6G: Towards a Fully Digital and Connected World · 6G: Towards a Fully Digital and Connected World 6G Wireless Summit, 26th March 2019 6G Applications and Use Cases 8 MASSIVE SCALE

6G: Towards a Fully Digital and Connected World

6G Wireless Summit, Levi, FinlandMarch 26th, 2019

Marco Giordani◦, Michele Polese◦, Marco Mezzavilla†, Sundeep Rangan†, Michele Zorzi◦◦University of Padova, Department of Information Engineering, Italy

† NYU WIRELESS, Tandon School of Engineering, New York University, Brooklyn, NY, USA

Fare clic per modificare stile

6G: Towards a Fully Digital and Connected World 6G Wireless Summit, 26th March 2019

Ø Introduction and Motivations

Ø 6G Key Performance Indices (KPIs)

Ø 6G Potential Applications

Ø 6G Enabling Technologies• Disruptive Communication Technologies • Novel 6G Communication Enabler• Innovative Network Architectures

• Integrating Intelligence in the Network

Ø Conclusions and Research Directions

Outline

2



Introduction and Motivations

3

• From 1G to 5G, passing through UMTS and LTE innovations, each generation of mobile technology has tried to meet the needs of network operators and final consumers

• The rapid development of data-centric and automated processes may exceed even the capabilities of emerging 5G systems, thereby calling for a new wireless generation

time

1G

Voice callingMassive broadband Internet of Things

1-10 Gbps

1980

inno

vati

on

64 Kbps

SMS

1990

Internet

2 Mbps

2000

Internet of Applications

100-1000 Mbps

2010 2020

2.4 Kbps

2G3G

4G

5G

6G

2025-2030

Towards a Fully Digital and Connected World



Introduction and Motivations

4

5G is always associated with trade-offs: 6G will contribute to fill the gap between beyond-2020 societal and business demands and what 5G (and its predecessors) can support

• Consider potential applications for future connected systems and estimate the key requirements in terms of throughput, latency, connectivity and other factors.

• Identify use cases beyond the performance of 5G systems under development today.

• Survey emerging technologies that are not available in networks today but have significant promise for future 6G systems (including developments at all layers of the protocol stack)

Disruptive Technologies Energy Efficiency

Artificial Intelligence

Disaggregation/ virtualization

Cell-less Networks

New Spectrum


5

6G: Towards a Fully Digital and Connected World 6G Applications and Use Cases




6G Use Cases

6

Massive Scale Communication

Tactile Internet

Autonomous Driving

Virtual Reality

Industry 4.0

Smart Cities

E-Health

High-Speed Mobility

Holographic Telepresence



6G Applications and Use Cases

7

INDOOR COVERAGE

• 80% of the mobile traffic is generated indoor

• 5G current cellular networks never really targeted indoor coverageØ High-frequencies cannot penetrate solid materialØ 5G densification through femtocells (proposed as a solution) presents scalability

issues and high deployment and management costs for operators

6G should target cost-aware indoor connectivity solutions autonomously deployed by end-users

and managed by the network operators (e.g., wireless relays coupled with

indoor communications)

6G




8

MASSIVE SCALE COMMUNICATIONS

• 5G networks are designed to support more than 1’000’000 connections per km2

• Mobile traffic will grow 3-fold from 2016 to 2021, pushing the number of connected devices to the extreme (> 500 billion connected things worldwide by 2030)

• This will stress already congested networks, which will not guarantee the required QoS

6G targets capacity expansion to offer high throughput and continuous connectivity,

even when civil communication infrastructures may be compromised (public safety is main requirement)

6G




9

eHEALTH

• OBJECTIVE: revolutionize the health-care sector, e.g., eliminating time and space barriers through remote surgery and guaranteeing health-care workflow optimizations.

• Current communication technologies cannot be applied in future health-careØ high cost and lack of real-time tactile feedback

Ø mmWaves can support low-latency, but do not guarantee connection continuity.

6G enhancements will unleash the potential of eHealth applications through innovations like mobile edge computing, virtualization

and artificial intelligence

6G




10

INDUSTRY 4.0 and ROBOTICS

• OBJECTIVE: digital transformation of manufacturing through Cyber Physical Systems (CPS) and Internet of Things (IoT) services.

• Enabling, among other things, Internet-based diagnostics, maintenance, operation, and direct Machine to Machine (M2M) communications in a cost-effective, flexible and efficient way.

• CPS will break the boundaries between the physical factory and the cyber space computation.

6G will foster the Industry 4.0 revolution through new semiconductor and IC innovations (e.g., terahertz scale electronic packaging solutions)

6G




11

SMART CITY

• OBJECTIVE: life quality improvements, environmental monitoring, traffic control and city management automation

• Current cellular systems have been mainly developed for broadband applications

• Smart city applications build upon data generated by low-cost and low-energy consuming sensors, which efficiently interact with each other.

6G will seamlessly include support for user-centric M2M communication, promoting

ultra-long battery lifetime combined with energy harvesting approaches

6G




12

HOLOGRAPHIC TELEPRESENCE

• OBJECTIVE: remotely connect with an increasing amount of digital accuracy

• ISSUE: raw hologram, without any optimization nor compression, with colors, full parallax, and 30 fps, would require a daunting 4.32 Tbps data rate.

• ISSUE: latency requirement will hit the sub-millisecond, and thousands of synchronized view angles will be necessary (2 tiles for 4K/8K HD, and 12 tiles for VR/AR)

6G will develop architectures and network designs able to digitalize and transfer all the

5 human senses, increasing the overall target data rate

6G




13

UNMANNED MOBILITY (Autonomous Driving)

• OBJECTIVE: fully autonomous connected and intelligent transportation systems, offering safer traveling, improved traffic management, and support for infotainment applications (>7000B$)

• Unprecedented levels of communication reliability and low end-to-end latency, even in ultra-high mobility scenarios (up to an impressive 1000 km/h).

• Sensors (more than 200 per vehicle by 2020) will demand increasing data rates (in the order of terabytes per driving hour), saturating the capacity of traditional technologies.

6G will pave the way for the coming era of connected vehicles through hardware and

software advancements and new technologies to eliminate typical 5G

latency/reliability/throughput trade offs

6G


14

6G: Towards a Fully Digital and Connected World 6G Technologies and Innovations




6G Technologies and Innovations

15

Disruptive Communication Technologies – Terahertz

4

Increasing energy and bandwidth

Increasing wavelength

Legacy Spectrum Millimeter Waves TeraHertz Visible Light6 GHz 30 GHz 300 GHz 300 GHz 10 THz 430 THz 770 THz

0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 0.020 W 0.02 WReceived Power [W]0 dB 150 dBPathloss [dB]

10510-5

0.1

10-2

102

Frequency [THz]

Dist

ance

[m]

6 150 30010

100

200

Frequency [GHz]

Dist

ance

[m]

70 dB 150 dBPathloss [dB]

24

68

24

68

X Room [m]Y Room [m]

Rece

ived

Pow

er [W

] LED1

LED2

Distance [m]100 500 1000

20

80

120

Path

loss

[dB]


Micro Macro Smart City

Fig. 3: Pathloss for sub-6 GHz, mmWave and terahertz bands, and received power for Visible Light Communications (VLC). Notice that thelimits of the axis and the legends are different in each frequency band, to better illustrate the differences and the possible scenarios in whicheach band could be exploited. The sub-6 GHz and mmWave pathloss is computed using 3GPP models and considers both Line-of-Sight(LOS) and Non-Line-of-Sight (NLOS) conditions, while LOS-only is considered for terahertz (with the model from [8]) and VLC (usingthe model described in [9]).

mature than that on terahertz communications, also thanksto a lower cost and higher availability of experimentalplatforms. A standard for VLC (i.e., IEEE 802.15.7 [1])has also been defined; however, this technology has neverbeen considered so far for inclusion in a cellular networkstandard. As reported in Fig. 3, VLC have limited cover-age range, require an illumination source and suffer fromshot noise from other light sources (e.g., the sun), thuscan be mostly used indoors [12]. Moreover, they needto be complemented by RF for the uplink. Nonetheless,VLC could be used to introduce cellular coverage inindoor scenarios, which, as mentioned in Sec. II, is ause case that has not been properly addressed by cellularstandards. In indoor scenarios, VLC can exploit a verylarge unlicensed band, and be deployed without cross-interference among different rooms and with relativelycheap hardware.

