 Per mi és un plaer introduir-nos l'Àngel Lothano. Jo crec que l'Àngel Lothano necessita una introducció a tot. I avui dia he de parlar de 5G. Aquest és l'únic hot topic que tenim ara, en l'arrenar de les comunicacions. Però, abans de dir alguna cosa sobre l'Àngel Lothano, volia mostrar una documentada que vaig trobar a la nit quan vaig pensar una mica sobre l'Àngel. Em vaig sorprèn perquè vaig trobar aquest publicament en januari 2000, eliero. Està bé. A tu et vaig trobar un plaer. Va ser l'aniversari de l'experiment markets de Marconi en l'U.S. Marconi va ser la pioner que va fer alguna gelat confrontada a la comunicació de la via i les dades. Aquest va ser el lloc. I l'Àngel va mirar per a tu i a mi, però no era possible, perquè la qualitat de la foto no és... Quan hi ha una d'aquestes imatges de processar gurus per enganxar les imatges... Jo pensava en la mateixa, doncs... Comoma o marcemo, unes vegades... Tant de reconstrucció. Aquest was the first time we met, I think. We were just... We were younger, I'd say. A little bit, it's not a long time at all. So I wanted to share with you and with all the audience this picture at the day we first met, and some years after, a few years after, we met here at the UPF. So it was a pleasure to have Angel here, joining the department seven or eight years ago. He has been really committed with the university, not only building the research group and doing brilliant research with us, but also with a high level of commitment with the UPF, being vice-rector for research, almost... Six years, yeah. Six years, so. Now he is in his sabbatical, so, right? But he's here sharing with us the last research thoughts about 5G and see how this is going to evolve in the next year. So thank you for accepting the invitation and... Okay, thank you, Mr. Thomas. So these seminars are a good opportunity for us to showcase our research, and especially for those of us in minority topics. Because even though this is the department on information and communication technologies, those of us doing communications are actually a small minority here. So this is a one-com opportunity. And just to show you graphically where my research is positioned within the IEEE, my view on this communication theory, which is forming within the IEEE Communication Society, although having some intersection with the Information Theory and Synoprocessing. Also, there is a very large committee on communication theory, a tipus on committee, which I actually chaired for a few years. And my work revolves mostly around the activities, the publications and the conferences organized by this committee. Okay, so without further ado, let me get into the talk itself. And the subject are wireless networks. The talk has six sections, about 40, 45 minutes. So my challenge ironically is to try to keep your attention away from your smartphones for about 40 minutes when I explain things about those very smartphones. And as a general comment, before I start, I'm not going to specifically talk about Wi-Fi. I'm going to focus on why there are networks that can be accessed anywhere. Wi-Fi is a great complement for hotspots and we need it. But it's kind of unpredictable and spotty. And I'm sure you've all had your share of frustration with it. So let me go ahead and start with the first section, where I'll try to establish what those of us in communication theory, what we do for a living. So, there are two metrics that drive our work, which are in fact intertwined. They are the user bitrate and the area capacity. And here I've placed them within a triangle that contains the three resources that we need to communicate data. So the first is bandwidth, which is to say spectrum. The second is energy or what's the same power. And the third is computational complexity. So let me spend just a few minutes determining all these quantities and how they interplay. The user bitrate measures the number of bits per second that flow reliably to specific users, like yourselves. It's an important quantity, especially for certain applications, but not as important as the area capacity, which measures the aggregate bitrate over all users within a certain area, say a square kilometer, for instance. These two quantities measure your degree of satisfaction as consumers of data. A challenge we have here are the expectations. And this chart makes that point. Just ignore all the acronyms, which correspond to specific standards. Just look at the two lines. So the blue line depicts the evolution over time of the user bitrates over wireless networks. So we've had ISDN, then ADSL, now fiber optics. And the yellow line depicts the evolution over time of the bitrates over wireless networks. And the issue is that wireless is always one order of magnitude behind. This is not surprising, even that a radio channel is much more challenging than a cable channel. But it does mean that your expectations are of a certain performance that you're used to, and you expect it over our networks as well. So this is what we're up against. This is a quote from a report by the company Ericsson, which I think describes very well these expectations. It says that wait times during web page and video loads causes mobile users heart rates to increase 38%, which apparently is equivalent to the anxiety of watching a horror movie alone, which is a very remarkable baseline for stress. And then once a video begins and an additional pause increases frustration dramatically. So this is what we're up against as engineers. Now let's talk about the resources we need to quench these thirsts for data that you have. The first one, the first thing that we need is radio spectrum. And even though there is an infinite amount of it, only a finite amount is actually usable. The portion roughly between one and six gigahertz, which we usually call the radio window. Below this window, the wavelength is too long, so the antennas are too big to be practical. And above it, the signals don't propagate very well. They suffer a lot of attenuation. In fact, that's what the chart depicts. So this chart depicts the attenuation in the V-perkylometer. So, as you can see, it's almost zero over this radio window, and then it spikes up into tens