Angel Lozano!

"" " "

Towards 5G
A Research Perspective
 Angel Lozano!
"" " "
Towards 5G
A Research Perspective
5G - Wikipedia, the free encyclopedia http://en.wikipedia.org/wiki/5G

5G
From Wikipedia,
5G - Wikipedia, the free encyclopedia
the free encyclopedia http://en.wikipedia.org/wiki/5G

5G (5th generation mobile networks or 5th generation wireless systems) is a term used in some
Based onpapers
research the above
and observations, somethe
projects to denote sources suggest
next major that of
phase a new
mobilegeneration of 5G standards
telecommunications may to

standards
[4][5]
beyond the current
be introduced 4G/IMT-Advanced
approximately standards.
in the early 2020s. 5G However,
is also referred to transnational
still no as beyond 2020
5G mobile
development
communications technologies.
projects have officially 5G doesand
been launched, not there
describe anyaparticular
is still large extent specification
of debate onin what
any official
5G is exactly
about. Prior to 2012, some industry representatives have expressed skepticism towards 5G[6] but the
document published by any telecommunication standardization body.
trends clearly changed since 2012.[citation needed]
Although updated standards that define capabilities beyond those defined in the current 4G standards
are
Newunder consideration,
mobile generations those new capabilities
are typically assigned are
newstill being grouped
frequency bands and under thespectral
wider current 4G
bandwidth
standards.
per frequency channel (1G up to 30 kHz, 2G up to 200 kHz, 3G up to 20 MHz, and 4G up to
Outline

Introduction
5 Research Reflections (Inspired by 5G)
Final Remarks
1G
(1981)
2G
(1992)
3G
(2002)
4G
(2011)

Analog Digital Multimedia Citius

Altius
Fortius
 !"#$%&'"()*+,%&+-(+.(/"01-+#+2&"3(

LTE
4G
Relative Subscriptions

3G
UMTS/HSPA

2G
GSM/EDGE

1990 2000 2010 2020 2030

Mobile Broadband Explosion, Rysavy Research

2012 white paper
2011 45(
5G
(2020) .!.!?!
MASSIVE INCREASE IN WIRELESS TRAFFIC
AND IN NUMBER OF CONNECTED DEVICES
 Traffic

1000
Smartphones
The Cloud
M2M Devices

100
Vast Range of Bit Rate
& Latency Requirements

www.qualcomm.com/1000x
www.ericsson.com/res/docs/whitepapers/wp-5g.pdf
Can we improve the utilization of radio
resources by 1000?!

bits/s bits/s node

2 = 2 Hz
Km Hz node Km
Massive
MIMO Extreme
Densification New Spectrum
(Including D2D) (including mm-Wave)

Interference Management
New Signal Waveforms
Full-Duplex Relays
Data Caching
Etc
International activities on 5G getting momentum
EU
Framework Program 7, e.g. METIS and 5GNow projects
5G PPP in Horizon 2020
UK 5G Innovation Centre (5GIC) at University of Surrey
US
Intel Strategic Research Alliance (ISRA)
NYU Wireless Research Center
China
863 Research Program
Future Forum
IMT-2020 (5G) Promotion Group
Japan 2020 and Beyond Ad-Hoc Group under ARIBs Advanced Wireless
Communications Study Committee
Korea 5G Forum as PPP
Taiwan Ministry of Economic Affairs, National Science Council
Russia 5GRUS by Russias Icom-Invest
NGMN White paper on future requirements
Company internal research
Source: 5G Infrastructure Association.
 5G Infrastructure PPP: The next generation of
communication networks will be Made in EU.
b Building the foundations of the next
decade communications networks
b Addressing the future connectivity
needs in key societal and economic
domains
b Boosting European industrial

Thinkstock / Getty
leadership in telecommunications

What is the challenge?