Besides the new spectrum, 6G will also transform wirelessnetworks by leveraging a set of technologies that have beenrecently enabled by advancement in physical layer and circuitsresearch, but are not part of 5G. The following will be keyenablers for 6G:

• Integration of full-duplex capabilities in the com-

munication stack. With full-duplex communications,the transceiver in base stations and User Equipments(UEs) will be capable of receiving a signal while alsotransmitting, thanks to self-interference-suppression cir-

cuits [13]. This enables continuous downlink transmissionwith uplink acknowledgments or control messages (orvice versa), to increase the multiplexing capabilities andthe overall system throughput without using additionalbandwidth. Nonetheless, 6G networks will need carefulplanning in the design of allowed full-duplex proceduresand deployments to avoid interference, and novel designsof the cellular network schedulers [13].

• Novel channel estimation techniques (e.g., out-of-band

estimation and compressed sensing). The channel esti-mation for Initial Access (IA) and beam tracking will be akey component of ultra-high frequencies communicationsin a cellular context, as for mmWaves. However, it isdifficult to design efficient procedures for directionalcommunications, considering multiple frequency bandsand possibly a very large bandwidth. Therefore, 6Gsystems will need new channel estimation techniques.Recently, out-of-band estimation (e.g., for the angulardirection of arrival of the signal) has been proposed toimprove the reactiveness of beam management schemes,by exploiting the omnidirectional propagation of sub-6 GHz signals and mapping the channel estimation tommWave frequencies [14]. Similarly, given the sparsityin terms of angular directions of mmWave and terahertzchannels, it is possible to exploit compressive sensing toestimate the channel using a reduced number of samples.

• Sensing and network-based localization. The usage ofRF signals to enable simultaneous localization and map-

• Frequency bands between 100 GHz and 1 THz à bring to the extreme the potentials and challenges of high-frequency communications.

• Huge data rates (possible to allocate contiguous chunks of up to 200 GHz of spectrum)

PROPAGATION LOSS (compensated using directional antenna arrays, also enabling spatial multiplexing without increasing the interference)

MOLECULAR ABSORPTION (it is important to choose deployments in frequency bands not severely affected by molecular absorption)

4





10510-5

0.1

10-2

102

Frequency [THz]

Dist

ance

[m]

6 150 30010

100

200

Frequency [GHz]

Dist

ance

[m]


24

68

24

68


Rece

ived

Pow

er [W

] LED1

LED2

Distance [m]100 500 1000

20

80

120

Path

loss

[dB]















16

Disruptive Communication Technologies – Visible Light Communications

• Frequency bands between 430 GHz and 770 THz à complement RF communications by piggybacking on the wide adoption of LED luminaries

• VLC devices can switch between different light intensities to modulate a signal

• More mature research than THz (standard for VLC – IEEE 802.15.7 – has been defined)

4





10510-5

0.1

10-2

102

Frequency [THz]

Dist

ance

[m]

6 150 30010

100

200

Frequency [GHz]

Dist

ance

[m]


24

68

24

68


Rece

ived

Pow

er [W

] LED1

LED2

Distance [m]100 500 1000

20

80

120

Path

loss

[dB]












INTERFERENCE (limited coverage range, require an illumination source and suffer from shot noise from other light sources)

MULTI-CONNECTIVITY: need to be complemented by RF for the uplink

4





10510-5

0.1

10-2

102

Frequency [THz]

Dist

ance

[m]

6 150 30010

100

200

Frequency [GHz]

Dist

ance

[m]


24

68

24

68


Rece

ived

Pow

er [W

] LED1

LED2

Distance [m]100 500 1000

20

80

120

Path

loss

[dB]















17

Disruptive Communication Technologies – Full-Duplex

• The transceiver in base stations and UEs will be capable of TX a signal while also TX

• Continuous downlink transmission with uplink acknowledgments or control messages àincrease multiplexing and system throughput without using additional bandwidth.

• IDEA: leverage channel state information acquired at a lower frequency as a form of side information on a higher frequency channel.

• Need to define a “transformation function” to relate the spatial correlation matrix derived at one frequency to another at a much different frequency

Disruptive Communication Technologies – OBB channel estimation

11

Gaussian or truncated Laplacian), and the complex coefficients ↵rc,i according to a suitable

fading model (e.g., Rayleigh or Ricean).

III. SYSTEM AND CHANNEL MODEL

We consider a multi-band MIMO system shown in Fig. 1, where ULAs of isotropic point

sources are used at the TX and the RX. The ULAs are considered for ease of exposition, whereas,

the proposed strategies can be extended to other array geometries with suitable modifications. We

assume that the sub-6 GHz and mmWave arrays are co-located, aligned, and have comparable

apertures. Both sub-6 GHz and mmWave systems operate simultaneously.FOR SUBMISSION TO IEEE 14

Base stationUser equipment

MmWave SystemSub-6 GHz System

H , H

FOR

SUB

MIS

SIO

NTO

IEEE

30

[40]

R.J

.Wei

ler,

M.P

eter

,T.K

hne,

M.W

isot

zki,

and

W.K

eusg

en,“

Sim

ulta

neou

smill

imet

er-w

ave

mul

ti-ba

ndch

anne

lsou

ndin

g

inan

urba

nac

cess

scen

ario

,”in

Proc

.Eur

.Con

f.An

tenn

asPr

opag

.(Eu

CAP

),M

ay20

15,p

p.1–

5.

[41]

A.S

.Poo

nan

dM

.Ho,

“Ind

oor

mul

tiple

-ant

enna

chan

nel

char

acte

rizat

ion

from

2to

8G

Hz.

”in

Proc

.IEE

EIn

t.C

onf.

Com

mun

.(IC

C),

2003

,pp.

3519

–352

3.

[42]

S.Ja

ecke

l,M

.Pet

er,K

.Sak

aguc

hi,W

.Keu

sgen

,and

J.M

edbo

,“5G

Cha

nnel

Mod

els

inm

m-W

ave

Freq

uenc

yB

ands

,”in

Proc

.Eur

.Wire

less

Con

f.,M

ay20

16,p

p.1–

6.

[43]

A.O

.Kay

a,D

.Cal

in,a

ndH

.Vis

wan

atha

n.(2

016)

28G

Hz

and

3.5

GH

zW

irele

ssC

hann

els:

Fadi

ng,D

elay

and

Ang

ular

Dis

pers

ion.

[44]

R.C

.Qiu

and

I.-T.

Lu,“

Mul

tipat

hre

solv

ing

with

freq

uenc

yde

pend

ence

forw

ide-

band

wire

less

chan

nelm

odel

ing,

”IE

EE

Tran

s.Ve

h.Te

chno

l.,vo

l.48

,no.

1,pp

.273

–285

,199

9.

[45]

K.H

aned

a,A

.Ric

hter

,and

A.F

.Mol

isch

,“M

odel

ing

the

freq

uenc

yde

pend

ence

oful

tra-w

ideb

and

spat

io-te

mpo

rali

ndoo

r

radi

och

anne

ls,”

IEEE

Tran

s.An

tenn

asPr

opag

.,vo

l.60

,no.

6,pp

.294

0–29

50,2

012.

