MOBILE CO M MUNICATIONS AND NET WORKS To ward 6 G Net works: Use Cases and Technologies

Marc o Gior da ni, Mic hele P olese, Marc o Mezzavilla, S u n dee p Ra nga n, a n d Mic hele Z orzi

A bst r A c t • Ne w fr equency bands, such as milli me - T he a ut h ors disc uss tec h- ter- wave ( m m Wave) spectru m n ologies t hat will ev olve Relia ble data c o n nectivity is vital f or t he ever • Advanced spectru m usage and manage ment wireless net w orks t o war d increasingly intelligent, aut o mated, and u bi qui - i n li c e ns e d a n d u nli c e ns e d b a n ds a sixt h ge neratio n ( 6 G), t ous digital w orld. M obile net w orks are the data • A c o m pl et e r e d esi g n of t h e c or e n et w or k high ways an d, in a fully c onnecte d, intelligent Ho wever, the rapid develop ment of data-ce n - a n d t hat we c o nsider digital w orl d, will nee d t o c onnect everything, tric and auto mated processes, which require a as e na blers f or several including pe ople t o vehicles, sens ors, data, cl oud data rate on the order of tera bits per sec ond, a p ote ntial 6 G use cases. res ources, and even r ob otic agents. Fifth genera - latency of hundreds of microseconds, and 10 7 T hey pr ovide a f ull-stack, ti on (5 G) wireless net w orks, which are currently connections per k m 2 , may exceed even the capa - syste m-level pers pective being deployed, offer significant advances beyond biliti es of t h e e m er gi n g 5 G s yst e ms. LTE, but may be unable to meet the full connec - The above discussion has recently motivat - o n 6 G sce narios a n d tivity de man ds of the future digital s ociety. There - ed researchers t o l o ok int o a ne w generati on of re q uire me nts, a n d select f ore, t his article disc usses tec h n ol ogies t hat will wireless net w orks, that is, 6 G syste ms, t o meet 6 G tec h n ologies t hat ca n evolve wireless net works to ward a sixth genera - the de mands of a fully c onnected, intelligent dig - satisfy t he m eit her by tion (6 G) and which we consider as enablers for ital w orld. Al ong these lines, the br oad purp ose i m pr oving t he 5 G desig n, several p otential 6 G use cases. We pr ovide a full- of this article is t o un derstan d which techn ol o - stack, syste m-level perspective on 6 G scenari os gies can identify 6 G net works and provide more or by intr o d ucing c o m- and require ments, and select 6 G technologies capable and vertical-specific wireless net w orking pletely ne w c o m m u nica- that can satisfy the m either by i mpr oving the 5 G s ol uti o ns. S p e ci fi c all y, t h e arti cl e c o nsi d ers s e v er al tio n para dig ms. d esi g n or b y i ntr o d u ci n g c o m pl et el y n e w c o m m u - p ot e nti al s c e n ari os f or f ut ur e c o n n e ct e d s yst e ms, ni c ati o n p ar a di g ms. and atte mpts to esti mate their key require ments in ter ms of thr oughput, latency, c onnectivity, and Intr oduct Io n other fact ors. I m p ortantly, we i dentify several use Each generation of mobile technology, fro m the cases that go beyond the perfor mance of the 5 G first t o the fifth (5 G), has been designed t o meet syste ms under develop ment today, and de mon - the needs of end users and net work operators, as str at e w h y it is i m p ort a nt t o t hi n k a b o ut t h e l o n g- sho wn in Fig. 1. Ho wever, no wadays societies are ter m evolution beyond 5 G. Our analysis suggests beco ming more and more data-centric, data-depen - that, in order t o meet these de mands, radically d e nt, a n d a ut o m at e d. R a di c al a ut o m ati o n of i n d us - ne w co m munication technologies, net work archi - tri al m a n uf a ct uri n g pr o c ess es will driv e pr o d u ctivit y. tectures, and deploy ment models will be needed. Autono mous syste ms are hitting our roads, oceans, I n p arti c ul ar, w e e n visi o n: a n d air s p a c e. Milli o ns of s e ns ors will b e e m b e d d e d • Novel disruptive co m munication technolo - into cities, ho mes, and production environ ments, gi es : Although 5 G net works have already a n d n e w s yst e ms o p er at e d b y arti fi ci al i nt elli g e n c e been designed to operate at extre mely high residing in l ocal cl oud and f og envir on ments will frequencies, for exa mple, in the m m Wave e n a bl e a pl et h or a of n e w a p pli c ati o ns. bands in N R, 6 G could very much benefit Co m munication net works will provide the fro m even higher spectru m technologies, for nervous syste m of these ne w s mart syste m par - exa mple, through terahertz and optical co m - adig ms. The de mands, ho wever, will be daunt - m u ni c ati o ns. ing. Net works will need to transfer much greater • Inn ov ative net w ork architectures: Des pite a mounts of data at much higher speeds. Further - 5 G advance ments to ward more efficient ing a trend already started in 4 G and 5 G, sixth net w ork setups, the heter ogeneity of future generation (6 G) connections will move beyond net work applications and the need for 3 D personalized co m munication to ward the full real - c overage calls f or ne w cell-less architectural izati on of the Internet of Things (I oT) paradig m, paradig ms, based on the tight integrati on of connecting not just people, but also co mputing different co m munication technologies, for res ources, vehicles, devices, weara bles, sens ors, both access and backhaul, and on the disag - a n d e v e n r o b oti c a g e nts [ 1]. gregati on and virtualizati on of the net w ork - 5 G made a significant step to ward developing i n g e q ui p m e nt. a l o w-l at e n c y t a ctil e a c c ess n et w or k b y pr o vi di n g • Integrating intelligence in the net w ork: We n e w a d diti o n al wir el ess n er v e tr a cts t hr o u g h: ex pect 6 G t o bring intelligence fr o m central -

M arc o Gior d a ni (c orres p o n din g a ut h or), Mic hele P olese (c orres p o n din g a ut h or), a n d Mic hele Z orzi are wit h t he U niversity of P a d ov a; Digital O bject Ide ntifier: M arc o Mezz avilla a n d S u n dee p R a n g a n are wit h Ne w Y ork U niversity, Br o okly n. M arc o Gior d a ni a n d Mic hele P olese are t he pri m ary c o- a ut h ors. 10.1109/ MC O M.001.1900411

IEEE Co m munications Magazine • March 2020 0163-6804/20/$25.00 © 2020 IEEE 5 5 A ut h ori z e d li c e n s e d u s e li mit e d t o: N e w Y or k U ni v er sit y. D o w nl o a d e d o n S e pt e m b er 1 8, 2 0 2 0 at 1 4: 1 9: 0 7 U T C fr o m I E E E X pl or e. R e stri cti o n s a p pl y.