for leveraging 5G research to improve competitiveness and
IEEEThe Communications Magazine,
European telecommunications industrySpecial
has beenIssue on with
innovation 5Gthe Wireless Communication
ultimate aim of stimulating economic
Systems: Prospects and Challenges, Feb. 2014
historically in the forefront of the global competition since growth and more job creation in other industrial sectors.
the early days of the GSM technology, and still represents The 5G PPP brings a long term commitment from both the
approximately 40% of the worldwide market of nearly 200 private and the public actors to invest in achieving these
IEEE
billionSignal Processing
in 2012 in Magazine,
terms of network Special
infrastructure supply. Issue onand
objectives the PPP willProcessing
Signal for the the
play a key role in formulating 5G
The challenge for the 5G Public-Private Partnership (5G PPP) research and innovation priorities to be supported in Horizon
Revolution, Nov. 2014
is to secure Europes leadership in the particular areas where 2020.
Europe is strong or where there is potential for creating new
IEEE JSAC,
markets such as Special
smart cities, Issue
e-health, on 5G
intelligent Wireless
transport, WhatCommunication Systems,
results and benefits do we expect?Dec. 2014
education or entertainment & media.
The 5G PPP will deliver solutions, architectures, technologies The 5G PPP will result in a stronger, more competitive and
and standards for the ubiquitous next generation more innovative telecommunications industry through
communication infrastructures of the coming decade. It a closer cooperation between all key actors including the
will provide such advancements as 1 000 times increase research and academic institutions and the end-users. The
in wireless capacity serving over 7 billion people (while ultimate benefits are to provide solutions to important
5 Research Reflections
(Inspired by 5G)
Beware of hype
Interference is not the problem
Embrace messiness
Virtualization is coming to town
Cutting the wireless wire
 Beware of Hype

It is a good morning exercise for a researcher to discard

a pet hypothesis every day before breakfast
K. Lorenz
 Beware of Hype
 Beware of Hype
 Beware of Hype

The choice of modeling assumptions should illuminate the problem and

maximize insight, avoiding nuisances that would otherwise be
distracting