or even hundreds of V-perkylometers. Once you exceed a few gigahertz. So because the amount of spectra is limited, and we require more and more of it, it has become very expensive. It has become expensive to license it for exclusive use. So these days, it's roughly at one euro per megahertz per population unit. So a company like Vodafone or Tenafonica say they want to license 100 megahertz to use in Catalonia, which has 7 million people. They have to pony up 700 million euros, just to have the bandwidth to transmit. It's a lot of money, but that gives them exclusive access to this bandwidth, and that avoids all the interference, contamination that Wi-Fi suffers. Okay, so that's bandwidth. The second thing we need is energy to radiate, and well, what do I have to tell you that you don't know already? The most sought after things these days are airports and coffee shops or power plaques. And charging stations are now becoming pretty common, especially in the U.S. For now, there are a few of charge, but I suspect that will change. In Amsterdam, they have at the airport these eco-friendly charging stations, where you have to pedal to charge your phone, and in a sense, having to labor for power is already a form of payment, in a sense. So the critical point when it comes to batteries is to be able to spend the whole day without recharging your device, and we're close to that point already. So it's a resource we have to manage carefully. If we don't want to have to force people to carry around the chargers and spare batteries with them, which would be undesirable. Battery knives are one side of the energy equation, the more obvious one, perhaps, for us, but there's another side, which is that the biggest operating expense of companies like Tenafonica or Vodafone these days is the electricity bill. And that gets passed on to us. So that's one of the reasons that things are so expensive here. So reducing power consumption is also very important for them, not to mention all the initiatives on green ICT to reduce power consumption worldwide. Okay, so we have spectrum, we have energy, and the third thing we need is computational complexity. And this one has various implications. The first one is energy again, not energy-radiated, but energy-spent processing signals. The second one is size. So a small one as the bias is very different than a tablet, for instance, or a smartphone, in terms of the real state you have to install processors. And the third is cost, which is a very important variable in the marketplace. So these are the quantities we are with, bitrate, anaerocapacity, versus bandwidth, energy, and complexity. And we combine them into figures of merit, per instance, by combining the bitrate in bits per segon with the bandwidth in hertz, you obtain something called spectrum efficiency in bits per segon per hertz, which measures how well you're using your spectrum. Or if you combine the bitrate in bits per segon with the power in watts, you get the energy efficiency in bits per segon per watt, or in bits per jou. So in our work, we mostly try to find new ways to push these ratios higher, so we can maximize the bitrate and the capacity with certain resources, or we can minimize the resources to achieve a certain performance. All right, let me move on to the second section. I'm a history freak, and I could spend all the time I haven't more talking about it, so I'll have to contain myself here. I do think that a bit of history helps understand why things are the way they are. So I'm going to spend about five minutes discussing some key steps in the evolution of Warrangas communications, and Mikhail Bratamarkoni, who's going to come up in the story. I did a small experiment at some point and asked a bunch of young people some questions about the origins of this technology that they are so addicted to, and the outcome was disappointing, not surprising, perhaps, but disappointing. Younger generations seem to associate Warrangas communications with people and companies, such as these, and others like them, and these players have certainly played a part in the story, but they really are just the icing on the cake. These players have transformed the mobile phone into a dazzling computational platform that has exceeded anyone's expectations and that is an essential part of our lives. That's true. And yet, the concept of a smartphone is hardly new. As early as 1993, IBM launched a smart device called the Simon, which had the amazing amount of one megabyte of memory, and a touchscreen, Nokia 496, with their Nokia 9000, which was a similar device, both were very big, like, brick-size devices, very expensive and very heavy, and their performance was laughable by today's standards. They failed commercially, but not because the concept was wrong, just because it was ahead of its time. Now, 20 years later, of course, the time was right, i altres jugadors van fer el concepte d'un smartphone per ferir-se, com a favor d'avui. Deixem-ne, aquest experiment amb els joves em va fer pensar que hi havia un article recentment sobre l'espectrum d'Aitibán que es va aportar a l'espectrum de la direcció. I vaig pensar en què és el que ha de fer els brexos que han definit on som i on som, i vaig arribar a la segona de 5 milions. Per tant, potser el primer brexó va ser en l'1860s, quan Maxwell unified the electricity and magnetism in a wonderful set of equations. I spent some time at the University of Edinburgh, which is Maxwell's alma mater, and every morning I literally walked over the equations on the way to my office. These equations predicted the existence of electromagnetic waves, a prediction that was then verified by the German gentleman, Heinrich Hertz, a few years later, who then famously stated that he did not think that the world's waves he had invented or discovered would have any practical applications. He couldn't have been more wrong, and by the turn of the century, people had set out to use these waves to communicate information at a distance, initially, more signals. Arguably the first one to achieve it was Russian Alexander Popov, but the two people that history remembers are the eccentric genius Nikol Atesna i l'entrepreneurial Guillermo Marconi. Marconi