GI O R DA NI_LAY O UT.indd 55 3/ 1 0/ 2 0 1 1: 1 1 A M T he late ncy re q uire me nt will hit s u b- millisec- 6 G o v ati o n i n n ond, and thousands 6 G wi ll c o ntri b ut e t o fi ll the gap bet ween beyond - Ne w Spe ct r u m 2020 societal and business de mands and what 5 G of sy nc hr o nize d vie w (and its predecessors) can su p p ort 1- 1 0 G b ps Disruptive Technologies a ngles will be neces- 5 G sary, as o p p ose d t o t he 1 0 0- 1000 M b ps C ell-l e s s N et w o r k s

2 M b p s 4 G Disaggreg at i o n a n d fe w re q uire d f or V R/ virt u ali z ati o n 64 Kbps A R. M ore over, t o f ully 2.4 Kbps 3 G 2 G Energy E ff ici e ncy realize a n i m mersive 1 G re m ote ex perie nce, all Artifi ci al I nt elli g e n c e I nt er n et of M a s s ive broadband and fi ve hu man senses are V oi c e c alli n g S M S I nt er n et A p pli c ati o n s I nt er n et of T hi n g s T o w ar d s a F ull y Di git al a n d C o n n e ct e d W orl d 1980 1990 2 0 0 0 2 0 1 0 2020 2025-2030 ti m e destine d t o be digitize d Fig ur e 1. E v ol uti o n of c ell ul ar n et w or ks, fr o m 1 G t o 6 G, wit h a r e pr es e nt ativ e a p pli c ati o n f or e a c h g e n er ati o n. a n d tra nsferre d acr oss f ut ure net w orks, increas- ized c o m puting facilities t o end ter minals, A ug mented r e AlI t Y A n d v Irt uAl r e AlI t Y i ng t he overall target thereby providing concrete i mple mentation 4 G syste ms unlocked the potential of vid- data rate. to distributed learning models that have eo-over- wireless, one of the most data-hun- been studied fro m a theoretical point of gry a p plicati ons at the ti me. The increasing use vie w in a 5 G context. Unsupervised learning of strea ming and multi media services currently and kno wledge sharing will pro mote real- justifies the adoption of ne w spectru m (i.e., ti me net work decisions through prediction. m m Waves) to guarantee higher capacity in 5 G. Pri or pu blicati ons ( m ost n ota bly [ 2, 3]) have H o wever, this multi-gigabit-per-sec ond opp ortu- dis c uss e d 6 G c o m m u ni c ati o ns. T his arti cl e disti n c- nity is attracting ne w a p plicati ons that are m ore tively ad opts a syste matic appr oach in analyzing data-heavy than bi-di mensional multi media con- the research challenges associated with 6 G net- tent: 5 G will trigger the early a d o pti on of aug- w or ks, pr o vi di n g a f ull-st a c k p ers p e cti v e, wit h c o n- mente d/virtual reality ( A R/ V R). Then, just like siderati ons related t o s pectru m usage, physical, video-over- wireless saturated 4 G net works, the mediu m access, and higher layers, and net work pr oliferati on of A R/ V R a p plicati ons will de plete architectures and intelligence f or 6 G. We trans- the 5 G spectru m, and require a syste m capaci- fer i nt o o ur w ork a m ultifacete d critical s pirit t o o, ty above 1 Tb/s, as opposed to the 20 Gb/s tar- having selecte d, out of several p ossi ble inn ova- get defi ned f or 5 G [1]. Additi onally, t o meet the ti o ns, t h e s ol uti o ns t h at i n o ur vi e w s h o w t h e hi g h- latency require ments that enable real-ti me user est p ot e nti al f or f ut ur e 6 G s yst e ms. W hil e s o m e of interaction in the i m mersive environ ment, A R/ V R the m appear to be incre mental, we believe that cannot be co mpressed (coding and decoding is a the co mbination of breakthrough technologies ti me-c onsu ming pr ocess); thus, the per-user data and evolution of current net works deserves to rate needs t o t ouch the gigabit-per-sec ond, in be identifi ed as a ne w generati on of m obile net- c o ntr ast t o t h e m or e r el a x e d 1 0 0 M b /s 5 G t ar g et. w orks, as these s oluti ons have n ot been th or ough- ly addressed or cannot be properly included in h ol ogrAphIc t eleprese nce ( t elep ortAtI o n ) current 5 G standards develop ments, and there- The hu man tendency to connect re motely with fore will not be part of co m mercial 5 G deploy- i n cr e asi n g fi d elit y will p os e s e v er e c o m m u ni c ati o n ments. We expect our investigation to pro mote challenges in 6 G net w orks. Reference [4] details r es e ar c h e ff orts t o w ar d t h e d e fi niti o n of n e w c o m- the data rate require ments of a 3 D holographic munication and net working technologies to meet dis pl a y: a r a w h ol o gr a m, wit h o ut a n y c o m pr essi o n, t h e b ol d est r e q uir e m e nts of 6 G us e c as es. wit h c ol ors, f ull p ar all a x, a n d 3 0 f ps, w o ul d r e q uir e 4. 3 2 T b /s. T h e l at e n c y r e q uir e m e nt will hit s u b- milli- 6 g u s e c Ases second, and thousands of synchronized vie w angles 5 G presents tra de- off s on latency, energy, c osts, will be necessary, as opp osed t o the fe w required hard ware co mplexity, throughput, and end-to- f or V R / A R. M or e o v er, t o f ully r e ali z e a n i m m ersi v e en d relia bility. F or exa m ple, the re quire ments of re mote experience, all five hu man senses are des- m obile br oadband and ultra-reliable l o w-latency tine d t o be digitize d an d transferre d acr oss future co m munications are addressed by diff erent con- n et w or ks, i n cr e asi n g t h e o v er all t ar g et d at a r at e. fi gurati ons of 5 G net w orks. 6 G, on the c ontrary, will be devel o pe d t o j ointly meet stringent net- e h eAlt h w or k d e m a n ds ( e. g., ultr a- hi g h r eli a bilit y, c a p a cit y, 6 G will rev oluti onize the healthcare sect or, eli m- e ffi ci e n c y, a n d l o w l at e n c y) i n a h olisti c f as hi o n, i n inating ti me and space barriers through re mote vi e w of t h e f or es e e n e c o n o mi c, s o ci al, t e c h n ol o g- surgery and guaranteeing healthcare workflo w i c al, a n d e n vir o n m e nt al c o nt e xt of t h e 2 0 3 0 er a. opti mizati ons. Besides the high c ost, the cur- In this secti on, we revie w the characteristics re nt maj or li mitati o n is t he lack of real-ti me tac- and f oreseen require ments of use cases that, f or tile feed back [5]. M ore over, the pr oliferati on of t h eir g e n er alit y a n d c o m pl e m e nt arit y, ar e b eli e v e d e Health services will challenge the a bility t o meet t o well re present future 6 G services. Figure 2 pr o- their stringent quality of service ( Q oS) re quire- vides a co mprehensive vie w on the scenarios in ments, that is, continuous connection avail- ter ms of different key perfor mance indicators a bility ( 9 9. 9 9 9 9 9 perce nt relia bility), ultra-l o w ( K PIs). latency (sub- millisec ond), and m obility supp ort.

5 6 IEEE Co m munications Magazine • March 2020 A ut h ori z e d li c e n s e d u s e li mit e d t o: N e w Y or k U ni v er sit y. D o w nl o a d e d o n S e pt e m b er 1 8, 2 0 2 0 at 1 4: 1 9: 0 7 U T C fr o m I E E E X pl or e. R e stri cti o n s a p pl y.

