WEBVTT 00:00:33.000 --> 00:00:38.000 Good day, everyone 00:00:38.000 --> 00:00:37.000 Hello! 00:00:37.000 --> 00:00:37.000 Okay. Good day. Paris. 00:00:37.000 --> 00:00:41.000 Thank you. 00:00:41.000 --> 00:00:44.000 We'll give everyone a minute to join 00:00:44.000 --> 00:00:51.000 Sounds good 00:00:51.000 --> 00:00:56.000 There is. Would you like to do the introductions, or shall I 00:00:56.000 --> 00:01:13.000 If you would please. That'd be great. 00:01:13.000 --> 00:01:18.000 Okay. It's submitted after the hour. So let me get started. 00:01:18.000 --> 00:01:24.000 I'm at anvil from the National Science Foundation, and I'm delighted to welcome our size. 00:01:24.000 --> 00:01:30.000 Distinguished lecture here today at word nightly from Rice University, we in the next program here at Nsf. 00:01:30.000 --> 00:01:38.000 Are along familiar with Edwards work, and we're especially happy to walk on him. 00:01:38.000 --> 00:01:55.000 He and his group have been doing foundational work and wireless communication, sensing and security, and we're really looking forward to hearing how that work is being accepted and leveraged in the technology for all Project Edward is the Shaeffer Lindsey professor of electrical and computer engineering and 00:01:55.000 --> 00:02:06.000 computer science, rice is an acm, though, and I triple epsilon fella and recipient of the among many accomplishments this distinguished career. 00:02:06.000 --> 00:02:10.000 So without further ado, I will turn it over to Edward. 00:02:10.000 --> 00:02:15.000 I would like to just remind everyone that the presentation will last about an hour. 00:02:15.000 --> 00:02:20.000 There will be 30 min for Q. And a after. There is a assumption at the bottom of your screen. 00:02:20.000 --> 00:02:25.000 So please post your questions there, and not in the chat. 00:02:25.000 --> 00:02:32.000 And we'll get to those at the end. So again. Welcome, Edward, and go ahead. Thanks. 00:02:32.000 --> 00:02:48.000 Thank you so much. And for that introduction, and thanks to you, and in the entire size community for the invitation, it's a true honor to be giving this lecture today, and I'm delighted not only to give the presentation, but also for the office hours. 00:02:48.000 --> 00:02:52.000 Later, I'm looking forward to those discussions with with the Nsf. 00:02:52.000 --> 00:02:55.000 Team in recognition of the occasion. 00:02:55.000 --> 00:03:09.000 I wanted to begin with, some of the projects that sizes has funded in the past, because not only have they impacted my own research, but the community, the wireless community networking community at large. 00:03:09.000 --> 00:03:15.000 So I'm gonna start going back to a project that started 20 years ago now. 00:03:15.000 --> 00:03:14.000 And it's started with an Nsf. Itr Grant, that's a information technology research. 00:03:14.000 --> 00:03:33.000 And and that program was targeting interdisciplinary larger scale teams to to develop bold new designs for the next Internet wireless and wireless. 00:03:33.000 --> 00:03:50.000 And so we had a grant on what we call transit access points, and the idea was, it's an access point that that that not only serves the clients around it, but also that can control transit traffic in a mesh, and when this grant was awarded, it got announced in 00:03:50.000 --> 00:04:09.000 a front page story of the Houston Chronicle, and I got a call the next day from Will Read, Who's the founder of an organization called Technology for all, and he told me about his organization and their mission was to or is still to empower under 00:04:09.000 --> 00:04:20.000 resourced communities through technology. And he told me about how they had a refurbished PC program as picture in the lower left, where they were taking corporate Pcs. 00:04:20.000 --> 00:04:28.000 And putting new operating systems on them, and getting them ready for the community, and then also how they had tech training. 00:04:28.000 --> 00:04:40.000 So it was the PC. Plus the training of all the technology tools and what he told me is that the Missing link for the community was broadband access, and that was out of reach financially for those in the community. 00:04:40.000 --> 00:04:39.000 And he asked me whether we would like to deploy our technology. 00:04:39.000 --> 00:04:46.000 That we're developing as part of this Nsa project here. 00:04:46.000 --> 00:05:02.000 Read about in in that community and serve the actual users so as you can imagine, the original goal was in lab deployments and some in lab experiments. 00:05:02.000 --> 00:05:04.000 And and at that point we didn't have the the goal of doing an urban scale deployment. 00:05:04.000 --> 00:05:17.000 But we all rose to the challenge that technology for all gave us, and and launched an an urban scale deployment. 00:05:17.000 --> 00:05:22.000 And so once we had that goal, we developed a custom software and hardware transit access points. 00:05:22.000 --> 00:05:48.000 And so showing here in the lower left. These were some kids in the neighborhood that we employed to assemble and test the the tabs, and showing here an access node and some residents that we call them antenna host, and so what this antenna here would be and and the node 00:05:48.000 --> 00:05:50.000 Shown here was an access tier that provided access not only to this, this home itself, but all the the neighbors around it. 00:05:50.000 --> 00:06:12.000 In a couple of blocks, and then this antenna, also coordinateates with other nodes that form a backhaul tier to interconnect them, and then we had some directional links that we called capacity injection links to to scale up the 00:06:12.000 --> 00:06:18.000 Capacity, and provide the highest data rates possible and so shown here. 00:06:18.000 --> 00:06:21.000 This is will read, and this is a younger version of me, and then this is the one of the rooftop Umback hall links. 