era no només un ingenier d'expressió, però un gran espanyol, i va capitar els seus inventions per construir l'Empire, la corporació Marconi, que, per uns anys, monoponeixia el llibre a llibre i el llibre a llibre de la comunicació. Era l'únic company que va fer això. Aquesta companyia i la tecnologia van fer fama quan el titànic va sonar en 1912. El titànic, el llibre de la stress, va ser reaccionat per un altre llibre, la Carpaccia, que va fer rescat a 700 persones. Aquestes són les recorrescències de l'Empire Marconi, molt bèstia, bèstia feina per l'Empire Marconi. A mi em diu que, com que Masbel i Hertz, aquests dos menys Estesna i Marconi eren reals giants en el temps. Fins i tot, més signals han fet l'amplit de modulació, i la mena humana va ser transmetida. A l'any 1930, teníem la modulació de freqüència. Aquí hi ha una de les més angoçades de les pròpies de Carpaccia, un gran bonic, però una real wonder, després d'aquell moment. A l'any II, 25 ciutadans de l'estat espanyol ha desplorat el sistema de tenafònic públic. Es consisteix en una sola transmissió que es desborda d'una torre a la altra possible muntanya. I només un handful de persones que podien ser desbordats a l'estat espanyol. El tipus de manera d'anar a l'anar va ser d'anar d'anar d'anar d'anar d'anar d'anar d'anar. En aquest context, en 1946, un enginyer de Douglas Rink, treballant a Bell Ops, va tenir una idea transformació, el que va desenvolupar en un paper que es deia Cels Sites. El que va dir era que es deia d'una region d'interès en uns països petits, que es deia Cels, each featuring a separate transmitter receiver. A més d'una transmissió que es deia l'entera region, es deia que es deia smaller ones, each serving one of these cells. So, in time this would be called a cellular system and Rink illustrated his idea with this picture, which it's kind of hard to see, but it depicts the cost of New Jersey and the city of New York, this is Manhattan over here. And these are the cells he proposed using circles. This divide and conquer idea was visionary, as we will see, but the technology to implement it simply did not exist in 1946. Okay, well, there are places and times that are special and bound ups in the middle of the 20th century was such a place and such a time. So, only two years after Rink planted the seed for cellular systems and not far from his office, in fact, information theory was born. So that made it possible to quantify information with the bit as the universal currency and to establish the maximum amount of information that could be transmitted reliably through a noisy channel. So, in a very real sense, the information age began in 1948 with Shannon's work. The fifth and final breakthrough, in my opinion, came with the realization of a comment made by Richard Feynman, that is plenty of room at the bottom. The offspring of this comment was integrated circuit with pioneering companies such as Fairchild Semiconductor Intel or TI that made it a reality in the 1960s. Gordon Moore, in particular, one of the founders of Intel, is credited with the observation that the transistor density is double of roughly every 18 months. This Moore's law has driven the evolution of integration of processors and memory to our days. So, these five breakthroughs set the stage for the first generation of cellular systems which were trialed by AT&T in 1978, in Chicago and Newark. Here you can see the paperwork from those trials, each consisted of a bunch of cells which are here shown as hexagons. The commercial deployment came in 1983 and this gentleman here is Martin Cooper, the guy who designed and made the first call from a handheld device. So following Green's vision, these first generation systems were organized in cells, it's featuring a central site called the base station, which is a transmitter receiver that connects by radio with the mobile devices and then all these base stations are cabled into a network. So each mobile device, for instance, this guy here connects by radio with the base station in itself and then the signals go by cable to the network. And then as people move around, their calls get handed off from one cell to the next. This first generation was followed by subsequent generations 2G, 3G, now 4G, I guess those of us who are old enough have seen all these generations, probably one every 10 years, just riding more slow to increase bit-rays and capacity dramatically, although I have to say without any major conceptual breakthroughs. Do they set the progression of the added capacity over time? It's useful to express it as a product of three terms, so that would be the spectrum efficiency, that would be bits per second per hertz per cell, times the density of cells, so say cells per square kilometer, times the bandwidth in hertz. So that gives you bits per second per square kilometer. This guy we encountered a minute ago, Martin Cooper, the one who designed the first hand-hell phone, observed that the added capacity, this quantity, has roughly doubled every 30 months or so, which corresponds into a factor of about one million over time. So he actually, when I had, I broke down this factor of one million, and it turns out that the lion's share factor of 1600 has come from increasing the cell density, with smaller factors that have come from having more bandwidth and more efficiency. These are just ballpark numbers, but they do make the point that when it comes to increasing the added capacity, nothing beats densifying the cell structure. And indeed, Rink, surely, would be amazed if he could see how far his idea has gone. So today, in the US alone, there are close to 400,000 cells. As you can see, they correlate with population density. And worldwide, there are about 6 million cells today, roughly one for every 1,000 people. So this is certainly a successful idea that was ever one. So that's what we are today, with 4G deployed in most of the advanced world, in developing countries, they still have 2G and 3G. And just a couple of data points, there are 7,300 million subscribers worldwide, which is about 90% of the population. And just to put that in context, there's at least 2,000 million PCs in the world, personal computers, okay? So the