GI O R DA NI_LAY O UT.indd 56 3/ 1 0/ 2 0 1 1: 1 1 A M 5 G: 1 0 0 M b ps Overc o ming t he AR/ VR 6 G: 1 Gbps Per- user data r at e snoitacilppaesehtegremlliwsesacesuG6 5 G: 2 0 G b ps P e a k d at a r at e boundaries bet ween 6 G: > 1 T b ps

gnimaertsRA/RVhtiwytilibomdennamnu,.g.e Telepresence 5 G: > 1 ms s hari ng Air-i nterface late ncy 6 G: ~ 1 0 0 µs t he real fact ory a n d t he Ex perience

e Health 5 G: 9 9 .9 9 9 % (3 2 byte, 1 ms late ncy) cy ber c o m p utatio nal 6 G: > 9 9 . 9 9 9 9 9 R eli a bilit y

5 G: > 1 ms s pace will e na ble I nter- e e I n d ustrystry 4.4.0 0 6 G: ~ 100 µs Air-i nterface late ncy net- base d diag n ostics, 5 G: 1 x Spectru m e ffici e ncy 6 G: 5x fro m 5 G c o ntr ol Re mot Un manned o nl y f or u n manned mobility mainte na nce, o peratio n, m o bilitb ili yt y 5 G: 500 k m/h 6 G: 1000 k m/h Vehicle speed a n d direct mac hine 5 G: 1 x 6 G: 5x fro m 5 G Spectru m e ffici e ncy c o m m u nicatio ns in a Pervasive 5 G: 100 Mbps 6 G: 1 Gbps Per- user data r at e t n g conn ec t vitv it y y 5 G: 1 0 6 connec tons/k m 2 c ost-effective, fl exible 6 G: 1 0 7 connec t o ns/k m 2 Nu mber of devices a n d effi cie nt way.

everything 5 G: 1 x Connec 6 G: 1 0- 1 0 0 x E nergy e ffici e ncy

Net work manage ment L at e nc y a n d r eli a bilit y Ca pacity

Fig ur e 2. R e pr es e nt ati o n of m ulti pl e K PIs of 6 G us e c as es, t o g et h er wit h t h e i m pr o v e m e nts wit h r es p e ct t o 5 G n et w or ks, usi n g d at a fr o m [ 1 – 9].

The increased spectru m availability, co mbined micr osec ond delay jitter, and giga bit- per-sec ond with the refi ne d intelligence of 6 G net w orks, will peak data rates f or A R/ V R industrial a p plicati ons g u ar a nt e e t h es e K PIs, t o g et h er wit h 5 – 1 0  g ai ns ( e. g., f or tr ai ni n g, i ns p e cti o n). i n s p e ctr al e ffi ci e n c y [ 1]. u n mAnned m o bIlI t Y p ervAsIve c o n nectIvItY T h e e v ol uti o n t o w ar d f ull y a ut o n o m o us tr a ns p ort a- M o bile tra ffi c is ex pecte d t o gr o w three-f ol d fr o m ti o n s yst e ms off ers s af er tr a v eli n g, i m pr o v e d tr affi c 2016 to 2021, pushing the nu mber of mobile m a n a g e m e nt, a n d s u p p ort f or i nf ot ai n m e nt, wit h a devices t o the extre me, with 10 7 devices per k m 2 m ar k et of U S $ 7 trilli o n [ 8]. C o n n e cti n g a ut o n o m o us in dense areas (up fro m 10 6 in 5 G) [1] and more v e hi cl es d e m a n ds u n pr e c e d e nt e d l e v els of r eli a bilit y than 125 billion devices world wide by 2030. 6 G a n d l o w l at e n c y (i. e., a b o v e 9 9. 9 9 9 9 9 p er c e nt a n d will c onnect pers onal devices, sens ors (t o i m ple- b el o w 1 ms, r es p e ctiv ely), e v e n i n ultr a- hi g h- m o bilit y ment the s mart city paradig m), vehicles, and s o s c e n ari os ( u p t o 1 0 0 0 k m/ h) t o g u ar a nt e e p ass e n g er on. This will stress alrea dy c ongeste d net w orks, s af et y, a r e q uir e m e nt t h at is h ar d t o s atisf y wit h exist- which will not provide connectivity to every ing technologies. Moreover, the increasing nu mber device while meeting the require ments of Fig. of s e ns ors p er v e hi cl e will d e m a n d i n cr e asi n g d at a 2. Moreover, 6 G net works will require a higher rates ( with tera bytes generate d per driving h our [ 9]), overall energy e ffi ciency (10–100  with res pect b e y o n d c urr e nt n et w or k c a p a cit y. I n a d diti o n, fl yi n g to 5 G) to enable scalable, lo w-cost deploy ments v e hi cl es ( e. g., dr o n es) r e pr es e nt a h u g e p ot e nti al f or with lo w environ mental i mpact and better cover- v ari o us s c e n ari os ( e. g., c o nstr u cti o n, fi rst r es p o n d ers). age. In dee d, while 8 0 percent of m o bile tra ffi c is S w ar ms of dr o n es will n e e d i m pr o v e d c a p a cit y f or generated indoors, 5 G cellular net works, which e x p a n di n g I nt er n et c o n n e ctivit y. I n t his p ers p e ctiv e, are being mainly deployed outdoors and may be 6 G will p a v e t h e w a y f or c o n n e ct e d v e hicl es t hr o u g h operating in the m m Wave spectru m, will hard- advances in hard ware, soft ware, and the ne w con- ly pr ovide ind o or c onnectivity as high-frequency n e ctivit y s ol uti o ns w e dis c uss lat er. ra di o signals cann ot easily penetrate dielectric This wi de diversity of use cases is a uni que m at eri als ( e. g., c o n cr et e). 6 G n et w or ks will i nst e a d characteristic of the 6 G paradig m, whose poten- pr ovide sea mless and pervasive c onnectivity in a tial will be fully unleashe d only thr ough break- variety of different c ontexts, matching stringent through technological advance ments and novel QoS require ments in outdoor and indoor scenar- n et w or k d esi g ns, as d es cri b e d i n t h e n e xt s e cti o n. i os wit h a c ost- a w ar e a n d r esili e nt i nfr astr u ct ur e. 6 g e nAblI n g t echn ol ogIes In d ustrY 4. 0 A n d r ob otIcs In this secti on, we present the techn ol ogies that 6 G will fully realize the In dustry 4. 0 rev oluti on are ra pidly e merging as ena blers of the KPIs f or st art e d wit h 5 G, t h at is, t h e di git al tr a nsf or m ati o n the 6 G scenari os f oreseen ab ove. In particular, of manufacturing through cyber physical syste ms Table 1 su m marizes potentials and challenges of and IoT services. Overco ming the boundaries each proposed technological innovation and sug- bet ween the real factory and the cyber co mputa- gests which of the use cases intr oduced earlier ti o n al s p a c e will e n a bl e I nt er n et- b as e d di a g n osti cs, they e mpo wer. Although so me of these innova- m ai nt e n a n c e, o p er ati o n, a n d dir e ct m a c hi n e c o m- tions have already been discussed in the context m u nicati o ns i n a c ost-effective, flexi ble a n d effi- of 5 G, they were deli berately left out of early 5 G cient way [6]. Auto mation co mes with its o wn set standards develop ments (i.e., Third Generation of re quire ments in ter ms of relia ble an d is ochr o- Partnership Project [3 GPP] N R Releases 15 and nous co m munication [7], which 6 G is positioned 16) and will likely n ot be i mple mented in c o m- t o address thr ough the disruptive set of techn ol o- mercial 5 G deploy ments because of technologi- gi es w e d es cri b e l at er. F or e x a m pl e, i n d ustri al c o n- cal li mitati ons or because markets are n ot mature tr ol requires real-ti me operati ons with guaranteed enough to support the m. In this section we con-

IEEE Co m munications Magazine • March 2020 5 7 A ut h ori z e d li c e n s e d u s e li mit e d t o: N e w Y or k U ni v er sit y. D o w nl o a d e d o n S e pt e m b er 1 8, 2 0 2 0 at 1 4: 1 9: 0 7 U T C fr o m I E E E X pl or e. R e stri cti o n s a p pl y.

