome o, O. Sallen , R. Agus í, L. Wang, H. Agha mi, “Ne wo k-Con olled Cell-B ea hing o
Capaci y Imp o emen in He e ogeneous CDMA/TDMA Scena ios”, IEEE Wi eless Communica ions
and Ne wo king Con e ence WCNC’06 o be held in Las Vegas (USA) 3-6 Ap il 2006
― 4 pape s a IEEE Vehicula Technology Con e ence (VTC’06 Sp ing) o be held in Melbou ne
(Aus alia) 7-10 May 2006.
 J. Pé ez-Rome o, O. Sallen , R. Agus í, L Wang, H. Agha mi, “A No el Algo i hm o Radio
Access Technology Selec ion in He e ogeneous B3G ne wo ks” .
 J. Sánchez-González, J. Pé ez-Rome o, O. Sallen “A Loca ion-Awa e Resou ce Rese a ion
Algo i hm wi h Use Class Di e en ia ion in WCDMA”.
 F. Adelan ado, O. Sallen , J. Pé ez-Rome o, “On Modelling Spa ial T a ic and Se ice Non-
Uni o mi ies in WCDMA Re e se Link”.
 J. Pé ez-Rome o, O. Sallen , R. Agus í, “A No el F amewo k o Robus WCDMA Planning
unde Changing Spa ial T a ic Dis ibu ions”

Final Repo Page 49
Annex 2.- Rela ion wi h he S anda ds
As a ma e o ac , he EVEREST p ojec add esses opics such as Radio Resou ce Managemen
(RRM), Common Radio Resou ce Managemen (CRRM) and QoS Managemen ha a e ypically ou side
he scope o he S anda diza ion o a, being mainly implemen a ion dependen issues. Howe e , as i is
explained below, he EVEREST conso ium has p epa ed some con ibu ions e lec ing he main
ou comes o he p ojec . These con ibu ions ha e al eady been p esen ed o he TSG RAN WG2 (Radio
Laye 2 speci ica ion and Radio Laye 3 RRC speci ica ion) o he 3GPP o a e pending o submission.
Impac o Cu en o Eme ging S anda ds on P ojec ’s Wo k
The p ojec has wo ked in close con ac wi h he ele an s anda ds bodies cu en ly add essing he
aspec s ea ed by EVEREST. The e o e he p ojec acqui ed on a egula basis use ul documen a ion
and in o ma ion om s anda ds bodies (3GPP, IETF mainly), ega ding he Radio Resou ce managemen
issues and he QoS handling. This has ensu ed ha he p ojec e olu ion has been always aligned wi h
he guidelines p o ided by he s anda disa ion amewo k.
Impac o Ac ual o Expec ed Con ibu ions on he Wo k o S anda d Bodies
The ollowing pape s and con ibu ions ha e been p esen ed a di e en s anda disa ion g oups:
• “UTRAN as a Pa o a Mul i-access Ne wo k”, doc: REV-WS008: Sou ce: TeliaSone a, 3GPP
RAN Fu u e E olu ion Wo kshop, 2-3 No embe 2004, To on o, Canada
• “Radio Resou ce Managemen A chi ec u e conside a ions o He e ogeneous Wi eless Sys ems
- The EVEREST conso ium” doc numbe R3-051250, Sou ce: TeliaSone a, 3GPP RAN WG3
mee ing 49 in Seoul (Ko ea), 7-11 No embe 2005.
Mo eo e 3GPP TSG RAN WG2 is cu en ly de eloping a Technical Repo TR 25.922: "Examples o
RRM s a egies o HSDPA. EVEREST p ojec has p epa ed one con ibu ion, e lec ing he main
ou comes o he p ojec in e ms o schedule s o HSDPA, a he 3GPP RAN WG2 mee ing o be held in
Den e (USA), 13-17 Feb ua y 20062.