European Union exceeds 100% penetration, the US 110%, so there's more phones than people. And there are many nations, including big ones like India, that have more people with wireless access than with an electricity at home, which makes you wonder how they charge their phones. But never mind. And final data point, a smartphone today has more computing power than the entire aponga program that sends people to the moon. The expansion has been phenomenal, for instance, and very fast, for instance, in terms of internet access, wireless access has taken over from fixed access in just a few years. So in 10 years, we've gone from all the access was fixed to mostly 80% is mobile. So the future of the internet seems to be mobile. And using the air capacity breakdown I introduced with a small twist, we can gauge the performance that today's networks can deliver. So let's see. The efficiency on a 4G network is about 1B per second per hertz per cell. Every cell serves about 1,000 people. And the total amount of lights and spectrum, including 2G, 3G, 4G, and all operators is about 500 mHz, give or take. So you multiply these and you get about 500 kilobits per second for each of us. Now, most of the time, only a subset of us are active. So we actually get to enjoy a few megabits per second, which is a huge improvement over the 10 kilobits per second that was a peak rate 20 years ago. So we multiply that by about 1,000 in 20 years. In 4 generations. I couldn't possibly detail all the advances that have taken us from 1G to 4G. Let me just briefly touch on one that has been my main research focus since I finished my PhD. It's called MIMO. It stands for multiple input, multiple output. And it consists of vectorizing the transmissions. So rather than transmitting a single signal from a single antenna, we transmit a vector of signals from various antennas, say vector X. The vector of signals. And then, if at the receiver, we also have various antennas, what we observe can be construed as another vector, and it's called Y. And these vectors are related to a matrix called H here. So the ijth entry of this matrix H will be the specific channel between the i-th antenna here and the j-th antenna here. So by inverting this matrix, we can recover at the receiver or we transmit it at the other side. Of course, it's not so easy because the observations are noisy. This vector N represents the noise here. So you can just quite invert it. But with substantial signal processing, you can recover X from observing Y. That's what we do. This is MIMO in a nutshell, and it's already commercial. In fact, the smartphones you have in your pockets all have at least two internal antennas. And they are all receiving vector transmissions probably right now. So you know what we are with megabits per second for each of us on a good day, plus Wi-Fi on the hotspots. Do we need yet another generation? It's a good question, and apparently we do. Because the hunger for bitrate is relentless. Right now, it's growing at 57% a year, which means a factor of 10 every five years, a factor of 100 every 10 years, a factor of 1,000 every 15 years. It's mind-boggling. Right now, 2017, we're at about 11 exabytes a month worldwide, going to 16 next year. Remember, an exabyte is a million terabytes. So it's just, like I said, mind-boggling. Amounts of data that are being sent. So we do need another G apparently. And the research on 5G actually got momentum already four or five, even six years ago. And we had a period of much excitement, what I usually call 5G mania in my community. Which is now coming to an end as the target, the preliminary date of 2020 approaches. The idea is to have the first trials at the 2020 Olympics in Japan. So just to see, just so you see the scope of this 5G mania, these days you can get all sorts of 5G gear, you know, t-shirts, coffee mugs, hats, whatever you want. Not many research topics, I think, have this luxury. This frenzy has also been feeding all sorts of initiatives, especially funding instruments that have popped up like mushrooms all over the world. In Europe we have something called the 5G public-private partnership, 5GPP, which has been funneling millions of euros into 5G research for some time now. So in 2014 I wrote a paper on 5G with colleagues from Samsung, Huawei, Nokia, and other universities, which has had substantial impact. The paper outlined the direction that 5G has taken. Which drives to the limit the three existing mechanisms that are indicated here. So MIMO becomes Massi MIMO, densification becomes ultra-densification, and the frontiers of usable spectrum get pushed aggressively. So let me briefly comment on these three research directions. So with Massi MIMO the idea is to go from today's tower-top deployments or base stations, on what we can house, say, dozen antennas or so and most, so we can vectorize but just in limited amounts, to having vast arrays deployed, let me see, here it is, vast arrays of antennas deployed along the roofs and even the facades of buildings, or maybe on billboards, each one with tens or possibly even hundreds of antennas, and camouflaged so they don't disturb the environment. This is the vision for Massi MIMO. The prototype is already out there. Here on top you can see a 64 antenna prototype built by Agatán Lucer, which is now Nokia. And below there's a 128 antenna design that came out of a European project. So Massi MIMO is one direction to equip big cells with these huge arrays. And then ultra-densification is the second direction which would complement these big cells with lots and lots of very small cells. And terms such as picocel and now you are femtocel have been coined to refer to these tiny cells, which in a sense will be competing with Wi-Fi. The third direction in 5G has been to push different years of the usable spectrum. So remember that historically we haven't been able to use frequencies above 6 GHz because they don't propagate very well. Now with ultra-densification, well, we don't need the signals to travel very far because the transmitters and the receivers are going to be close by. So we actually can survive with these attenuations and there is reasonable hope that we may be able to tap frequencies