[46]

D.D

uple

ich,

R.S

.Tho

m,G

.Ste

inb,

J.Lu

o,E.

Schu

lz,X

.Lu,

G.W

ang

etal

.,“S

imul

tane

ous

mul

ti-ba

ndch

anne

lsou

ndin

g

atm

m-W

ave

freq

uenc

ies,”

inPr

oc.E

ur.C

onf.

Ante

nnas

Prop

ag.(

EuC

AP),

2016

,pp.

1–5.

[47]

P.K

y,I.

Car

ton,

A.K

arst

ense

n,W

.Fan

,G.F

.Ped

erse

net

al.,

“Fre

quen

cyde

pend

ency

ofch

anne

lpa

ram

eter

sin

urba

n

LOS

scen

ario

for

mm

wav

eco

mm

unic

atio

ns,”

inPr

oc.E

ur.C

onf.

Ante

nnas

Prop

ag.(

EuC

AP),

2016

,pp.

1–5.

[48]

K.

Han

eda,

J.-i.

Taka

da,

and

T.K

obay

ashi

,“E

xper

imen

tal

Inve

stig

atio

nof

Freq

uenc

yD

epen

denc

ein

Spat

io-T

empo

ral

Prop

agat

ion

Beh

avio

ur,”

inPr

oc.E

ur.C

onf.

Ante

nnas

Prop

ag.(

EuC

AP),

2007

,pp.

1–6.

[49]

V.N

urm

ela

etal

.,“M

ETIS

Cha

nnel

Mod

els,”

Mob

ilean

dw

irele

ssco

mm

unic

atio

nsEn

able

rsfo

rth

eTw

enty

-twen

ty

Info

rmat

ion

Soci

ety,

Tech

.Rep

.,20

15.

[50]

A.M

.Say

eed,

“Dec

onst

ruct

ing

mul

tiant

enna

fadi

ngch

anne

ls,”

IEEE

Tran

s.Si

gnal

Proc

ess.,

vol.

50,n

o.10

,pp.

2563

–257

9,

2002

.

[51]

A.A

lkha

teeb

,G.L

eus,

and

R.W

.Hea

thJr

.,“C

ompr

esse

dse

nsin

gba

sed

mul

ti-us

erm

illim

eter

wav

esy

stem

s:H

owm

any

mea

sure

men

tsar

ene

eded

?”in

Proc

.IEE

EIn

t.C

onf.

Acou

st.,

Spee

chSi

gnal

Proc

ess.

(IC

ASSP

),A

pril

2015

,pp.

2909

–291

3.

[52]

M.

L.B

ench

eikh

,Y.

Wan

g,an

dH

.H

e,“P

olyn

omia

lro

otfin

ding

tech

niqu

efo

rjo

int

DO

AD

OD

estim

atio

nin

bist

atic

MIM

Ora

dar,”

Sign

alPr

oces

s.,vo

l.90

,no.

9,pp

.272

3–27

30,2

010.

[53]

M.B

engt

sson

and

B.O

tters

ten,

“Low

-com

plex

ityes

timat

ors

ford

istri

bute

dso

urce

s,”IE

EETr

ans.

Sign

alPr

oces

s.,vo

l.48

,

no.8

,pp.

2185

–219

4,20

00.

[54]

R.

Key

s,“C

ubic

conv

olut

ion

inte

rpol

atio

nfo

rdi

gita

lim

age

proc

essi

ng,”

IEEE

Tran

s.Ac

oust

.,Sp

eech

,Si

gnal

Proc

ess.,

vol.

29,n

o.6,

pp.1

153–

1160

,198

1.

[55]

Z.G

ao,L

.Dai

,Z.W

ang,

and

S.C

hen,

“Spa

tially

com

mon

spar

sity

base

dad

aptiv

ech

anne

lest

imat

ion

and

feed

back

for

FDD

mas

sive

MIM

O,”

IEEE

Tran

s.Si

gnal

Proc

ess.,

vol.

63,n

o.23

,pp.

6169

–618

3,20

15.

[56]

M.M

isha

lian

dY.

C.E

ldar

,“R

educ

ean

dbo

ost:

Rec

over

ing

arbi

trary

sets

ofjo

intly

spar

seve

ctor

s,”IE

EETr

ans.

Sign

al

Proc

ess.,

vol.

56,n

o.10

,pp.

4692

–470

2,20

08.

[57]

I.F.

Gor

odni

tsky

and

B.D

.Rao

,“Sp

arse

sign

alre

cons

truct

ion

from

limite

dda

taus

ing

FOC

USS

:Are

-wei

ghte

dm

inim

um

norm

algo

rithm

,”IE

EETr

ans.

Sign

alPr

oces

s.,vo

l.45

,no.

3,pp

.600

–616

,199

7.

Mm

Wav

eSy

stem

Sub-

6GHz

Syst

em

RF

Cha

in

DA

C

FOR

SUB

MIS

SIO

NTO

IEEE

30

[40]

R.J

.Wei

ler,

M.P

eter

,T.K

hne,

M.W

isot

zki,

and

W.K

eusg

en,“

Sim

ulta

neou

smill

imet

er-w

ave

mul

ti-ba

ndch

anne

lsou

ndin

g

inan

urba

nac

cess

scen

ario

,”in

Proc

.Eur

.Con

f.An

tenn

asPr

opag

.(Eu

CAP

),M

ay20

15,p

p.1–

5.

[41]

A.

S.Po

onan

dM

.H

o,“I

ndoo

rm

ultip

le-a

nten

nach

anne

lch

arac

teriz

atio

nfr

om2

to8

GH

z.”

inPr

oc.

IEEE

Int.

Con

f.

Com

mun

.(IC

C),

2003

,pp.

3519

–352

3.

[42]

S.Ja

ecke

l,M

.Pet

er,K

.Sak

aguc

hi,W

.Keu

sgen

,and

J.M

edbo

,“5G

Cha

nnel

Mod

els

inm

m-W

ave

Freq

uenc

yB

ands

,”in

Proc

.Eur

.Wire

less

Con

f.,M

ay20

16,p

p.1–

6.

[43]

A.O

.Kay

a,D

.Cal

in,a

ndH

.Vis

wan

atha

n.(2

016)

28G

Hz

and

3.5

GH

zW

irele

ssC

hann

els:

Fadi

ng,D

elay

and

Ang

ular

Dis

pers

ion.

[44]

R.C

.Qiu

and

I.-T.

Lu,“

Mul

tipat

hre

solv

ing

with

freq

uenc

yde

pend

ence

for

wid

e-ba

ndw

irele

ssch

anne

lmod

elin

g,”

IEEE

Tran

s.Ve

h.Te

chno

l.,vo

l.48

,no.

1,pp

.273

–285

,199

9.

[45]

K.H

aned

a,A

.Ric

hter

,and

A.F

.Mol

isch

,“M

odel

ing

the

freq

uenc

yde

pend

ence

oful

tra-w

ideb

and

spat

io-te

mpo

rali

ndoo

r

radi

och

anne

ls,”

IEEE

Tran

s.An

tenn

asPr

opag

.,vo

l.60

,no.

6,pp

.294

0–29

50,2

012.

[46]

D.D

uple

ich,

R.S

.Tho

m,G

.Ste

inb,

J.Lu

o,E.

Schu

lz,X

.Lu,

G.W

ang

etal

.,“S

imul

tane

ous

mul

ti-ba

ndch

anne

lsou

ndin

g

atm

m-W

ave

freq

uenc

ies,”

inPr

oc.E

ur.C

onf.

Ante

nnas

Prop

ag.(

EuC

AP),

2016

,pp.

1–5.

[47]

P.K

y,I.

Car

ton,

A.

Kar

sten

sen,

W.

Fan,

G.