Only distractions that have no significant impact on the result can be

avoided!

Everything should be made as simple as possible, but not simpler

Albert Einstein
 Beware of Hype

Tractability Relevance
 Beware of Hype

Massive MIMO?
Mm-Wave Communication?
 Interference is not the Problem
 for!"#$%&'$!"#$%
interference alignment without explicitly
&'&() &)(%(*+% '* ,-()-$(. identifying gree-of-freedom
,/ +#(the+#(,)0
underlying structures
H;B<>;N>
M>;D>>H gain ;D?is
for
EL:K either oN
interferenc
K@<;=>:A
F>;9@F
cooperation.
ence Alignment
8.%< $8 hold
1+)*F the promise
'$'() %&1+#; of an
&) BCincreased
1& #&* number
1+)(%*/F of
$''/F9 G<(
output). Altogether then, IA cS
'*1 23))(*+40
numerical 5*,.*
insight of +(2#*$63(%
this paper is /,)that734+$82(44
such an 9:9;
approach over-the-air
<734+$&4(
appears .>J;<?H
to be bit
0?.
capableor
%%%A>BHKtwo
[P
of cooperat
M>P?HK5
@=;;AE
achieving
spatial$*&3+
degrees of freedom
?"@734+$&4(
')(%&1() +8 and,
,3+&3+= with that,
%&-'.*(1
2,,&()'+$,* DF of $*substantially
8&/,+#; $ 8(=.(#%(
.$)(4(%% higher&7 :*
*(+.,)5%> %&#,(5
interference in exchange for
spectral
MIMO
efficiencies.
&'*+-+6$*+&# Interference
Most results
')&D/(-89
cellular networks ?"@ is
available
.#1() not
in regimes (7()?(% to the
date,
'$)*+$/
where linear Problem
however,
#(*2&)3 have bit.
%&&)1+#$*+&#
beamforming *>:BP=MBF>K
$; I:><G<KK
dominates asymptotic decompos J??@>AB;<C
@AMMH
1(*%( *(+.,)5% .#()( $*+()/()(*2( '% +#( 5(0 2'&'2$+08
been obtained%$# &.*'()7&)-
4$7$+$*? /'2+,)@ I.734+$82(44
in simplified INTRODUCTION
BCsettings
.#1() featuring
7.//
2,,&()'+$,* #(*2&)32'*a small number$7&),-(
%&&)1+#$*+&#
1)'7'+$2'440
Naturally,
In this
$*of-&1()$*( paper,
R>H;B: the wequestion
study
@A?@BQB;<?H its aris
rec
J?HK<;
portion of sDoF elsewhere. Remarkably, polynomial identity test W1$1'X
appearsinterference
to E>=;KM:KK:A
transmitters
+#( ?@H89 and
%0%+(7 receivers
I#1() in isolation.
&,()/$''+#;
&()/,)7'*2(> While
%&&)1+#$*+&#
A(7')5'B40@ %32# such controlled
%/.8*()84
+(2#*$63(% ?"@ justified
Gaussian
')(%&1+#;
4$+()'440 in play
MP B::?D<HQthe acontext
keychannel
roleofinmo
J?RR9H<JB;
id
w
KHSKE=;E:=G :MD>AF<M
settingsof are
()*+,-$ excellent
$%<+(,(8 platforms
%&#8+1()$D/F
$*+()82(44 $*+()/()(*2( toinB0
<+;<() develop
*<)&.;<'.*
'44,.$*? ideas and
3%()build
%&-'$)(1
+#( 1'+' precisely
*&to-F&'+%
+,investigate theVR?M<:>
how topic of
limited this pap
transm