up to 70, 80 or maybe even 90 GHz. The jury is still out on how effectively we can use these frequencies, but the expectations right now are very high. So we'll be broadening the amount of spectrum that we can use. Ok, that's in a nutshell what 5G is shaping up to be, macrocells with massimimol, plus lots and lots of more cells using a wider range of spectrum. Now looking beyond 5G, let me turn to my current research. Despite the great progress in performance, the structure of today's networks and even the structure of the upcoming 5G networks is really no very different from those we had 20 years ago. So the cell is still at the heart of everything. And there's an inherent problem with cells, which is interference. So when a base station sends a signal to a user, that signal does not only reach this user, as the nature of radio propagation, it reaches everybody. And for all the unintended users, this signal is interference that adds up to the noise. So keeping interference under control has been a favorite research topic for years, the problem that has kept lots and lots of people busy for a very long time. But when you think about it, what is really interference? Interference is actually a signal that's meant for someone in a cell, which is being received by someone else in another cell. So interference is determined by the cell structure, which is an artificial construct. So, perhaps the time has come when we should move past the concept of a cell, which has, of course, served us very well, but which might have run its course by now. So we should perhaps start viewing the network as what it is, which is just a set of base stations serving a set of users. And rather than have each base station serve only the users in its cell, we can have all the base stations jointly serving all the users. That can be done by applying the MIMO idea of vectorization across multiple base stations rather than individually at each one. And this is one of the premises of the ERC advanced grant that I'm running, the idea of having wireless networks without cells. Consider this two-cell toy example, where with a similar structure that is cross-interference between the two cells. So, each base station does not care for the signal being sent by the user on the other cell. That's just noise for this base station. Now, if we remove the cells, and we know that two base stations will work together, and join the code both signals, then both base stations care about both signals, and the interference now has become a useful signal. So, not only do these crossings stop being interference, now they have become usable. Now, this looks easy when you have only two base stations, but when you consider a network with hundreds or thousands of base stations, it's impossible to jointly vectorize all the transmissions. So, one must carefully define local neighborhoods where each base station vectorizes its transmissions with a selected bunch of neighbors. Moreover, they should all be dynamic. So, we should have a particular vectorization structure at a given point, with base stations serving users in a certain way, and then, as people move around and the users come into the system, we should have a different arrangement, different vectorizations, and so on and so forth, okay? Without the rigidity of a predefined pattern of cells. This is certainly not easy, but now we have the computational power to phase it, and we have lots and lots of data being constantly gathered and reported by your smartphones, which can help the network assess the best course of action at each instant. That's something that I might assume, and I'm exploring under the event of the Maria M.S.2 program, in cooperation with Anders Johnson. So, we're playing with machine learning, which is a new tool for us, and we can credit this work and this cooperation with Anders to Maria M.S.2. This idea of getting rid of cells goes hand in hand with other currents of thought, chiefly the idea of deconstructing base stations, so just to take all the processors that reside physically at each base station right now and put them together into a cloud, okay? So, if we could do that, we would get something called a cloud radio access network, which is now a very fashionable concept. So, taking this idea to the limit, we could have the signal processing for the entire network on the cloud. And then the network structures, and then being pre-determined cells, would now become software-defined. The vectorizations could be software-defined. So, instead of having these hardware boxes that are today's base stations, we'd have software running on a data center somewhere. This would offer incredible elasticity because no resources would be unused or wasted. And especially because the upgrades would come in the form of software updates, rather than painful hardware replacements. So, going from 5G to 6G, when the time comes, would be not going from Windows 7 to Windows 8, as opposed to what happens today, which is a physical replacement of each of hundreds of thousands of base stations, okay? So, which is extremely expensive. So, it's a very attractive idea, this cloud radio access, but it has some caveats. One is that the data centers cannot be in remote places, like, say, Amazon has them. They have to be nearby for the device to be short. So, we need one of these data centers every, I'd say, at most 50 kilometers or so. And you'll see in a minute why the device is so important. But just to cover Catanguania, we'll take at least a dozen or so of these small clouds, so, Mocan clouds, we'll call them. Okay. So, this is the final section of the talk, where I'm going to outline things that I see in the horizon. So, it's a little more ill-defined. If we build the evolution of wireless communications from the viewpoint of the human senses, we've gone from transmitting just audio in the early days to now, nowadays, transmitting audio and video. And now, we started to think about transmitting also smell and touch. If you're curious about transmitting, I refer you to the website of this company, your notes, which is proposing devices to do just that. But here, let me focus on touch, which has more implications and it's now becoming