F.Pe

ders

enet

al.,

“Fre

quen

cyde

pend

ency

ofch

anne

lpa

ram

eter

sin

urba

n

LOS

scen

ario

for

mm

wav

eco

mm

unic

atio

ns,”

inPr

oc.E

ur.C

onf.

Ante

nnas

Prop

ag.(

EuC

AP),

2016

,pp.

1–5.

[48]

K.

Han

eda,

J.-i.

Taka

da,

and

T.K

obay

ashi

,“E

xper

imen

tal

Inve

stig

atio

nof

Freq

uenc

yD

epen

denc

ein

Spat

io-T

empo

ral

Prop

agat

ion

Beh

avio

ur,”

inPr

oc.E

ur.C

onf.

Ante

nnas

Prop

ag.(

EuC

AP),

2007

,pp.

1–6.

[49]

V.N

urm

ela

etal

.,“M

ETIS

Cha

nnel

Mod

els,”

Mob

ilean

dw

irele

ssco

mm

unic

atio

nsEn

able

rsfo

rth

eTw

enty

-twen

ty

Info

rmat

ion

Soci

ety,

Tech

.Rep

.,20

15.

[50]

A.M

.Say

eed,

“Dec

onst

ruct

ing

mul

tiant

enna

fadi

ngch

anne

ls,”

IEEE

Tran

s.Si

gnal

Proc

ess.,

vol.

50,n

o.10

,pp.

2563

–257

9,

2002

.

[51]

A.A

lkha

teeb

,G.L

eus,

and

R.W

.Hea

thJr

.,“C

ompr

esse

dse

nsin

gba

sed

mul

ti-us

erm

illim

eter

wav

esy

stem

s:H

owm

any

mea

sure

men

tsar

ene

eded

?”in

Proc

.IEE

EIn

t.C

onf.

Acou

st.,

Spee

chSi

gnal

Proc

ess.

(IC

ASSP

),A

pril

2015

,pp.

2909

–291

3.

[52]

M.

L.B

ench

eikh

,Y.

Wan

g,an

dH

.H

e,“P

olyn

omia

lro

otfin

ding

tech

niqu

efo

rjo

int

DO

AD

OD

estim

atio

nin

bist

atic

MIM

Ora

dar,”

Sign

alPr

oces

s.,vo

l.90

,no.

9,pp

.272

3–27

30,2

010.

[53]

M.B

engt

sson

and

B.O

tters

ten,

“Low

-com

plex

ityes

timat

ors

ford

istri

bute

dso

urce

s,”IE

EETr

ans.

Sign

alPr

oces

s.,vo

l.48

,

no.8

,pp.

2185

–219

4,20

00.

[54]

R.

Key

s,“C

ubic

conv

olut

ion

inte

rpol

atio

nfo

rdi

gita

lim

age

proc

essi

ng,”

IEEE

Tran

s.Ac

oust

.,Sp

eech

,Si

gnal

Proc

ess.,

vol.

29,n

o.6,

pp.1

153–

1160

,198

1.

[55]

Z.G

ao,L

.Dai

,Z.W

ang,

and

S.C

hen,

“Spa

tially

com

mon

spar

sity

base

dad

aptiv

ech

anne

les

timat

ion

and

feed

back

for

FDD

mas

sive

MIM

O,”

IEEE

Tran

s.Si

gnal

Proc

ess.,

vol.

63,n

o.23

,pp.

6169

–618

3,20

15.

[56]

M.M

isha

lian

dY.

C.E

ldar

,“R

educ

ean

dbo

ost:

Rec

over

ing

arbi

trary

sets

ofjo

intly

spar

seve

ctor

s,”IE

EETr

ans.

Sign

al

Proc

ess.,

vol.

56,n

o.10

,pp.

4692

–470

2,20

08.

[57]

I.F.

Gor

odni

tsky

and

B.D

.Rao

,“Sp

arse

sign

alre

cons

truct

ion

from

limite

dda

taus

ing

FOC

USS

:Are

-wei

ghte

dm

inim

um

norm

algo

rithm

,”IE

EETr

ans.

Sign

alPr

oces

s.,vo

l.45

,no.

3,pp

.600

–616

,199

7.

Mm

Wav

eSy

stem

Sub-

6GHz

Syst

em

RF

Cha

in

DA

C

FOR

SUB

MIS

SIO

NTO

IEEE

30

[40]

R.J

.Wei

ler,

M.P

eter

,T.K

hne,

M.W

isot

zki,

and

W.K

eusg

en,“

Sim

ulta

neou

smill

imet

er-w

ave

mul

ti-ba

ndch

anne

lsou

ndin

g

inan

urba

nac

cess

scen

ario

,”in

Proc

.Eur

.Con

f.An

tenn

asPr

opag

.(Eu

CAP

),M

ay20

15,p

p.1–

5.

[41]

A.S

.Poo

nan

dM

.Ho,

“Ind

oor

mul

tiple

-ant

enna

chan

nel

char

acte

rizat

ion

from

2to

8G

Hz.

”in

Proc

.IEE

EIn

t.C

onf.

Com

mun

.(IC

C),

2003

,pp.

3519

–352

3.

[42]

S.Ja

ecke

l,M

.Pet

er,K

.Sak

aguc

hi,W

.Keu

sgen

,and

J.M

edbo

,“5G

Cha

nnel

Mod

els

inm

m-W

ave

Freq

uenc

yB

ands

,”in

Proc

.Eur

.Wire

less

Con

f.,M

ay20

16,p

p.1–

6.

[43]

A.O

.Kay

a,D

.Cal

in,a

ndH

.Vis

wan

atha

n.(2

016)

28G

Hz

and

3.5

GH

zW

irele

ssC

hann

els:

Fadi

ng,D

elay

and

Ang

ular

Dis

pers

ion.

[44]

R.C

.Qiu

and

I.-T.

Lu,“

Mul

tipat

hre

solv

ing

with

freq

uenc

yde

pend

ence

forw

ide-

band

wire

less

chan

nelm

odel

ing,

”IE

EE

Tran

s.Ve

h.Te

chno

l.,vo

l.48

,no.

1,pp

.273

–285

,199

9.

[45]

K.H

aned

a,A

.Ric

hter

,and

A.F

.Mol

isch

,“M

odel

ing

the

freq

uenc

yde

pend

ence

oful

tra-w

ideb

and

spat

io-te

mpo

rali

ndoo

r

radi

och

anne

ls,”

IEEE

Tran

s.An

tenn

asPr

opag

.,vo

l.60

,no.

6,pp

.294

0–29

50,2

012.

[46]

D.D

uple

ich,

R.S

.Tho

m,G

.Ste

inb,

J.Lu

o,E.

Schu

lz,X

.Lu,

G.W

ang

etal

.,“S

imul

tane

ous

mul

ti-ba

ndch

anne

lsou

ndin

g

atm

m-W

ave

freq

uenc

ies,”

inPr

oc.E

ur.C

onf.

Ante

nnas

Prop

ag.(

EuC

AP),

2016

,pp.

1–5.

[47]

P.K

y,I.

Car

ton,

A.K

arst

ense

n,W

.Fan

,G.F

.Ped

erse

net

al.,

“Fre

quen

cyde

pend

ency

ofch

anne

lpa

ram

eter

sin

urba

n

LOS

scen

ario

for

mm

wav

eco

mm

unic

atio

ns,”

inPr

oc.E

ur.C

onf.

Ante

nnas

Prop

ag.(

EuC

AP),

2016

,pp.

1–5.

[48]

K.

Han

eda,

J.-i.

Taka

da,

and

T.K

obay

ashi

,“E

xper

imen

tal

Inve

stig

atio

nof

Freq

uenc

yD

epen

denc

ein

Spat

io-T

empo

ral

Prop

agat

ion

Beh

avio

ur,”

inPr

oc.E

ur.C

onf.