GI O R DA NI_LAY O UT.indd 57 3/ 1 0/ 2 0 1 1: 1 1 A M Enabling Technology P ot e ntial Challenges Use cases Ne w spectru m Hig h ba n d widt h, s mall a nte n na size, Pervasive c o n nectivity, in d ustry 4.0, h ologra p hic Tera hertz Circ uit desig n, hig h pr o pagatio n loss focused bea ms telepresence L o w-c ost har d ware, lo w interfere nce, VL C Li mite d c overage, nee d f or RF u plink Pervasive c o n nectivity, e Healt h unlicensed spectru m Novel P HY techniques F ull d u plex C o ntin u o us TX/ R X a n d relaying Ma nage me nt of interfere nce, sc he d uling Pervasive c o n nectivity, in d ustry 4.0 O ut- of- ba n d c ha n nel esti matio n Flexible m ulti-s pectr u m c o m m u nicatio ns Nee d f or relia ble fre q ue ncy ma p ping Pervasive c o n nectivity, h ologra p hic tele prese nce Efficie nt m ultiplexing of c o m m u nicatio n Se nsing a n d localizatio n N ovel services a n d c o ntext- base d c o ntr ol e Healt h, u n ma n ne d m o bility, in d ustry 4.0 a n d localizatio n Innovative net work architectures M ulti-c o n nectivity a n d cell-less Sea mless m o bility a n d integratio n of Pervasive c o n nectivity, u n ma n ne d m o bility, Sc he d uling, nee d f or ne w net w ork desig n arc hitect ure differe nt kin ds of links h ologra p hic tele prese nce, e Healt h M o deling, t o p ology o pti mizatio n a n d Pervasive c o n nectivity, e Healt h, u n ma n ne d 3 D net w ork arc hitect ure U biq uit o us 3 D c overage, sea mless service energy efficiency m o bility L o wer c osts f or o perat ors f or massively- Hig h perf or ma nce f or P HY a n d M A C Pervasive c o n nectivity, h ologra p hic tele prese nce, Disaggregatio n a n d virt ualizatio n dense deploy ments pr ocessing i n d ustry 4.0, u n ma n ne d m o bility Advanced access-backhaul Flexible de ploy me nt o ptio ns, o ut d o or-t o- Scala bility, sc he d uling a n d interfere nce Pervasive c o n nectivity, e Healt h i ntegratio n i n d o or relaying E nergy- harvesting a n d lo w- E nergy-efficie nt net w ork o peratio ns, Nee d t o integrate e nergy s o urce Pervasive c o n nectivity, e Healt h p o wer o peratio ns resilie ncy c haracteristics in pr ot oc ols I nt ellig e nc e i n t h e n et w ork Lear ning f or value of I ntellige nt a n d a ut o n o m o us selectio n of Pervasive c o n nectivity, e Healt h, h ologra p hic C o m plexity, u ns u pervise d lear ning infor mation assess ment t he inf or matio n t o tra ns mit tele prese nce, in d ustry 4.0, u n ma n ne d m o bility Nee d t o desig n n ovel s haring K n o wle dge s haring S pee d u p lear ning in ne w sce narios Pervasive c o n nectivity, u n ma n ne d m o bility mechanis ms Distrib ute d intellige nce t o t he e n d p oints of User-ce ntric net w ork arc hitect ure Real-ti me a n d e nergy-efficie nt pr ocessing Pervasive c o n nectivity, e Healt h, in d ustry 4.0 t he net w ork N ot c o nsidere d in 5 G Wit h ne w feat ures/ca pa bilities in 6 G T a ble 1. Co mparison of 6 G enabling technologies and relevant use cases.

sider physical layer breakthr oughs, ne w architec - to m m Waves, bring to the extre me the potential tural an d pr ot oc ol s oluti ons, an d finally disru ptive of hi g h-fr e q u e n c y c o n n e cti vit y, e n a bli n g d at a r at es a p pli c ati o ns of arti fi ci al i nt elli g e n c e. on the order of hundreds of gigabits per second, in line with the boldest 6 G require ments. On the d Isr upt Iv e c o m mun Ic A t Io n t echnolog Ie s ot h er si d e, t h e m ai n iss u es t h at h a v e pr e v e nt e d t h e A ne w generation of mobile net works is generally a d o pti o n of t er a h ert z li n ks i n c o m m er ci al s yst e ms s o c h ar a ct eriz e d b y a s et of n o v el c o m m u ni c ati o n t e c h - far are pr opagati on l oss, m olecular abs orpti on, high nologies that provide unprecedented perfor mance penetrati on l oss, and engineering challenges f or (e.g., in ter ms of availa ble data rate an d latency) antennas and radi o fre quency ( RF) circuitry. As f or a n d c a p a biliti es. F or e x a m pl e, m assiv e m ulti pl e-i n p ut m m Waves, the propagation loss can be co mpen - multiple-output ( MI M O) and m m Wave co m muni - s at e d usi n g dir e cti o n al a nt e n n a arr a ys, als o e n a bli n g cations are both key enablers of 5 G net works. In s patial multi plexing with li mited interference. Fur - order t o meet the require ments described earlier, ther more, terahertz co m munication perfor mance 6 G net works are expected to rely on conventional can be maxi mized by operating in frequency bands s p e ctr u m (i. e., s u b- 6 G H z a n d m m W a v es) b ut als o n ot severely affected by m olecular abs orpti on [10], on frequency bands that have not yet been con - as sh o wn in Fig. 3. Finally, such high fre quencies, si d er e d f or c ell ul ar st a n d ar ds, n a m el y t h e t er a h ert z w h e n li mit e d t o i n d o or-t o-i n d o or s c e n ari os, e n a bl e b a n d a n d visi bl e li g ht c o m m u ni c ati o ns ( V L C). Fi g ur e n e w ki n ds of ultr a-s m all-s c al e el e ctr o ni c p a c k a gi n g 3 r e pr es e nts t h e p at h l oss f or e a c h of t h es e b a n ds s ol uti o ns f or t h e R F a n d a nt e n n a cir c uitr y. i n t y pi c al d e pl o y m e nt s c e n ari os i n or d er t o hi g hli g ht V L C: These have been proposed to co mple - the differences and opportunities that each portion ment RF co m munications by piggybacking on of t h e s p e ctr u m c a n e x pl oit. I n t h e f oll o wi n g p ar a - the wide adoption of cheap light e mitting diode graphs, we focus on the t wo novel spectru m bands (LE D) lu minaries. These devices can indeed t h at will b e us e d i n 6 G. q uickly s witch bet ween different light intensities Terahertz Co m munications: These operate t o m odulate a signal that can be trans mitted t o bet ween 100 G Hz and 10 T Hz [10] and, co mpared a pr oper receiver [12]. The research on VL C is

5 8 IEEE Co m munications Magazine • March 2020 A ut h ori z e d li c e n s e d u s e li mit e d t o: N e w Y or k U ni v er sit y. D o w nl o a d e d o n S e pt e m b er 1 8, 2 0 2 0 at 1 4: 1 9: 0 7 U T C fr o m I E E E X pl or e. R e stri cti o n s a p pl y.