possible thanks to haptic technology, which uses motion and pressure to recreate touch. This requires very short delays on the order of one millisecond or so, which is the time that it takes the brain to process físic en contacte. The end-to-end delay in one of the networks has come down substantially over time. It used to be as much as 100 milliseconds in 3G, even more. Today, in 4G, it's 20, sometimes 15 milliseconds. That's the round-trip delay. And in 5G, the target is about one millisecond, so that we can recreate touch at a distance. This is the target of the system, this is the reason that the data centers cannot be very far. These 50 kilometers I brought up have to do with the speed of light, and that's pretty much it. One millisecond speed of light, 50 kilometers at most. Now, if we succeed with this one millisecond delay, which is far from easy, we'll open the door to so-called tactile wireless networks with lots of exciting applications, especially in remote medicine. So, this could really be a game changer in healthcare. So, you can, for instance, imagine someone who has had an accident being taken to the hospital in an ambulance and having surgery in the ambulance, right, from a surgeon who is at the hospital. This could be possible, for instance. There's a lot of trials now on medical equipment being controlled remotely, and with one millisecond delays, you can actually recreate the touch at the other end. So, you can actually operate at a distance. So, that's very exciting. Another interesting issue in the horizon concerns the position of the smartphone as a dominant platform for wireless access. Nowadays, smartphones have become almost an extension of people's bodies, especially for youngsters. In extreme cases, literally, an extension of their bodies. But this privileged position of the smartphone may be coming to an end. New platforms are about to proliferate, and we may soon be transitioning to an ecosystem of communication devices that, rather than portable, are wearable. I'm talking about smartwatches, smartbands, helmets, eyeglasses, even shoes and clothes, say, jackets, such as this one, with lots of embedded antennas. My student, Jordi George, recently finished his thesis on the topic of wearable wireless devices, and we just cried the surface. There's a lot of things to be done in this space. We could even see contact nights as equipped with wireless capabilities, although this is probably further into the future. Keyboards and screens may become a thing of the past, progressively replaced by projections. So, on-skin projections and off-wall projections. So, this evolution is going to present major challenges for us, because these new platforms are very small devices with tiny batteries. So, we're going to have new constraints in terms of power consumption and complexity. The extreme of these will be in-body devices, say, monitoring sensors, direct delivery systems, medical robots on implants. The spectrum is already reserved worldwide for these devices, però moltes coses han de ser pensades abans que puguin ser viables d'un punt de comunicació. A l'altre extrem, tenim una altra màxima platforma, un car connectat, fins i tot el car self-driving, en què tenim grans batteries i molts processing-power. Les conseqüències, aquí, són drastically diferents. A veure una gran foto, anirem de moltes devices homogíneos, els smartphones, a diverses devices heterogíneos, so, we certainly have research challenges for years to come here, in terms of communications. Ok, so, my 40 minutes are up, so, let me finish by acknowledging the support of various institutions, the European Research Council, NICREA, in the public arena, and then also Intel Corporation, through the 5G University Research Program. I also wanna thank my students and POSDOX, both the current ones, I mentioned a couple of them, in relation to projects, i aquests que han vingut a altres llocs. I l'enclosc i deixo una cota de Marconi. Vaig fer aquest comentari en 1932, potser abans que va morir. I crec que, si pot veure com s'ha quedat, avui dia, serà prou. Gràcies. És cert. Vam just arreglar el tracte. Això és allò? Això és l'evolució. ¿Aquest comentari? I, si hi ha hagut problemes relacionats amb la comunicació que s'ha d'accedir al futur? Sí, és una bona qüestió. A vegades, em quedi les meves estudis, perquè no tenen sentit de por. No entenem la diferència entre megawatts i megawatts. Una cosa que ha passat a l'hora és que, quan les cels han tornat més i més, les cels han tornat més i més. Així, tots els phones han de fer un poble de por. Aquestes dies, els smartphones transmeten més de 100 megawatts, més o menys, perquè tenen control de la base. Aquestes són les seves amuntes de por. En principi, no hi ha problemes. I la tendència continuarà de tenir més i més por, perquè de les problemes de bateria, etcètera. Així, si posem això en context, l'aigua de microwave és blastant 800 morts, quan et cuinis alguna cosa. Així, el llibre a la porta de l'aigua de microwave és més que un poble de megawatts. Així, no hi ha problemes amb portables de devices. Aleshores, no hi ha una xifra del meu cervell l'endemà. Ha de fer molt amb temps, com amb por. Així, és l'aigua de por que m'ha matat. Així, l'exposició sporàtica a la por és també no molt malament. L'exposició sustain, l'exposició acumulatòria és el que passa. Ara, hi ha una altra qüestió, que és la base de stations. I el que passa aquí és que no volem tenir gran cobertura, però no volem la base de stations per a casa, que és un oximor, en el sentit, o una contradicció. Perquè per tenir un bon servei, necessitem una base de stations per a tu. Ok, això és el que fem. Una de les grans obstacles per a les netes densificades és zones de permets. És molt difícil fer permets per instal·lar base de stations, perquè no volen tenir-los. Ells volen els serveis i no volen l'equipament. Així, això és el més gran obstacle per a la densificació, és fer permets per instal·lar nova base de stations. És entendible. La base de stations ha de tenir més poder, unes tècniques de quants, a més. És still not a lot, però és entendible que no vols ser almenys a la nit. Però el poder és molt fàcil per a distància. Així que, si ets una