Ante

nnas

Prop

ag.(

EuC

AP),

2007

,pp.

1–6.

[49]

V.N

urm

ela

etal

.,“M

ETIS

Cha

nnel

Mod

els,”

Mob

ilean

dw

irele

ssco

mm

unic

atio

nsEn

able

rsfo

rth

eTw

enty

-twen

ty

Info

rmat

ion

Soci

ety,

Tech

.Rep

.,20

15.

[50]

A.M

.Say

eed,

“Dec

onst

ruct

ing

mul

tiant

enna

fadi

ngch

anne

ls,”

IEEE

Tran

s.Si

gnal

Proc

ess.,

vol.

50,n

o.10

,pp.

2563

–257

9,

2002

.

[51]

A.A

lkha

teeb

,G.L

eus,

and

R.W

.Hea

thJr

.,“C

ompr

esse

dse

nsin

gba

sed

mul

ti-us

erm

illim

eter

wav

esy

stem

s:H

owm

any

mea

sure

men

tsar

ene

eded

?”in

Proc

.IEE

EIn

t.C

onf.

Acou

st.,

Spee

chSi

gnal

Proc

ess.

(IC

ASSP

),A

pril

2015

,pp.

2909

–291

3.

[52]

M.

L.B

ench

eikh

,Y.

Wan

g,an

dH

.H

e,“P

olyn

omia

lro

otfin

ding

tech

niqu

efo

rjo

int

DO

AD

OD

estim

atio

nin

bist

atic

MIM

Ora

dar,”

Sign

alPr

oces

s.,vo

l.90

,no.

9,pp

.272

3–27

30,2

010.

[53]

M.B

engt

sson

and

B.O

tters

ten,

“Low

-com

plex

ityes

timat

ors

ford

istri

bute

dso

urce

s,”IE

EETr

ans.

Sign

alPr

oces

s.,vo

l.48

,

no.8

,pp.

2185

–219

4,20

00.

[54]

R.

Key

s,“C

ubic

conv

olut

ion

inte

rpol

atio

nfo

rdi

gita

lim

age

proc

essi

ng,”

IEEE

Tran

s.Ac

oust

.,Sp

eech

,Si

gnal

Proc

ess.,

vol.

29,n

o.6,

pp.1

153–

1160

,198

1.

[55]

Z.G

ao,L

.Dai

,Z.W

ang,

and

S.C

hen,

“Spa

tially

com

mon

spar

sity

base

dad

aptiv

ech

anne

lest

imat

ion

and

feed

back

for

FDD

mas

sive

MIM

O,”

IEEE

Tran

s.Si

gnal

Proc

ess.,

vol.

63,n

o.23

,pp.

6169

–618

3,20

15.

[56]

M.M

isha

lian

dY.

C.E

ldar

,“R

educ

ean

dbo

ost:

Rec

over

ing

arbi

trary

sets

ofjo

intly

spar

seve

ctor

s,”IE

EETr

ans.

Sign

al

Proc

ess.,

vol.

56,n

o.10

,pp.

4692

–470

2,20

08.

[57]

I.F.

Gor

odni

tsky

and

B.D

.Rao

,“Sp

arse

sign

alre

cons

truct

ion

from

limite

dda

taus

ing

FOC

USS

:Are

-wei

ghte

dm

inim

um

norm

algo

rithm

,”IE

EETr

ans.

Sign

alPr

oces

s.,vo

l.45

,no.

3,pp

.600

–616

,199

7.

Mm

Wav

eSy

stem

Sub-

6GHz

Syst

em

RF

Cha

in

DA

C

FOR

SUB

MIS

SIO

NTO

IEEE

30

[40]

R.J

.Wei

ler,

M.P

eter

,T.K

hne,

M.W

isot

zki,

and

W.K

eusg

en,“

Sim

ulta

neou

smill

imet

er-w

ave

mul

ti-ba

ndch

anne

lsou

ndin

g

inan

urba

nac

cess

scen

ario

,”in

Proc

.Eur

.Con

f.An

tenn

asPr

opag

.(Eu

CAP

),M

ay20

15,p

p.1–

5.

[41]

A.

S.Po

onan

dM

.H

o,“I

ndoo

rm

ultip

le-a

nten

nach

anne

lch

arac

teriz

atio

nfr

om2

to8

GH

z.”

inPr

oc.

IEEE

Int.

Con

f.

Com

mun

.(IC

C),

2003

,pp.

3519

–352

3.

[42]

S.Ja

ecke

l,M

.Pet

er,K

.Sak

aguc

hi,W

.Keu

sgen

,and

J.M

edbo

,“5G

Cha

nnel

Mod

els

inm

m-W

ave

Freq

uenc

yB

ands

,”in

Proc

.Eur

.Wire

less

Con

f.,M

ay20

16,p

p.1–

6.

[43]

A.O

.Kay

a,D

.Cal

in,a

ndH

.Vis

wan

atha

n.(2

016)

28G

Hz

and

3.5

GH

zW

irele

ssC

hann

els:

Fadi

ng,D

elay

and

Ang

ular

Dis

pers

ion.

[44]

R.C

.Qiu

and

I.-T.

Lu,“

Mul

tipat

hre

solv

ing

with

freq

uenc

yde

pend

ence

for

wid

e-ba

ndw

irele

ssch

anne

lmod

elin

g,”

IEEE

Tran

s.Ve

h.Te

chno

l.,vo

l.48

,no.

1,pp

.273

–285

,199

9.

[45]

K.H

aned

a,A

.Ric

hter

,and

A.F

.Mol

isch

,“M

odel

ing

the

freq

uenc

yde

pend

ence

oful

tra-w

ideb

and

spat

io-te

mpo

rali

ndoo

r

radi

och

anne

ls,”

IEEE

Tran

s.An

tenn

asPr

opag

.,vo

l.60

,no.

6,pp

.294

0–29

50,2

012.

[46]

D.D

uple

ich,

R.S

.Tho

m,G

.Ste

inb,

J.Lu

o,E.

Schu

lz,X

.Lu,

G.W

ang

etal

.,“S

imul

tane

ous

mul

ti-ba

ndch

anne

lsou

ndin

g

atm

m-W

ave

freq

uenc

ies,”

inPr

oc.E

ur.C

onf.

Ante

nnas

Prop

ag.(

EuC

AP),

2016

,pp.

1–5.

[47]

P.K

y,I.

Car

ton,

A.

Kar

sten

sen,

W.

Fan,

G.

F.Pe

ders

enet

al.,

“Fre

quen

cyde

pend

ency

ofch

anne

lpa

ram

eter

sin

urba

n

LOS

scen

ario

for

mm

wav

eco

mm

unic

atio

ns,”

inPr

oc.E

ur.C

onf.

Ante

nnas

Prop

ag.(

EuC

AP),

2016

,pp.

1–5.

[48]

K.

Han

eda,

J.-i.

Taka

da,

and

T.K

obay

ashi

,“E

xper

imen

tal

Inve

stig

atio

nof

Freq

uenc

yD

epen

denc

ein

Spat

io-T

empo

ral

Prop

agat

ion

Beh

avio

ur,”

inPr

oc.E

ur.C

onf.

Ante

nnas

Prop

ag.(

EuC

AP),

2007

,pp.

1–6.

[49]

V.N

urm

ela

etal

.,“M

ETIS

Cha

nnel

Mod

els,”

Mob

ilean

dw

irele

ssco

mm

unic

atio

nsEn

able

rsfo

rth

eTw

enty

-twen

ty

Info

rmat

ion

Soci

ety,

Tech

.Rep

.,20

15.

[50]

A.M

.Say

eed,

“Dec

onst

ruct

ing

mul

tiant

enna

fadi

ngch

anne

ls,”

IEEE

Tran

s.Si

gnal

Proc

ess.,

vol.