GI O R DA NI_LAY O UT.indd 58 3/ 1 0/ 2 0 1 1: 1 1 A M Increas i n g e ne r g y a n d T he de nsity a n d hig h b a n d wi dt h access data rate of

Increas i n g wavelength tera hertz c o m m u nica- tio ns will increase t he Legacy Spect r u m Mi ll i meter Waves Tera Hertz Visi bl e Li g ht capacity de mands on 6 G Hz 30 G Hz 3 0 0 GHz 3 0 0 G Hz 1 0 T Hz 430 T Hz 7 7 0 T Hz t he u n derlying tra ns p ort net w ork, w hic h has 20 d B Pat hl o ss [ d B] 140 d B 7 0 d B Pat hl o ss [ d B] 1 5 0 d B 0 d B Pat hl o ss [ d B] 1 5 0 d B 00 W 0.002 0.004 0.006Received 0.008 Po 0.01 w er 0.012[ W] 0.014 0.016 0.0180. 0 2 0.02 W t o pr ovide b ot h m ore p oints of access t o fi ber L E D Micro Macro S mart City 1 2 0 0 2 1 0 L E D 2 [ W] a n d hig her ca pacity

1 2 0 er m] m] than today’s backhaul [ [ d B]

[ 1 0 0 1 0 - 2 ss 8 0

Received Po w net w orks. M ore over, Dista nce Dista nce

Pat hl o 8 6 8 6 2 0 1 0 1 0 - 5 4 4 t he wide ra nge of dif- 1 0 0 5 0 0 1 0 0 0 6 1 5 0 3 0 0 0. 1 5 1 0 2 2 Dista nce [ m] F re quency [ G Hz] F re quency [ T Hz] Y R o o m [ m] X R o o m [ m] fere nt c o m m u nicatio n Fig ur e 3. Path loss for sub-6 G Hz, m m Wave, and terahertz bands, and received po wer for VL C. The sub-6 tec h n ologies availa ble G H z a n d m m W a v e p at h l oss f oll o ws t h e 3 G P P m o d els c o nsi d eri n g b ot h li n e- of-si g ht ( L o S) a n d n o n- L o S will increase t he heter o- ( NL oS) c onditi ons, while L oS- only is c onsidered f or terahertz [10] and VL C [11]. ge neity of t he net w ork, w hic h will nee d t o be more mature than that on terahertz co m munica- and depl oy ments t o av oid interference, as well as ti ons, als o thanks t o l o wer c ost of experi mental n o v el r es o ur c e s c h e d ul er d esi g ns [ 1 3]. managed. platf or ms. As re p orte d in Fig. 3, VL C have li mite d Novel Channel Esti mation Techniques (e.g., c overage range, require an illu minati on s ource, Out-of-Band Esti mation and Co mpressed Sensing): and suff er fr o m sh ot n oise fr o m other light s ourc- C h a n n el esti m ati o n f or dir e cti o n al c o m m u ni c ati o ns es (e.g., the sun), and thus can be m ostly used will be a key co mponent of co m munications at indoors [12]. Moreover, they need to be co mple- m m Waves and terahertz frequencies. Ho wever, it is mented by RF for the uplink. Nonetheless, VL C di ffi c ult t o d esi g n e ffi ci e nt pr o c e d ur es f or dir e cti o n al c ould be used t o intr oduce cellular c overage in co m munications, considering multiple frequency indoor scenarios, which, as mentioned previously, bands and p ossibly a very large band width. There- is a use case that has n ot been pr operly addressed f or e, 6 G s yst e ms will n e e d n e w c h a n n el esti m ati o n b y c ell ul ar st a n d ar ds. techniques. For exa mple, out-of-band esti mation Although standardization bodies are pro- (e. g., f or t he a n g ular directi o n of arrival of t he si g- m oting study ite ms that are oriented t o ward the nal) can i mprove the reactiveness of bea m manage- investigati on of terahertz and VL C s oluti ons f or m e nt b y m a p pi n g t h e o m ni dir e cti o n al pr o p a g ati o n future wireless syste ms (i.e., IEEE 802.15.3d and of s u b- 6 G H z si g n als t o t h e c h a n n el esti m ati o n f or 802.15.7, respectively), these techn ol ogies have m m W a v e fr e q u e n ci es [ 1 4]. Si mil arly, giv e n t h e s p ar- n ot yet been included in a cellular net w ork stan- sit y i n t er ms of a n g ul ar dir e cti o ns of m m W a v e a n d dard, and will be targeting beyond 5 G use cases. t er a h ert z c h a n n els, it is p ossi bl e t o e x pl oit c o m pr es- M ore over, a d diti o nal researc h is still re q uire d t o siv e s e nsi n g t o esti m at e t h e c h a n n el usi n g a r e d u c e d enable 6 G m obile users t o operate in the tera- n u m b er of s a m pl es. hertz and VL C spectra, including hard ware and Sensing and Net work-Based Localization: T h e alg orith ms f or fl exi ble multi- bea m ac quisiti on and usage of RF signals t o enable si multane ous l ocal- tracking in n on-line- of-sight ( NL oS) envir on ments. ization and mapping has been widely studied, but Besides the ne w spectru m, 6 G will als o trans- such capabilities have never been deeply inte- f or m wir el ess n et w or ks b y l e v er a gi n g a s et of t e c h- grated with the operati ons and pr ot oc ols of cel- n ol o gi es t h at h a v e b e e n e n a bl e d b y r e c e nt p h ysi c al lular net w orks. 6 G net w orks will ex pl oit a unifi ed layer an d circuit research, but are n ot part of 5 G. interface for localization and co m munications to T h e f oll o wi n g will b e k e y e n a bl ers f or 6 G. i mprove control operations, which can rely on Full- Duplex Co m munication Stack: Wit h context infor mation to shape bea mfor ming pat- full-duplex co m munications, the transceivers t er ns, r e d u c e i nt erf er e n c e, pr e di ct h a n d o v ers, a n d will be ca pa ble of receiving a signal while als o offer i n n ovative user services (e.g., f or ve hic ular trans mitting, t hanks t o caref ully designe d self-in- a n d e H e alt h a p pli c ati o ns). terference-su p pressi on circuits [13]. Practical full-duplex depl oy ments require inn ovati ons in In n ovAtIve n et work A rc hItect ures antenna and circuit design t o reduce the cr oss- T h e disr u pti o n br o u g ht b y t h e c o m m u ni c ati o n t e c h- talk bet ween trans mitter an d receiver circuits in a n ol ogies described earlier will enable a ne w 6 G wir el ess d e vi c e; t h us, t h e y h a v e n ot b e e n i n cl u d e d net w ork architecture, but als o p otentially re quire in current cellular net w ork s pecifi cati ons. Future structural updates with respect t o current m obile technology advance ments, ho wever, will enable n et w or k d esi g ns. F or e x a m pl e, t h e d e nsit y a n d t h e concurrent do wnlink and uplink trans mission to high access data rate of terahertz co m munications increase the multiplexing capabilities and the will increase the capacity de mands on the under- overall syste m throughput without using addi- l yi n g tr a ns p ort n et w or k, w hi c h h as t o pr o vi d e b ot h tional band width. Nonetheless, 6 G net works will m or e p oi nts of a c c ess t o fi b er a n d a hi g h er c a p a c- need careful planning f or full-duplex pr ocedures ity than today’s backhaul net works. Moreover, the