I
the achievable sDoF region the former case. K>C<J> ?A ;<EIA>J
MBF>X B
N modern
B( BC4
C,$*+40wireless
(8'(%+$//F
&),2(%%(1 communication
2<(# B0 *<( %/.8*()8
%(-()'4 systems,
$)(
$*+()/()$*? interference
/$);(9B'%(
intuition, they also conceal important aspects that are inherent gate interference. %+'+$,*%@ +#3% G=> BAI;G:;J
WeA>:BP=MBF>K
setK:A;
outMHGE:DG<Z:;C
AtoJ??@>AB;<C>
compare
natural coope;Ath
=@@
has become
7$7$25$*?
to actual the systems.
wireless +#(major factor
B(*(!+% ,/
Chiefthatamong
' limits these
4')?( performance.
-$)+3'4 9:9;
is the Inter-
'))'0>
fact 934+$8
that %&&'()$0
2(44 ()*+, -+%.#!J(//./$)
9:9; 2,,&()'+$,* #(*2&)34
2,*2(&+% ')( %&#,(5 &'*+-+6$*+&#4
(D'7$*(1 /),7 IA relative
mitters
1$//()8 M>
:; toprior
eaf2non-cooperativ
is <HF;A9R>H;B:
that, e"f2to<H=;B
each blocD
R<;<QB;<
EL<
any small
ference(*+ set
arises
*+&#4 of cooperating
whenever
+#*()7()(#%( transmitters
multiple and
transmitter-receiver
-+*+;$*+&#4 receivers
/+#($) ')(%&1+#;4 is
I.pairs bound
are
I-./*+'/(0$#*(##$9
to be
&()%&(2+$-(%@
embedded within
$*2431$*?
a large
'* (D'7$*'+$,*
system featuring
,/ +#(
many
NTRODUCTION
/3*1'7(*+'4
other mitters hold1A><AF<>2
the subscriptsa conference
R9:;<@B;N A>:B;>KX
IA[P and to
:K tell
<H NC@?
C<;
present $*/,)7'+$,*8+#(,)(+$2
and each receiver is only 4$7$+%@interested
' )(-$(.in,/retrieving
+#( 2,1$*? infor-
'*1 %$?*'4 J?RR9H<JB;<?HF5 .?=JB::>
transmitters and receivers. This paper studies the fundamental sages. Hence
quantities, :; the
e.g.,
e`f2messages
SINR
e6f4 are cl
(signal-
0:QQ<><;E
mation from itsnetworks
Cellular
&),2(%%$*? own transmitter.
'4?,)$+#7$2 Due to the broadcast
are fundamentally
1(-(4,&7(*+%@ limited and
'*1@ ?,$*? super- interference.
by inter-cell
B(0,*1 +#'+@ In this context,
J?R@A>FF=I?ADBAK J??@>AB d
performance of IA, in the context of a large cellular network, operative messages,
respectively with
D><K<;E<B which
IA and are
with t
position2,*%$1()'+$,*
nature of ,/ -()0
wireless &)'2+$2'4
channels, one
and contrasts it with that of non-cooperative MIMO.
$%%3(%
users )(4'+(1 +, %2'4'B$4$+0 '*1
information-car-
ters dueH?K>F ?II>A :;
to the;?conference;
e7f2 e!9f
B @?D>AI9:
and 2
has emerged
rying
%0%+(784(-(4
signal causes
as a$*+(?)'+$,*>
useful yetEtractable
interference
$4/(.
to other
metric'*1
$ */)'0(./$'*
&),7$%$*?
users. The
in quantifying
63$+( /3*1'7(*+'4
interference In order
the extent to
to:Kwhich
KLAI;keep :;thee!!f4
interferencecompa PA
Index TermsDistributed
)(%(')2# '-(*3(% ')( '4%, cooperation,
%3??(%+(1>Interference alignment, arethat those unknown
TN<:>
are highly to the
J?HC>H;<?HB:
favorable toother tra
K<C
IA
channel
Interference
this The
issue,
isolated
M5S4G4,
"
is the */%)P%)%*.%
simplest information
management,
./0()are
authors :; 1(%2#!F,,&()'+$,*@
where with the
each
I:><G<KK
Spatial
The Edward
transmitter
@<GGHG=>
theoretic
M=;=C<M<;E
multiplexing.
9:9;@
S. Rogers
(receiver)
KRKE<MK4
model
:K = for
$;2(4434')
Sr.
is assumed
EL:K
$*+()/()(*2(@ B('7/,)7$*?@ 2,,)1$*'+$,*@ 734+$82(44@ 1$%+)$B3+(1>
from othere-mail:
BAI;G:;J
Canada transmitters
@<GGHG=> (receivers).
;<EIA>J2 =;BIn:;F<KE:C=E<
I.(gsridharan@comm.utoronto.ca,
I NTRODUCTION
studying @L=GG<;C<
QH;B=M<;E=G
*(+.,)5%@
Department
D=D<>2 to
link
be
variousweiyu@comm.utoronto.ca).
practical
EL< D<>QA>M=;@<
fectand
)(4'0%@
I<ofbe@A;K:B<>
Electrical
ing
<R@A?C>
capacities
EL<
EL<
transmit
Computer
EL<
:<;;:> ;?
classified
overheads
F>C>A>
ADE:M:d=E:A;
;N>
are
and
HK<>
based<HJA>BF>
EL<<H;>A=J>::
on
:<HE=:>C>:
finite.
receive
Engineering,
>=E<K
neglected,
On
their
E>=;KM:E D><@AB<
A>
AQ
the
CSI
-$*
;N>targe
@
Univ
W9B
and
<H;>AI>A>
B
o