50,n

o.10

,pp.

2563

–257

9,

2002

.

[51]

A.A

lkha

teeb

,G.L

eus,

and

R.W

.Hea

thJr

.,“C

ompr

esse

dse

nsin

gba

sed

mul

ti-us

erm

illim

eter

wav

esy

stem

s:H

owm

any

mea

sure

men

tsar

ene

eded

?”in

Proc

.IEE

EIn

t.C

onf.

Acou

st.,

Spee

chSi

gnal

Proc

ess.

(IC

ASSP

),A

pril

2015

,pp.

2909

–291

3.

[52]

M.

L.B

ench

eikh

,Y.

Wan

g,an

dH

.H

e,“P

olyn

omia

lro

otfin

ding

tech

niqu

efo

rjo

int

DO

AD

OD

estim

atio

nin

bist

atic

MIM

Ora

dar,”

Sign

alPr

oces

s.,vo

l.90

,no.

9,pp

.272

3–27

30,2

010.

[53]

M.B

engt

sson

and

B.O

tters

ten,

“Low

-com

plex

ityes

timat

ors

ford

istri

bute

dso

urce

s,”IE

EETr

ans.

Sign

alPr

oces

s.,vo

l.48

,

no.8

,pp.

2185

–219

4,20

00.

[54]

R.

Key

s,“C

ubic

conv

olut

ion

inte

rpol

atio

nfo

rdi

gita

lim

age

proc

essi

ng,”

IEEE

Tran

s.Ac

oust

.,Sp

eech

,Si

gnal

Proc

ess.,

vol.

29,n

o.6,

pp.1

153–

1160

,198

1.

[55]

Z.G

ao,L

.Dai

,Z.W

ang,

and

S.C

hen,

“Spa

tially

com

mon

spar

sity

base

dad

aptiv

ech

anne

les

timat

ion

and

feed

back

for

FDD

mas

sive

MIM

O,”

IEEE

Tran

s.Si

gnal

Proc

ess.,

vol.

63,n

o.23

,pp.

6169

–618

3,20

15.

[56]

M.M

isha

lian

dY.

C.E

ldar

,“R

educ

ean

dbo

ost:

Rec

over

ing

arbi

trary

sets

ofjo

intly

spar

seve

ctor

s,”IE

EETr

ans.

Sign

al

Proc

ess.,

vol.

56,n

o.10

,pp.

4692

–470

2,20

08.

[57]

I.F.

Gor

odni

tsky

and

B.D

.Rao

,“Sp

arse

sign

alre

cons

truct

ion

from

limite

dda

taus

ing

FOC

USS

:Are

-wei

ghte

dm

inim

um

norm

algo

rithm

,”IE

EETr

ans.

Sign

alPr

oces

s.,vo

l.45

,no.

3,pp

.600

–616

,199

7.

Mm

Wav

eSy

stem

Sub-

6GHz

Syst

em

RF

Cha

in

DA

C

FOR SUBMISSION TO IEEE 32

[51] M. L. Bencheikh, Y. Wang, and H. He, “Polynomial root finding technique for joint DOA DOD estimation in bistatic

MIMO radar,” Signal Process., vol. 90, no. 9, pp. 2723–2730, Sep. 2010.

[52] M. Bengtsson and B. Ottersten, “Low-complexity estimators for distributed sources,” IEEE Trans. Signal Process., vol. 48,

no. 8, pp. 2185–2194, Aug. 2000.

[53] R. Keys, “Cubic convolution interpolation for digital image processing,” IEEE Trans. Acoust., Speech, Signal Process.,

vol. 29, no. 6, pp. 1153–1160, Dec. 1981.

[54] Z. Gao, L. Dai, Z. Wang, and S. Chen, “Spatially common sparsity based adaptive channel estimation and feedback for

FDD massive MIMO,” IEEE Trans. Signal Process., vol. 63, no. 23, pp. 6169–6183, Dec. 2015.

[55] M. Mishali and Y. C. Eldar, “Reduce and boost: Recovering arbitrary sets of jointly sparse vectors,” IEEE Trans. Signal

Process., vol. 56, no. 10, pp. 4692–4702, Oct. 2008.

[56] I. F. Gorodnitsky and B. D. Rao, “Sparse signal reconstruction from limited data using FOCUSS: A re-weighted minimum

norm algorithm,” IEEE Trans. Signal Process., vol. 45, no. 3, pp. 600–616, Mar. 1997.

[57] “IEEE standard for information technology–Telecommunications and information exchange between systems–Local and

metropolitan area networks–Specific requirements-Part 11: Wireless LAN medium access control (MAC) and physical

layer (PHY) specifications amendment 3: Enhancements for very high throughput in the 60 GHz band,” pp. 1–628, Dec.

2012.

[58] M. Grant, S. Boyd, and Y. Ye, “CVX: Matlab software for disciplined convex programming,” 2008. [Online]. Available:

http://www.stanford.edu/ boyd/cvx

Transmitter Receiver

FOR SUBMISSION TO IEEE 32

[51] M. L. Bencheikh, Y. Wang, and H. He, “Polynomial root finding technique for joint DOA DOD estimation in bistatic

MIMO radar,” Signal Process., vol. 90, no. 9, pp. 2723–2730, Sep. 2010.

[52] M. Bengtsson and B. Ottersten, “Low-complexity estimators for distributed sources,” IEEE Trans. Signal Process., vol. 48,

no. 8, pp. 2185–2194, Aug. 2000.

[53] R. Keys, “Cubic convolution interpolation for digital image processing,” IEEE Trans. Acoust., Speech, Signal Process.,

vol. 29, no. 6, pp. 1153–1160, Dec. 1981.

[54] Z. Gao, L. Dai, Z. Wang, and S. Chen, “Spatially common sparsity based adaptive channel estimation and feedback for

FDD massive MIMO,” IEEE Trans. Signal Process., vol. 63, no. 23, pp. 6169–6183, Dec. 2015.

[55] M. Mishali and Y. C. Eldar, “Reduce and boost: Recovering arbitrary sets of jointly sparse vectors,” IEEE Trans. Signal

Process., vol. 56, no. 10, pp. 4692–4702, Oct. 2008.

[56] I. F. Gorodnitsky and B. D. Rao, “Sparse signal reconstruction from limited data using FOCUSS: A re-weighted minimum

norm algorithm,” IEEE Trans. Signal Process., vol. 45, no. 3, pp. 600–616, Mar. 1997.

[57] “IEEE standard for information technology–Telecommunications and information exchange between systems–Local and

metropolitan area networks–Specific requirements-Part 11: Wireless LAN medium access control (MAC) and physical

layer (PHY) specifications amendment 3: Enhancements for very high throughput in the 60 GHz band,” pp. 1–628, Dec.

2012.

[58] M. Grant, S. Boyd, and Y. Ye, “CVX: Matlab software for disciplined convex programming,” 2008. [Online]. Available:

http://www.stanford.edu/ boyd/cvx

Transmitter Receiver

Fig. 1: The multi-band MIMO system with co-located sub-6 GHz and mmWave antenna arrays.The sub-6 GHz channel is H and the mmWave channel is H.

A. Sub-6 GHz system and channel model

The sub-6 GHz system is shown in Fig. 2. Note that, we underline all sub-6 GHz variables

to distinguish them from the mmWave variables. The sub-6 GHz system has one RF chain per

antenna and as such, fully digital precoding is possible. We assume narrowband signaling at sub-

6 GHz. Extending the proposed approach to wideband sub-6 GHz systems is straight forward

because only the directional information is retrieved from sub-6 GHz, which is not expected

to vary much across the channel bandwidth. We adopt a geometric channel model for H based

on (1). The MIMO channel matrix for sub-6 GHz can be written as

H =

sMRXMTX

⇢pl

CX

c=1

RcX

rc=1

↵rcp(�⌧ c � ⌧ rc)aRX(✓c + #rc)a⇤TX(�c

+ 'rc), (3)




18

Disruptive Communication Technologies – Channel Sparsity

• Estimating the mmWave channel is equivalent to estimating the parameters of the channel paths, i.e., the AoA, the AoD, and the gain of each path.