IEEE Co m munications Magazine • March 2020 5 9 A ut h ori z e d li c e n s e d u s e li mit e d t o: N e w Y or k U ni v er sit y. D o w nl o a d e d o n S e pt e m b er 1 8, 2 0 2 0 at 1 4: 1 9: 0 7 U T C fr o m I E E E X pl or e. R e stri cti o n s a p pl y.

GI O R DA NI_LAY O UT.indd 59 3/ 1 0/ 2 0 1 1: 1 1 A M virt u ali z ati o n c o n c e pts. M or e o v er, c urr e nt 5 G st u d- ies have n ot yet addressed the challenges related t o the design of disaggregated architectures that can operate under the higher c ontr ol latency that might be intr oduced by centralizati on, and t o the security of virtualized net w ork functi ons, which could be subjected to cyber-attacks. 6 G net works will bri n g dis a g gr e g ati o n t o t h e e xtr e m e b y virt u al- izing mediu m access control ( M A C) and physical (P HY) layer co mponents, which currently require d e di c at e d h ar d w ar e i m pl e m e nt ati o ns, a n d r e ali zi n g l o w- c ost distri b ut e d pl atf or ms wit h j ust t h e a nt e n- n as a n d mi ni m al pr o c essi n g. T his will d e cr e as e t h e Fig ur e 4. Ar c hit e ct ur al i n n o v ati o ns i ntr o d u c e d i n 6 G n et w or ks. cost of net working equip ment, making massively d e ns e d e pl o y m e nt e c o n o mi c ally f e asi bl e. wi d e r a n g e of di ff er e nt c o m m u ni c ati o n t e c h n ol o gi es Advanced Access-Backhaul Integration: T h e a v ail a bl e will i n cr e as e t h e h et er o g e n eit y of t h e n et- massive data rates of the ne w 6 G access tech- w or k, w hi c h will n e e d t o b e m a n a g e d. nologies will require adequate gro wth of the T h e m ai n ar c hit e ct ur al i n n o v ati o ns t h at 6 G will backhaul capacity. Moreover, terahertz and VL C intr o duce are descri be d in Fig. 4. In this c ontext, depl oy ments will increase the density of access we envision the introduction and/or deploy ment p oints, which need backhaul c onnectivity t o their of t h e f oll o wi n g p ar a di g ms. neighbors and the core net work. The huge capac- Tight integration of multiple frequencies and ity of 6 G techn ol ogies can thus be expl oited f or co m munication technologies and cell-less archi- self- backhauling s oluti ons, where the ra di os in the tect ure: 6 G devices will support a nu mber of het- base stations provide both access and backhaul. erogeneous radios in the devices. This enables W hil e a si mil ar o pti o n is alr e a d y b ei n g c o nsi d er e d m ulti- c o n n e cti vit y t e c h ni q u es t h at c a n e xt e n d t h e f or 5 G, the scale of 6 G depl oy ments will intr o- current b oundaries of cells, with users c onnected d u c e n e w c h all e n g es a n d o p p ort u niti es; f or e x a m- t o the net w ork as a wh ole (i.e., thr ough multiple ple, the net works will need higher autono mous c o m pl e m e nt ar y t e c h n ol o gi es) a n d n ot t o a si n gl e c o n fi g ur ati o n c a p a biliti es. c ell. T h e c ell-l ess n et w or k pr o c e d ur es will g u ar a nt e e Energy- Harvesting Strategies for Lo w Po wer sea mless mobility support without overhead due Consu mption Net work Operations: Inc or p orating to handovers ( which might be frequent when con- energy-harvesting mechanis ms into 5 G infrastruc- si d eri n g s yst e ms at t er a h ert z fr e q u e n ci es), a n d will tures currently faces several issues, inclu ding c oex- pr ovide Q oS guarantees that are in line with the istence with the co m munications, and efficiency m ost c h all e n gi n g m o bilit y r e q uir e m e nts e n visi o n e d l oss when c onverting harvested signals t o electric f or 6 G, as in the vehicular scenari os. The devices c urr e nt. Gi v e n t h e s c al e e x p e ct e d i n 6 G n et w or ks, will b e a bl e t o s e a ml essly tr a nsiti o n a m o n g di ff er e nt it is necessary t o design syste ms where b oth the heterogeneous links (e.g., sub-6 G Hz, m m Wave, circuitry and the c o m municati on stack are devel- t er a h ert z, a n d V L C) wit h o ut m a n u al i nt er v e nti o n or oped with energy a wareness in mind. One option c o n fi g ur ati o n. Fi n all y, a c c or di n g t o t h e s p e ci fi c us e is using energy-harvesting circuits t o all o w devic- case, the user may als o c oncurrently use different es t o be self- p o were d, which c oul d be critical, f or n et w or k i nt erf a c es t o e x pl oit t h eir c o m pl e m e nt ar y e x a m pl e, t o e n a bl e o ff - gri d o p er ati o ns, l o n g-l asti n g c h ar a ct eristi cs, f or e x a m pl e, t h e s u b- 6 G H z l a y er f or I o T d e vi c es a n d s e ns ors, or l o n g st a n d- b y i nt er v als c o ntr ol, a n d a t er a h ert z li n k f or t h e d at a pl a n e. f or devices and equip ment that are rarely used. 3 D Net work Architecture: 5 G net works (and previous generations) have been designed to pro- Inte grAtI n g I ntellI ge nce I n t h e n et work vi de c onnectivity f or an essentially bi- di mensi onal The co mplexity of 6 G co m munication technolo- s p a c e, t h at is, n et w or k a c c ess p oi nts ar e d e pl o y e d gi es a n d n et w or k d e pl o y m e nts will pr o b a bly pr e v e nt t o o ff er c o n n e cti vit y t o d e vi c es o n t h e gr o u n d. O n closed-for m and/or manual opti mizations. While the contrary, we envision future 6 G heterogeneous i nt elli g e nt t e c h ni q u es i n c ell ul ar n et w or ks ar e alr e a d y architectures to provide three-di mensional cover- being discussed for 5 G, we expect 6 G deploy ments a g e, t h er e b y c o m pl e m e nti n g t err estri al i nfr astr u c- to be much denser (i.e., in ter ms of nu mber of t ures with n on-terrestrial platf or ms (e.g., dr ones, access points and users) and more heterogeneous b all o o ns, a n d s at ellit es). M or e o v er, t h es e el e m e nts (i n t er ms of i nt e gr ati o n of di ff er e nt t e c h n ol o gi es a n d could also be quickly deployed to guarantee sea m- a p pli c ati o n c h ar a ct eristi cs), a n d h a v e stri ct er p erf or- less service c o nti n uity a n d relia bility, f or exa m ple, mance require ments with respect to 5 G. Therefore, in rural areas or during events, av oi ding the o per- i nt elli g e n c e will pl a y a m or e pr o mi n e nt r ol e i n t h e ational and manage ment costs of al ways-on, fi xed net w ork, g oing bey ond the classificati on and pre- infrastructures. Despite such pr o mising opp ortu- di cti o n t as ks t h at ar e b ei n g c o nsi d er e d f or 5 G s ys- nities, there are vari ous challenges t o be s olve d t e ms. N oti c e t h at t h e st a n d ar d m a y n ot s p e cif y t h e b ef or e fl yi n g pl atf or ms c a n eff e cti v el y b e us e d i n t e c h ni q u es a n d l e ar ni n g str at e gi es t o b e d e pl o y e d i n wir el ess n et w or ks, f or e x a m pl e, air-t o- gr o u n d c h a n- net works, but data-driven approaches can be seen n el m o d eli n g, t o p ol o g y a n d tr aj e ct or y o pti mi z ati o n, as tools that net work vendors and operators can resource manage ment, and energy e ffi ciency. us e t o m e et t h e 6 G r e q uir e m e nts [ 1 5]. I n p arti c ul ar, Disaggregation and Virtualization of the Net- 6 G r es e ar c h will b e ori e nt e d t o w ar d t h e f oll o wi n g working Equip ment: Even though net works have as p e cts. r e c e ntly st art e d t o tr a nsiti o n t o w ar d t h e dis a g gr e g a- Learning Techniques for Data Selection and ti o n of o n c e- m o n olit hi c n et w or ki n g e q ui p m e nt, t h e Feat ure Extracti o n: T h e l ar g e v ol u m e of d at a g e n- 3 GPP does not directly specify ho w to introduce erated by future c onnected devices (e.g., sens ors