sages,
the which
optimum are those
IA aimed
precoders at i
scenarios, however,
S<;<!EK they
AQ are
=GGAI:;C not isolated
@AAD<>=E:A;and cooperation
=;B TA:;E among
D>A@<KK:;C
This work was supported by the Natural Science and Engineering Research Council =MA;C ><G=E:F<
J>:: (NSERC)
M?9HKBAP EA EL<
of Canada. :;E<>Q<
BA>BFat?I The
Intercell interference is arguably theS=K<
most severe@AAD<>=E:A;2
impairment ones, which
iterative are those
processes aimed
or the the
addit
transmitters EL<or receivers
S=K< can
KE=E:A;K4
presented in part at Canadian Workshop be induced.
U:ELAHE
%5 % +0*(1)/0%(+ For example,
KE=E:A; in down-
Inf. Theory, Jun. 2013, IEEE Int. EL< KRKV
Symp. I:EL
<H;>AI>A>HJ>
Inf. AF<>G=DD:;C
Theory, FN?9:K
Jul. 2013,M>and@A
K>
in contemporary wireless systems, and its there are four kinds of messages
link cellular E<Msystems, base stations
:K :;E<>Q<><;@<VG:M:E<B2 are :4<42 EL<mitigation
connected via has long to actually
infrastruc-
K:C;=GVEAV:;E<>Q<><;@<VDGHKV compute
@L=;;<G
dC<A;9B:d :K2013. such pre
@A;K:B<><B
?A10,dH>;D?AEd 2%
Dec. 2013.
!"
beenbackhaul
Manuscript
#$%&'()
a favorite
submitted
*'+,
research
to IEEE
'(+$-.$-$(/$0
theme.
Transactions
/1(2$(+'1(%&
OfMAS:G<K
late, base
on Information
station%(3 (12$& for
Theory
%45
cooper- K:MDGR theonwhole
Decembersystem. 1 Now the q
ture ;A:K< networks.
>=E:A W-$*)X =E EL< @=;;AE S< :MD>AF<B <;@AB<B
?I J?=JNBHH>: SR
:<HEFK<D=>=E<
;NB; JB
ation
4-1%/,$6
has gained the perception these messages? II. .+*S YST
In our previous
SR work
:;@><=K:;C EL< we ofhave
[1],S=K< being
KE=E:A; best way
studied
E>=;KM:E to counter
the DAI<>2
two-user K:;@< L:CL<> F>AC<J>5
DAI<>%H ;N> N<QN
@A;KE>=:;EK W?+A
Generally speaking, Gaussian
intercell
Gaussian
G ,1%+L
December interference.
interference
10, 2013
E>=;KM:E
B<DGARM<;E
BHK
standsAQ
<H;>AI>A>HJ>
Among
channel
DAI<>
IA
=
the
=GKA with
@><=E<K
(interference
!Z<B
IBK<HQ @9;F :<R<;F ?H ;N> J?C>ABQ>
BA>
various
;HMS<>
;N>
conferencing
;D?
KE>A;C<>
MP K>F<QH>AF ?I R?M<:> J?RR9H<JB;<?H FPF;>RF5 TN<:>
being considered AQ S=K<
E>P
cooperation JNB::>HQ>F
schemes
receivers
:;E<>Q<><;@<4
alignment), which=;B
KE=E:A;K
IBJ>K
to Y:F<; We
has=;E<;;=K2
BHK A>:<BM<:<;P ?I Gaussian
http://mc.manuscriptcentral.com/t-it BHP
FJ>HBA<?X3
EL<
transmitter
consider
H9RM>A
e!7fge5!f2
ter is equipped
;N<F
cooperation
G<F<G ?I
interference
!Q9A>
a wireless
=;B 0?.
J?HJ9AA>H;
with
D<>QA>M=;@<
?I R
netw
are <H;
Nt tran
channels
m
C=
w
the advantage of admitting distributed implementations [1], A>I>AA>K ;? BF@AA>B:;=E:A;
;<EIA>J ;N> R9:;<@
theisfollowing
equippedextreme with Nrcase. receive
Manuscript A;< =DD>A=@L
received April EA :;@><=K<
29, 2010;J?HH>J;<?H3 KRKE<M
revised September EL>AHCLDHE :K
16, 2010; accepted EA =GGAI EL<
@?<H;=;?=@?<H;
[2]. In12, a nutshell,
D<A>:>FF
IA ensures that the
>5Q53 M>;D>>H
interference
B MBF> F;B;<?H
at each H9RM>A ?I 3$)-$$6 1.Wh .-
January TA:;E
2011. <;@AB:;C
Date of currentAQ HK<>
version K:C;=GK
April 20, =@>AKK
2011. EL<
This S=K<
work was
BHK B R?M<:> ;>AR<HB:3 <H;>AI>A>HJ> A>F;A<J;F ;N> A>9FBM<:<;Pin?Ian unlimitedKE=E:A;K4
sup- receiver
$; EL:K link
U< supports
@A;K:B<>
fashion, d
the
EL< scem
M
ported in part
receiver by the=KKHM:;C
aligns Nationala Science
along certain Foundation
subspace under Grant
leaving CCF-0830796
the remaining ;>AR<H?:?QP5
@=K<2 D<>Q<@E @AAD<>=E:A; =MA;C EL< S=K< KE=E:A;K2 Nr 1
 Interference is not the Problem
 Interference is not the Problem