• IDEA: exploit the poor scattering nature of the mmWave channel to formulate the mmWave channel estimation problem as a sparse compressed sensing problem: the channel power is concentrated in a few entries of a virtual channel matrix

• It is sufficient to estimate the AoAs and AoDsof the dominant paths to be resolved.

6

(a) Grayscale of angular-delay domain (b) Surface plot of angular-delay domain

Fig. 4. An illustration of a 6-path SFW channel, where M = 128, N = 128, d/�c = 0.5, fs/fc = 0.2, and SNR = 10dB.

D. Sparse Channel Representation and Angular-Delay Or-

thogonality

Define the vectorizing SFW channel hp , vec(Hp) as

hp =

Lp�1X

l=0

↵p,l

⇥vec (⇥( p,l)) � vec

�a( p,l)b

T (⌧p,l)�⇤

=

Lp�1X

l=0

↵p,ldiag(vec(⇥( p,l))) (b(⌧p,l)⌦ a( p,l))

,Lp�1X

l=0

↵p,lp( p,l, ⌧p,l) 2 CMN⇥1, (18)

where p( p,l, ⌧p,l) , diag (vec (⇥( p,l))) (b(⌧p,l)⌦ a( p,l))is the corresponding item and serves as the basis vector forspanning hp. The following theorem proved in Appendix Cindicates the orthogonality of the basis vectors of hp.

Theorem 2: When M ! 1, N ! 1, the followingproperty holds

limM,N!1

1

MNp( 1, ⌧1)

Hp( 2, ⌧2) =

(1 1 = 2, ⌧1 = ⌧2

0 otherwise.

(19)

Based on Theorem 2, if any two users do not share thesame path (paths with both the same DOA and the same timedelay), then their vectorized SFW channels are asymptoticallyorthogonal. This phenomenon is called the angular-delay

orthogonality, which indicates the orthogonality for users atdifferent locations, or for users at the same location but withdifferent path delays. Interestingly, as shown in Fig. 4, twosquares in Fig. 4(a) corresponding to two different paths maypartially overlap with each other, which, however, does notbreak the angular-delay orthogonality in terms of Theorem 2unless the two squares locate at the exact identical position.

An important application of Theorem 2 is in channel esti-mation and user scheduling. By exploiting the angular-delayorthogonality, orthogonal users in angular-delay domain can

be simultaneously scheduled without pilot contamination ormutual interference, as demonstrated in the next section.

IV. CHANNEL ESTIMATION FOR DUAL-WIDEBANDMMWAVE MASSIVE MIMO SYSTEMS

The above discussion has demonstrated that the SFWchannel model of the massive MIMO system is not just anextension of the traditional MIMO channel model. This crit-ical difference demands for redesign of most communicationstrategies, such as channel estimation, signal detection, beam-forming, precoding, user scheduling, etc. Due to the spacelimitation, we present a simple channel estimation algorithmin this section to deal with the dual-wideband effects by theaid of array signal processing.

A. Preamble for Initial Uplink Channel Estimation

In the preamble phase, we apply the conventional leastsquare MIMO-OFDM channel estimation algorithm [39] foreach antenna at the BS. Denote bHp as the uplink preamblechannel between the pth user and the BS. The main purposeof the preamble is to obtain the initial DOA and the timedelay of each path for each user and facilitate the subsequentuplink and downlink channel estimations with a small numberof pilot resources.

B. Extracting Angular-Delay Signature

From Theorem 1, #p,l, ⌧p,l, and Lp can be immediatelyobtained from the non-zero square of bGp = F

H

MbHpF

⇤N

.However, when M and N are finite in practice, the regionof the non-zero square will be expanded due to the powerleakage effect [15], [31]. Hence, #p,l, ⌧p,l, and Lp should beobtained by a more sophisticatedly designed way.

Denote

M (� p,l) = diag⇣1, ej� p,l , . . . , ej(M�1)� p,l

⌘(20)

Example of 6 dominant paths (the location of each square reflects the AoA and time delay of each path)

B. Wang, et al., "Spatial- and Frequency-Wideband Effects in Millimeter-Wave Massive MIMO Systems," in IEEE Trans. Sig. Proc., vol. 66, no. 13, pp. 3393-3406, Jul. 2018.




19

Innovative Network Architectures – Disaggregation/virtualization

• IDEA: decouple network/control plane (CP) and forwarding/user plane (UP).Ø SDN offers network programmability and centralization of the control

Ø SDN is agile and responsive (traffic flow meets fluctuating needs and demands).

Ø SDN is standards-based (e.g., OpenFlow) and vendor-neutral.

• IDEA: replace network services provided by dedicated hardware (e.g., network switches) with virtualized software.

Ø NFV saves capital and operating expenses

• C. J. Bernardos et al., "An architecture for software defined wireless networking," in IEEE Wireless Communications, vol. 21, no. 3, pp. 52-61, June 2014.• R. Mijumbi, et al., "Network Function Virtualization: State-of-the-Art and Research Challenges," in IEEE COMST, vol. 18, no. 1, pp. 236-262, 2016

NFV and SDN are complementary technologies(SDN executes on an NFV infrastructure)




20

Innovative Network Architectures – Access/Backhaul integration

• Massive 6G data rates technologies àadequate growth of the backhaul capacity

• THz and VLC deployments will call for a massive increase in the density of access points, which should be provided with backhaul connectivity to the core network à expensive

• IDEA: deploy a fraction of BSs with traditional fiber-like backhaul capabilities and the rest of the BSs connecting to the fiber infrastructures wirelessly.

• 6G deployments will introduce new challenges and opportunities• The networks will need higher autonomous configuration capabilities• Out-of-band IAB can be realized to increase the overall network throughput.

• M. Polese, et al. , "End- to-End Simulation of Integrated Access and Backahul at mmWaves,” to appear on IEEE CAMAD, Sep. 2018.• M. Polese, et al., "Distributed Path Selection Strategies for Integrated Access and Backhaul at mmWaves", to appear on IEEE GLOBECOM, Dec. 2018

Donor gNB

IAB nodeIAB node

MACRLC

PDCP

MACRLC

PDCP

Backhaul Access

SCHED

IAB node stack

PHY PHY




21

Integrating Intelligence in the Network – Learning

• BACKGROUND: the signal received at multiple BSs renders a defining signature for the user location and its interaction with the surrounding environment.

• BACKGROUND: UEs typically move through predefined paths, and some movements are impossible due to the presence of obstacles, e.g., buildings, walls.

• IDEA: account for previous access statistics and use machine learning tools to predict the network behaviors (e.g., by remembering/observing consequences of previous decisions).