6 0 IEEE Co m munications Magazine • March 2020 A ut h ori z e d li c e n s e d u s e li mit e d t o: N e w Y or k U ni v er sit y. D o w nl o a d e d o n S e pt e m b er 1 8, 2 0 2 0 at 1 4: 1 9: 0 7 U T C fr o m I E E E X pl or e. R e stri cti o n s a p pl y.

GI O R DA NI_LAY O UT.indd 60 3/ 1 0/ 2 0 1 1: 1 1 A M [2] W. Saad, M. Bennis, and M. Chen, “ A Vision of 6 G Wire - in autono mous vehicles) will put a strain on co m - W hile intellige nt tec h - m u ni c ati o n t e c h n ol o gi es, w hi c h c o ul d n ot g u ar a n - less Syste ms: Applications, Trends, Technologies, and Open R es e ar c h Pr o bl e ms,” IEEE Net work ( E arl y A c c ess), 2 0 1 9. niq ues in cellular net - tee t he re q uire d q uality of service. It is t heref ore [ 3] E. Calva nese Stri nati et al ., “ 6 G: The Next Fr ontier,” I E E E funda mental t o discri minate the value of inf or - V e hi c. T e c h. M a g ., v ol. 1 4, n o. 3, S e pt. 2 0 1 9, p p. 4 2 – 5 0. w orks are alrea dy being mati on t o maxi mize the utility f or the en d users [ 4] X. X u et al ., “3 D Holographic Display and Its Data Trans - missi on Re quire ment,” Pr oc. I nt’l. C o nf. I nf o. Ph ot o nics a n d disc usse d f or 5 G, we with (li mite d) net w ork res o urces. In this c ontext, Optical Co m mun ., O ct. 2 0 1 1, p p. 1 – 4. m a c hi n e l e ar ni n g ( M L) str at e gi es c a n e v al u at e t h e [5] Q. Zhang, J. Liu, and G. Zhao, “To wards 5 G Enabled Tac - expect 6 G deploy ments degree of c orrelati on in o bservati ons, or extract tile R ob otic Telesurgery,” ar Xiv preprint ar Xiv:1803.03586, to be much denser and features fr o m input vect ors and predict the a p os - 2 0 1 8. [6] J. Lee, B. Bagheri, and H.- A. Ka o, “ A Cyber-Physical Syste ms m ore heter oge ne o us, teri ori pr o ba bility of a se quence given its entire Ar c hit e ct ur e f or I n d ustr y 4. 0- B as e d M a n uf a ct uri n g S yst e ms,” hist ory. M ore over, in 6 G, unsupervised and rein - Manufacturing Letters , v ol. 3, J a n. 2 0 1 5, p p. 1 8 – 2 3. a n d have stricter per - force ment learning approaches do not need label - [ 7] M. W ollschlaeger, T. Sauter, an d J. Jas perneite, “The Futur f or ma nce re q uire me nts ing and can be used to operate the net work in a of Industrial Co m munication: Auto mation Net works in the Er a of t h e I nt er n et of T hi n gs a n d I n d ustr y 4. 0,” I E E E I n d. El e c - wit h res pect t o 5 G. truly autono mous fashion. tr o n. M a g. , v ol. 1 1, n o. 1, M ar. 2 0 1 7, p p. 1 7 – 2 7. Inter- User Inter- Operator Kno wledge Sharing: [ 8] N. L u et al. , “ Connected Vehicles: Solutions and Challenges,” T heref ore, intellige nce Spectru m and infrastructure sharing is beneficial I E E E I nt er n et of T hi n gs J ., v ol. 1, n o. 4, A u g. 2 0 1 4, p p. 2 8 9 – 9 9. in cellular net w orks t o maxi mize the multiplexing [ 9] J. C h oi et al ., “ Milli meter- Wave Vehicular Co m munication to will play a m ore Support Massive Auto motive Sensing,” IEEE Co m mun. Mag. , ca pa bilities. With learning- driven net w orks, o pera - v ol. 5 4, n o. 1 2, D e c. 2 0 1 6, p p. 1 6 0 – 6 7. pr o mine nt r ole in t he tors and users can also share learned/processed [10] J. M. J ornet and I. F. Akyildiz, “ Channel M odeling and net w ork, g oing bey o n d representations of specific net work deploy ments Capacity Analysis for Electro magnetic Wireless Nanonet - and/or use cases, for exa mple, to speed up the w or ks i n t h e T er a h ert z B a n d,” IEEE Trans. Wireless Co m mun. , t he classificatio n a n d v ol. 1 0, n o. 1 0, O ct. 2 0 1 1, p p. 3 2 1 1 – 2 1. n et w or k c o n fi g ur ati o n i n n e w m ar k ets or t o b ett er [11] T. Ko mine and M. Nakaga wa, “Funda mental Analysis for pre dictio n tasks t hat are adapt to ne w unexpected operational scenarios. Visible-Light Co m munication Syste m Using LE D Lights,” T h e tr a d e- o ffs i n l at e n c y, p o w er c o ns u m pti o n, s ys - IEEE Trans. Consu mer Electron. , v ol. 5 0, n o. 1, Fe b. 2 0 0 4, being c o nsidere d f or 5 G te m overhead, and cost will be studied in 6 G for p p. 1 0 0 – 0 7. syste ms. [ 1 2] P. H. P at h a k et al. , “ Visi bl e Li g ht C o m m u ni c ati o n, N et w or k - both onboard and edge-cloud-assisted solutions. ing, and Sensing: A Survey, P otential and Challenges,” I E E E User- Centric Net work Architecture: M L- dri v - Co m mun. Surveys & Tutorials , v ol. 1 7, n o. 4, 4t h qtr. 2 0 1 5, e n net w orks are still i n t heir i nfa ncy, b ut will be a p p. 2 0 4 7 – 7 7. funda mental co mponent of co mplex 6 G syste ms, [ 1 3] S. G o y al et al. , “ F ull D u pl e x C ell ul ar S yst e ms: Will D o u bli n g Interference Prevent Doubling Capacity?” IEEE Co m mun. w hic h e nvisi o n distri b ute d artificial i ntelli ge nce, M a g ., v ol. 5 3, n o. 5, M a y 2 0 1 5, p p. 1 2 1 – 2 7. t o i mple ment a fully user-centric