altra vegada, és per cert que no és fàcil. Així, no hi ha gaire preocupació. Gràcies. Moltes gràcies, Toc. Vull saber, en additione, que els issues de la ciutat, els issues d'etica relacionats a aquest tipus de recerca, de la mobilitat, on estem a cada moment, què fem a cada moment. És una bona qüestió. Aquesta és una cosa que no hem de preocupar en el passat, però ara ho hem de fer, perquè, quan començarem a utilitzar data amb les devises, per fer decisions, potser ha de ser una qüestió. I ha de fer una anonimia, crec que. Però això és una cosa molt nova per a nosaltres, certament. Jo acho que parlem de ser que es fa una qüestió, essayer de fer una qüestió com una qüestió a la gent. Veig que both t'miyorco. I per Si et veuen en una qüestió de ràpid, s'ha de fer és un missatge molt important. Per ordre, però és a la qüestió de la teva ciutat, per donar una qüestió de sobre per a vosaltres. Per doneus, quan es fan una qüestió de les agraïments, pot ser una qüestió de la ciutat, que no es abaixin per a una qüestió, que et donin un integrating a lescies. I la qüestió de la ciutat, No, no. No, no, I don't have to get your permission to do it, yes. I can have a hard time thinking about evil things that can be done with the data. But then again, once you begin, things can happen that one hasn't thought about. I can hardly imagine anything bad that Tenguafonega could do knowing where we are. But maybe I might. A mi em sembla que la 5G és molt fòstima amb l'acord amb la comunicació. A la mateixa hora hi ha una gran trend d'IoT que té l'aplicació de les nens d'enıy, com l'Aura, el Forx, etc. Llavors hi ha una convergència entre les dues. A mi em sembla que, fins i tot per les nens d'enıy, amb les devises com les xarxes, poder tenir l'aplicació de l'aigua, les notificacions de pressa... com si tinguessis una aplicació tèpica en IoT. Això serà molt interessant. Veus la convergència, o serà més o menys diferents tecnologies per a diferents aplicacions? Sí, és una bona qüestió. En el passat hem tingut diferents networks per a aquestes coses. I crec que és el way to go. Però, fortuna, tinc la majoria aquí. La decisió ha estat per a integrar-hi tot. 5G també suporta IoT. Aquesta és la biquetica, perquè aquests són molts diferents types de sistemes i ara han de coexistir. En IoT el problema no és l'amunt de data, sinó l'amunt de devices. Els devices senden molt de nit de data. S'hi senden reportes, mesuraments, temperatures, alarmes i altres. És molt poc de bits. Però molts, molts, molts devices. S'haurien de complir amb les xarxes de control, per exemple. És un diferent problema, però ens han de posar en contacte amb el mateix network. Hi ha diversos interessos de comerciació, de llibertat, d'enginyament. D'un enginyament de temps, ho seguiré separant, però han decidit integrar-los. S'haurien de ser unes. 5G també suporta IoT i smartphones i altres altres gadgets. És una xarxa. Ok, anem-hi a ser positius. S'haurien d'ajudar que hi hagi un job per molt de temps. Papers de llibertat i coses a fer. Anem-hi a posar-li en aquesta xarxa positiva. Ok. IoT, IoT gaire, Rafa. D'acord amb tu, què és l'estranya més tècnica per arribar a aquestes altres eficiencies? L'estranya més gran, és que els buscats de la manera més clara queailleurs ho han fet i el que és més clara, és que, per ara, només un setembre de la pròxima part que es queixi al mig d'aquí una xarxa i el que és més important és que es faci un costat. La capacitat per fer les companyies com més clars que el deixar que no pot serviria la mateixa Contacte. Un dia, i és un costat, un costat de molt més. un clau que és el que el processament ha fet. L'antena és un dom. Ells només tenen un converter, un amplificador, i ells tenen la informació raó, informació al clau, i llavors cranquen-ho. S'haurien de no aconseguir processar-ho com avui. El problema ara és l'amunt de data que va a aquestes cables que van al clau, que és molt gran. A vegades 100 MHz a vegades, una mica d'antena, potser 4 per d'antena, per a un número de bites per a digititzar els senyors, es pot fer un gigabit per segon per a l'amunt. Per tant, hi ha fibra-optics, que es pot fer en llocs on hi ha moltes fibra-optics. Barcelona pot suportar això. Hi ha fibra-optics a més d'aparements d'aquests dies, però hi ha d'anar a l'Africa i no hi ha res. A Barcelona hi ha diversions, entre Corea, Xina, Japó i totes aquestes places, que hi ha fibra a totes les places. I també partis d'Europa, per a les restes del món, on no hi ha l'infrastructure que suporta. I aquestes diversions és extenció. Però això és un lloc molt tècnica, i és un lloc d'informació i de teoria. Compressos, segons, hi ha moltes redundàncies en què hi ha d'observacions de diferents antenes, i les persones que treballen en això. Sobretot puc posar això al top de l'estat. Oh, el meu own estudiador, és molt d'entrada. Moltes gràcies per una bona taula. Aleshores, vaig vendre en les últimes 20 anys, ser en l'engineering de virus, i treballar en la propagació de model, i how to improve the... We trade all by analysing the signal, we receiving all the environment. So when we come to the idea of this cloud, so we have lots of data, we have so also you know this machine learning coming to almost all the area. Do you think is there any possibility, so we lose our job in the future for this machine learning or all the technique, because our job was to analyse these things, create algorithm, improve efficiency. So if machine learning can do better than us, why they need us. Yeah, no, I'm not worried about losing my job to machine learning yet. But this is an interesting debate. It's almost philosophical, right? It's the expert skills versus the... I mean, okay. He's the right guy to answer that question, right? Machine learning people. But... Or many other people here, better than me. But I would say that machine learning can do certain things very, very well, but nothing else. Like, you know, find patterns in data. Identify faces of people. It's amazing, actually, yeah? But, you know, a robot cannot fall the towel, right, Hector? And probably won't for a while, thank goodness. So there's still, I think, a lot of room for experts like us doing work that cannot be said, complemented, right? So