SUPERVISED LEARNINGThe amount of data generated will be massive, thus labelingthe data may be infeasible. UNSUPERVISED LEARNING

Does not need labeling, used to autonomously build complex

network representations




22

Integrating Intelligence in the Network – Knowledge sharing and learning

• 5G: At high frequency, the massive bandwidth and spatial degrees of freedom are unlikely to be fully used by any one cellular operator. Spectrum can be shared, in time and in space, with several performance and energy benefits:§ Reducing deployment costs, if operators share bands and infrastructures

§ Inter-operators access and interference coordination

• 6G: operators and users may also be interested in sharing learned representations of specific network deployments and/or use cases§ Speed up the network configuration in new markets

§ Better adapt to new unexpected scenarios which may emerge during network operations

Controllers

Reference Planner

Mission Planning

Behavioral Planner

Perception

Lane Centering

ACC Controller

Merge Assist

Fig. 3: The proposed behavioral planning framework

III. BEHAVIORAL PLANNING FRAMEWORK

As discussed in Section II, there are shortcomings ofboth the current hierarchical and parallel robot decisionmaking architectures. Due to real-time constraints, the mo-tion planner cannot consider the effects of imperfect vehi-cle controllers or cooperation between cars. In this paper,we propose a novel behavioral planning framework thatcombines the strengths of the hierarchical and parallel ar-chitectures. It is based on the hierarchical architecture sothat fully autonomous driving with high-level intelligencecan be achieved. However, it also uses the independentcontrollers as in the parallel autonomous vehicle architectureto ensure basic performance and driving quality. The designgreatly reduces the necessary search space for the motionplanner without sacrificing any performance due to coarsergranularity. The proposed framework is shown in Figure 3.

A. Mission PlanningThe mission planning module takes charge of decompos-

ing driving missions such as “go from point A to point B”into lane-level sub-missions such as which lane we should bein, whether we need to stop at an intersection, etc. It takes ahuman drivers desired destination and computes the shortestpath to it from the robots current position. It outputs a setof future lane-level sub-missions which describe the desiredlane and turn of the vehicle at each intersection. In addition,this module also controls the transition of goals. When thevehicle completes the current lane-level sub-mission, it willautomatically send the next set of goals to the lower layers.When there are intersections with stop signs, traffic lightsor yielding requirements, this module directly talks to theperception system to decide whether the car can proceed tothe next sub-mission.

B. Traffic-free Reference PlanningThe reference planning layer takes the lane-level sub-

missions, and outputs a path and speed profile for theautonomous vehicle to drive [5]. In this layer, the plannerassumes there is no traffic on the road. It uses non-linearoptimization to find a smooth and human-like path and speedwith consideration of the kinematics and dynamics of theautonomous vehicle, as well as the geometry of the road.

Fig. 4: Reference planners result: the desired path cut cornersat the turn to minimize path length and generate betterhandling

Candidate Strategy Generation

Prediction Engine

Cost function-based Evaluation

Best Strategy

Multiple candidate strategies

Predicted surroundingtraffic scenarios

Progress Comfort SafetyFuel

Consum-ption

Fig. 5: The block diagram of Prediction- and Cost-functionBased algorithm (PCB)

For example, if the road is not straight, as shown in Figure4, instead of driving exactly in the center of the road, ahuman driver will drive slightly offset from the center line tominimize the required steering maneuver, which is emulatedin this module. In addition, traffic rules such as speed limitsare applied in this layer.

C. Behavioral Planning

Since the road geometry has been considered in theprevious layer, the behavior planner focuses on handling on-road traffic, including moving obstacles and static objects.It takes the traffic-free reference, moving obstacles andall road blockages as input. It outputs controller directivesincluding lateral driving bias, the desired leading vehiclewe should follow, the aggressiveness of distance keeping,and maximum speed. In this module, a Prediction- andCost-function Based (PCB) algorithm is implemented [17].There are three primary steps: candidate strategy generation,prediction, and cost function-based evaluation, as shown inFigure 5.

In the PCB algorithm, the world is abstracted as shownin Equation 1. V Shost is the state vector for the hostvehicle and surrounding vehicles, with station (longitudinaldistance along the reference path), velocity, acceleration andlateral offset information. V Sother is the state vector foreach surrounding vehicle. Compared with V Shost, it hasadditional information about the intention and probability ofintention. Sroad is the state vector of the reference path withmaximum speed information at each of M stations. SPCB

is the state vector for the behavioral planner.



6G Challenges

23

5

Disaggregated and virtualized RANThe networking equipment will not require dedicated hardware

Cell-less architectureThe UE connects to the RAN and not to a single cell

Extreme multi-connectivityExploit THz, VLC, mmWave and sub-6 GHz links

Efficient and low-power network operationsEnergy will be at the core of 6G protocols design

Generic hardware

Virtual MAC

Virtual PHY

EdgeCloud

Ener

gy h

arve

stin

g

Low

-pow

er n

odes

Fig. 4: Architectural innovations introduced in 6G networks.

ping has been widely studied [15], but such capabilitieshave never been deeply integrated with the operations andprotocols of cellular networks. 6G networks will exploita unified interface for localization and communicationsto (i) improve control operations, which can rely oncontext information (without the need for exchangingGPS coordinates with the user) to control beamformingpatterns, reduce interference, predict handovers; and (ii)offer innovative user services, e.g., for vehicular andeHealth applications.

B. Innovative Network Architectures

The disruption brought by the communication technologiesdescribed in Sec. III-A will require new 6G network architec-ture, but also potentially require structural updates with respectto current mobile network designs. For example, the densityand the high access data rate of terahertz communications willcreate constraints on the underlying transport network, whichhas to provide both more points of access to fiber and a highercapacity than today’s backhaul networks.

The main architectural innovations that 6G will introduceare described in Fig. 4. In this context, we envision theintroduction and/or deployment of the following architecturalparadigms:

• Cell-less architecture and tight integration of multiple

frequencies and communication technologies. 6G willbreak the current boundaries of cells, with UEs connectedto the network as a whole and not to a single cell.This will guarantee a seamless mobility support, withoutoverhead due to handovers (which might be frequentwhen considering systems at terahertz frequencies), andwill provide Quality of Service (QoS) guarantees even inchallenging mobility scenarios such as vehicular ones.

The overcoming of the cell concept will also enablea tight integration of the different 6G communicationtechnologies. The users will be able to seamlessly tran-sition among sub-6 GHz, mmWave, terahertz or VLClinks without manual interventions or configurations inthe device, which will automatically select the best avail-able communication technology. Finally, according to thespecific use case, the UE may also concurrently use dif-ferent network interfaces to exploit their complementarycharacteristics, e.g., the sub-6 GHz layer for control, andterahertz link for the data plane.

• Disaggregation and virtualization of the networking

equipment: from the physical layer to NFV. Networkshave recently started to transition towards the disaggre-gation of once-monolithic networking equipments: forexample, 5G networks base stations can be deployedwith distributed units with the lower layer of the protocolstack, and centralized units in data centers at the edge.Following this direction, 6G networks will adopt an evenmore disruptive architecture, where the units deployedon the ground will contain just the physical antennas andthe lowest amount of processing units possible. Moreover,virtualization will be brought to the extreme, thanks tothe advances in the capabilities of general purpose pro-cessors: 6G will virtualize additional components, suchas those related to the Medium Access Control (MAC)and Physical (PHY) layers, which currently require ded-icated hardware implementations. The virtualization willdecrease the costs of networking equipment, making amassively dense deployment economically feasible.

• Advanced access-backhaul integration. The massivedata rates provided by the new 6G access technologieswill require an adequate growth of the backhaul capacity.

• Circuit design, high propagation loss • Limited coverage, need for RF uplink• Need for reliable frequency mapping

• Scheduling, need for new network design • Scalability, and interference management

• Slower network operations/security concerns• CU/UP functions are tightly coupled

• Integrate energy characteristics in protocols • Energy vs. high-deployment / MIMO

6G: Towards a Fully Digital and Connected World


Marco Giordani◦, Michele Polese◦, Marco Mezzavilla†, Sundeep Rangan†, Michele Zorzi◦◦University of Padova, Department of Information Engineering, Italy

† NYU WIRELESS, Tandon School of Engineering, New York University, Brooklyn, NY, USA

Documents

6G: Towards a Fully Digital and Connected World · 6G: Towards a Fully Digital and Connected World 6G Wireless Summit, 26th March 2019 6G Applications and Use Cases 8 MASSIVE SCALE