net w ork archi - [14] A. Ali, N. G onzález-Prelcic, and R. W. Heath, “ Milli meter tecture. In this way, en d ter minals will be a ble Wave Bea mselection Using Out-of-Band Spatial Infor ma - to make autono mous net work decisions based ti o n, ” IEEE Trans. Wireless Co m mun. , v ol. 1 7, n o. 2, Fe b. 2 0 1 8, p p. 1 0 3 8 – 5 2. on the outco mes of previous operations without [15] M. Wang et al. , “ Machine Learning for Net working: Work - c o m m u ni c ati o n o v er h e a d t o a n d fr o m c e ntr ali z e d flo w, Advances, and Opportunities,” IEEE Net work , v ol. 3 2, controllers. Distributed methods can process ML n o. 2, M ar. 2 0 1 8, p p. 9 2 – 9 9. al g orit h ms i n r e al ti m e, t h at is, wit h a s u b- ms l at e n - cy, as re quired by several 6 G services, there by b Io g r A p h Ie s yielding more responsive net work manage ment. M arc o G I o r d a nI [ M’20] was a Ph. D. student in infor mation engineering at the University of Pad ova, Italy (2016-2019), c o ncl us Io n s where he is no w a postdoctoral researcher and adjunct profes - sor. He visited the Ne w York University ( NY U) and the Toyota In this article, we revie w use cases and techn ol - Infotechnology Center, Inc. In 2018 he received the Daniel ogies that we believe will characterize 6 G net - E. Noble Fello wship A ward fro m the IEEE Vehicular Technol - works. Table 1 su m marizes the main challenges, ogy Society. His research focuses on protocol design for 5 G m m W a v e c ell ul ar a n d v e hi c ul ar n et w or ks. p otentials, and use cases of each enabling tech - n ol o g y. 6 G wir el ess r es e ar c h c a n disr u pt t h e tr a di - M I c hElE P olEsE [ M’20] was a Ph. D. student in infor mation engi - ti o nal cell ular net w orki ng para dig ms t hat still exist n e eri n g at t h e U niv ersit y of P a d o v a ( 2 0 1 6- 2 0 1 9), w h er e h e is n o w in 5 G, intr oducing, f or exa mple, supp ort f or tera - a postdoctoral researcher and adjunct professor. He visited N Y U, AT &T Labs, and Northeastern University. His research focuses on hertz a n d visi ble li g ht s pectra, cell-less a n d aerial protocols and architectures for 5 G m m Wave net works. ar c hit e ct ur es, a n d m assi v e distri b ut e d i nt elli g e n c e, a mong others. These technologies, ho wever, are M arc o M Ezz avIll a [S M’19] is a research scientist at the N Y U n ot market-ready: this represents a unique opp or - Tandon School of Engineering. He received his Ph. D. (2013) in infor mation engineering fro m the University of Padova. His tunity f or the wireless research c o m munity t o f os - research focuses on design and validation of co m munication ter inn ovati ons that will ena ble unf oreseen digital pr ot o c ols a n d a p pli c ati o ns of 4 G / 5 G t e c h n ol o gi es. us e c as es f or t h e s o ci et y of 2 0 3 0 a n d b e y o n d. s u n dEE P r a n G a n [F’15] is an E CE professor at N Y U and ass oci - A ckno wledg ments ate director of NY U WI RELESS. He received his Ph. D. fro m the University of Calif ornia, Berkeley. In 2000, he c o-f ounded ( with T his w or k w as p arti all y s u p p ort e d b y NI S T t hr o u g h f our others) Flari on Techn ol ogies, a s pin off of Bell La bs that A ward No. 70 N A NB17 H166, by the U.S. AR O developed the first cellular OF D M data syste m. It was acquired under Grant no. W911 NF1910232, by MI UR (Ital - by Qualco m m in 2006, where he was a director of engineering pri or t o j oi ni n g N Y U i n 2 0 1 0. ian Ministry for Education and Research) under the initiative “ Depart ments of Excellence” (La w M I c hElE z o r zI [F’07] is with the Infor mation Engineering 232/2016), by NSF grants 1302336, 1564142, Depart ment of the University of Padova, focusing on wireless a n d 1 5 4 7 3 3 2, t h e S R C, a n d t h e i n d ustri al a ffili at es co m munications research. He was Editor-in- Chief of IEEE Wire - less Co m munications fro m 2003 to 2005, IEEE Transactions on of N Y U WI R E L E S S. Co m munications fro m 2008 to 2011, and IEEE Transacti ons on Cognitive Co m munications and Net working fro m 2014 to r efere nces 2018. He served/is serving Co mSoc as a Me mber-at-Large of [1] Z. Zhang et al ., “6 G Wireless Net works: Vision, Require - the Board of Governors fro m 2009 to 2011, as Director of Edu - ments, Architecture, and Key Techn ol ogies,” IEEE Vehic. c ati o n a n d Tr ai ni n g fr o m 2 0 1 4 t o 2 0 1 5, a n d as Dir e ct or of J o ur - T e c h. M a g ., v ol. 1 4, n o. 3, S e pt. 2 0 1 9, p p. 2 8 – 4 1. n als fr o m 2 0 2 0 t o 2 0 2 1.

IEEE Co m munications Magazine • March 2020 6 1 A ut h ori z e d li c e n s e d u s e li mit e d t o: N e w Y or k U ni v er sit y. D o w nl o a d e d o n S e pt e m b er 1 8, 2 0 2 0 at 1 4: 1 9: 0 7 U T C fr o m I E E E X pl or e. R e stri cti o n s a p pl y.

GI O R DA NI_LAY O UT.indd 61 3/ 1 0/ 2 0 1 1: 1 1 A M