I think this is a tool we have to use and you're trying to do that in your thesis to do certain things. And there is going to be better powerful. But I think that there's still a lot of things that we can do better in other ways. With other tools, like analyses or computer simulations. And that's it. So I didn't fully understand the world without cells idea. So to me what defines a cell is that this finite range of transmission. And what you were mentioning was a more like a moving cells with more efficient capabilities. But what do you mean by a world without cells? Yeah, so in a cell what happens is that as a user, your smartphone connects to the closest base station, right? That's actually what it does. It connects to the strongest base station, which is the closest one. And the rest of the signals that are received from other base stations are interference to you. So the point is to change all these and have all the base stations send signals that are meant for all the users at the same time, right? So you can vectorize. The problem is that it's a little hard to explain. You can vectorize the signals as they do any DSL, for instance, so that you receive only your share, right? And the rest of the interference avoids you, right? And I don't have a good picture for that, right? Except that maybe example, right? But when you vectorize well, right? There are 10 base stations sending signals and you observe a single signal and no interference, right? You do it right, right? It's just mathematics, right? It's essentially matrix algebra. You can vectorize properly so that you receive a clean signal with no interference. And vice versa, when you send your signal and everybody does at the same time, the base station joining can decode the signals and get all of them without interference. That's the idea. You could do that with cells, yes. You can imagine there are cells that cooperate somehow, right? In fact, people have thought about this already, right? Imagine that the cells are cooperating cells, right? That's one way to see it. Another way to see it is forget the cells, right? You have base stations because the cells really are artificial. They don't exist. There's no line on the ground saying, yep, end of the cell. No, it's based on powers. Hope that helps. Okay, so related to previous questions about machine learning and so on, but it's not machine learning. So you're trying in some sense to maximize bit rate or efficiency or the defiance or some function, okay? So the very sort of model-based approach is, let's try to set up the design space and let's try to optimize, okay, this, okay? And then look for some optimization for the space of designs. Well, here it looks that, so sometimes you bypass this by these regular cells or then moving something else. So this explicit optimization approach, have it been pursued or is it? We do, we do. A problem is that you could get into non-complex problems. In fact, even the two, I showed a baby example with two base stations and two users, that's a non-complex problem already. It's a correct one. No, no, the optimization, you can formulate the optimizations, they're very complicated, right? But, yeah, we optimize all the time. Sure, the problem is that we, so if I use machine learning, I don't know what I'm getting, I'm doing, right? It's garbage in garbage out. Yeah, yeah, yeah, sure. We're trying to do that now, we, it was all. I think some problems can be helped that way, yes. But not all. The last one, right? Yeah, yeah. So when you said that there was this decision of IOT to move it together with 5G, I'm curious about the decision process. So where is this decided? Yeah, there's a body that standardizes all the systems, right? So all these, you know, 2G, 3G, 4G, there's a forum called 3GPP, which actually should be 5GPP now, but it's still called 3GPP, which holds about 200 companies, including all the ones that you can imagine more. And they get together, they have meetings once a month and they standardize it. They produce all the standards, physically the standards, right? So that's why they debate patterns and technologies, directions, and all that stuff, right? And that's what it was decided. And it makes for a very complicated system. Okay, I think, is it short? Well, that you're not hindered by what you had before, right? So you have sort of, the idea is to have a clean slate and start from scratch. Some attempts to have sales cooperating have not worked very well because you're still bound by the sales structure to some extent. So maybe just, I mean, you could do that, right? And eventually you get to the same answer, right? You should get to the same answer, if you do it properly. But maybe just best to start from scratch and say, okay, now, in the old days, sales make sense because they simplify the program very much. But now we have so much computational power. Let's forget about it, let's see what we can do. We have transmitters and receivers and nothing else. And, you know, so much capacity to process signals and see what we can do. But like I said, if you do it right, you should get to the same answer both ways, yes. Okay, I have one last provocative question for Angelo. So you know that there are some wifi guys at the department, right? Really? So we don't use sales. 3G tried to beat wifi, but it couldn't. 4G, the same. It's going 5G to beat. Hopefully, it's about time. This has been very frustrating, actually, because wifi is a very poor design. But it's working. It's working, yes. Most of the people right now, it's using wifi. There is this saying that says, better is the enemy of good, right? So it's a very poor design, very simple. And it works very well over short distances. So every attempt to overrun it in hotspots has failed, right? And 5G will be another attempt, which may have failed again. What should happen, I think, is an integration, right? So 5G should stop worrying about hotspots, cover everything else, cover IoT well, cover short delays, cover the outdoors, and then integrate with wifi. So sort of declare defeat for hotspots and make a pact with wifi and join forces. That's what I propose. It sounds good. Okay, thank you. Okay, thank everybody.