Pi Pi + j 1st tier Pj
SINRi = SINRi =
Pj + Pj + N Pj + N
j 1st tier j / 1st tier j / 1st tier
 Interference is not the Problem

Some examples
Company Increase in Report
Traffic
Qualcomm 20% CTW 2011 Keynote
Alcatel-Lucent 10% ICC 2012 Keynote
Huawei 8% [2013]

Interference is a problem
but its not the problem!
 Interference is not the Problem

Interference!
Management!
 Interference is not the Problem

We should join the noise, rather than fight it

Tom Cover
 Embrace Messiness

For every complex problem there is a simple solution,

and its wrong
U. Eco
 Embrace Messiness
HISTORY OF COMMUNICATIONS

Cellular Service Trial Chicago Cellular testbed Newark

Figure 3. Coverage maps for the Chicago and Newark trials.

(Continued from page 20) ting and reduced the number of cells would simulate operation in a few 1-mi
that were needed at most points in cells, surrounded by six interfering cells
 Embrace Messiness

HetNets: macrocells, microcells, picocells, femtocells

CoMP
Separate uplink/downlink routes
Direct D2D
Figures courtesy of Prof. Jeff Andrews (UT Austin)
and Prof. Martin Haenggi (U. of Notre Dame)
 Embrace Messiness

Voronoi diagrams of line segments
Farthest-point Voronoi diagrams
Roundness
Roundness Voronoi diagrams of line segments Higher-order Voronoi diagrams
Higher-order Voronoi diagrams Farthest-point Voronoi diagrams Computing the farthest-point Voronoi diagram
Computing the farthest-point Voronoi diagram Roundness
Roundness

First order Voronoi diagram Second order Voronoi diagram

,&-!!."
!!" ,!+-!!."
,#-!!."
!+" ,#-&."
,!-!!."
#" &" ,*-!+." ,!-!+." ,!-#."
V1 (d1 )
V2 (d1 , d2 )
!" ,#-%."
,!-*." ,%-&."
,!-$."
,)-!+."
*" %" ,'-&."
,$-%."
$" ,$-*." ,%-'."
,)-*." ,%-(."
'" ,(-*." ,$-(."
)"
(" ,'-(."
,(-)."

Roundness
Voronoi diagrams of line segments
(a)!
Farthest-point Voronoi diagrams
Higher-order Voronoi diagrams
Computing the farthest-point Voronoi diagram (b)!
Computational Geometry Roundness
Lecture 13: More on Voronoi diagrams Computational Geometry Lecture 13: More on Voronoi diagrams

Second order Voronoi diagram

,&-!!."
,!+-!!."
,#-!!."
,#-&."
,!-!!."
,*-!+." ,!-!+."
,!-#." ,#-%."
,!-*." ,%-&."
,!-$."
,)-!+."
,$-%."
 Embrace Messiness

Embrace Messi-ness
 Virtualization is Coming to Town

He that will not apply new remedies

must expect new evils:
for time is the greatest innovator
Francis Bacon
 Virtualization is Coming to Town

Extreme densification
Massive MIMO
 Virtualization is Coming to Town
 Virtualization is Coming to Town

Antennas

Cloud
RAN
DSP

Amplifiers
 Virtualization is Coming to Town

Physical structures get blurred

Virtual structures rule
Cloud RAN
Cloud RAN

Indoor
Radio Head
+ Antenna

Outdoor Low-Power
Radio Head + Antenna Massive MIMO
Outdoor High-Power
Radio Head + Antenna
Radio Head + Antenna
 Virtualization is Coming to Town

How fast can we reconfigure?

(Average Google query < 200 ms)
Cloud Caching at
RAN the Edges
 Cutting the Wireless Wire
 Cutting the Wireless Wire
 Cutting the Wireless Wire

A solution to the perfect storm?

The curse of Moores Law
 Cutting the Wireless Wire

b/s
Hz
High energy efficiency

Eb
N0
(dB) Low spectral efficiency
Mobile User Perspective

bits/s
bits/s = antennas Hz
Hz antenna

Careful with the high-SNR infatuation

Final Remarks
 Watch out for hype
Lets not lose sleep over interference
There are no cells, only nodes
Think messy, think virtual
Energy efficiency also matters
Our results are only as relevant as our models
DONTs of Mathematical Modeling
Cum grano salis
(Don't believe the 33rd order consequences of a 1st order model)
Use only as directed
(Don't apply any model till you understand the simplifying
assumptions on which its based, and can test their ability)
Don't go off the deep end
(Don't extrapolate beyond the region of fit)
Don't beat a dead horse
(Don't retain a discredited model)
Solomon W. Golomb, 1970