00:06:21.000 --> 00:06:31.000 There on the top of the tall building, so we could get a good connection to to the fiber. 00:06:31.000 --> 00:06:37.000 And then you can see here that the entire network deployed every red pushpin is one of the nodes that we deployed in any, either a home or a building such as the Y.M.C.A. 00:06:37.000 --> 00:06:54.000 There, and so we became, and still are, the Internet service provider for a high density of that. 00:06:54.000 --> 00:07:10.000 A 1,000 users per square kilometer in in one of Houston's most under, served in under resourced neighborhoods, and it provided the first multi-tier urban mesh where we it was access back hall and also the capacity injection 00:07:10.000 --> 00:07:18.000 And really a decade of Phd. Topics from my students and our collaborators. 00:07:18.000 --> 00:07:24.000 Some about asymmetric topology, and how some nodes, if we didn't take to for action, could starve network architecture. 00:07:24.000 --> 00:07:25.000 How do you design this network? Where do the nodes go? 00:07:25.000 --> 00:07:37.000 Multi-hop congestion mechanisms. How do we make sure that we, the nodes closest to the gateway? Don't get all the bandwidth, and we throttle them to the right level to make sure no it's farther away. 00:07:37.000 --> 00:07:46.000 I still can get access network measurement studies, network modeling studies. 00:07:46.000 --> 00:07:57.000 So it's research impact. And so you you might think everyone's happy at this stage with with the the network and the deployment. 00:07:57.000 --> 00:08:16.000 And so now this is 2 years later, in 2,005, and there was a bill sponsored in the State of Texas that would have outlawed our network, and and every network that was not deployed by a quote license telecom operator so you can guess 00:08:16.000 --> 00:08:22.000 who's behind the bill it was one of the major cellular providers who, who, I won't mention today. 00:08:22.000 --> 00:08:31.000 But there, you know, at the time the model was every wireless byte should be charged to the cellular operators, and if you can get it through Wi-fi, then that's a big threat. 00:08:31.000 --> 00:08:38.000 So this was a sponsored bill. It came very close to passing. 00:08:38.000 --> 00:08:47.000 Will re was the hero of pretty much camping out in Austin to try to talk to to all the legislators here. 00:08:47.000 --> 00:08:57.000 But so, you know, looking back, it's it's it is a disruptive industry shift that as Wi-fi as a public Internet entry point that that beyond cellular network that Wi-fi can can serve as an access point. 00:08:57.000 --> 00:09:05.000 And in public spaces, the mesh networking industry for multi hopping. 00:09:05.000 --> 00:09:10.000 It's 5 billion dollar industry today. And there's there's multi hopping probably in your home. 00:09:10.000 --> 00:09:16.000 I have it in my home and and and of course, in many other areas, outdoors to indoors in the standards. 00:09:16.000 --> 00:09:16.000 It was originally 8 to 11 s. And now it's mandatory. 00:09:16.000 --> 00:09:36.000 Since wi-fi. 5. These mechanisms I mentioned about making sure you you can report congestion and throttle different nodes in the mesh to make sure that you get coverage in the more remote aspects that was one of the important findings and made it to the standard but lastly 00:09:36.000 --> 00:09:45.000 this societal side about providing broadband for underserved communities, and what we heard from those in the community, like the antenna host that I mentioned, we would ask them about. 00:09:45.000 --> 00:09:51.000 You know what? Not just the technology, but what they were using it for, and they would tell us that it was transformative. 00:09:51.000 --> 00:10:11.000 Most for education, for their kids. That was critical for them to have access job search for them and their families, and and and really the notion of digital citizen where you have effective access to online resources and are able to use that and that was something that was previously limited to those with higher incomes. 00:10:11.000 --> 00:10:25.000 And then providing that for all, if hence the wonderfully named organization technology for all. 00:10:25.000 --> 00:10:27.000 And then I'll fast forward. One more story from the past. 00:10:27.000 --> 00:10:34.000 And so this will fast forward. 10 years, and this is the Houston super Wi-fi trial. 00:10:34.000 --> 00:10:42.000 And this started from one of the residents, who was at very much at the cell edge, and she had, you know, that one bar of service, and it wasn't highly reliable. 00:10:42.000 --> 00:10:50.000 It was really at a coverage gap, and she would regularly complain to us that the service was not what she needed to do. 00:10:50.000 --> 00:11:08.000 The things I mentioned for education, for kids, and so on. And so we decided that we needed to use new spectrum that we couldn't just keep deploying higher and higher density nodes. 00:11:08.000 --> 00:11:06.000 But we needed diverse spectrum to solve this problem. 00:11:06.000 --> 00:11:15.000 And so we wanted to use UHF TV bands. And so we got 2 grants in this domain. 00:11:15.000 --> 00:11:24.000 As F. Let's see large. It's called Nancy at the time, and and MRI. 00:11:24.000 --> 00:11:30.000 And this led to the first residential super Wi-fi trial. 00:11:30.000 --> 00:11:33.000 So super wi-fi! That's the name the tech press gave it. 00:11:33.000 --> 00:11:43.000 We don't know it. Name our own projects, super, but they called it Super Wi-fi, and the idea was to use unused UHF TV channels. 00:11:43.000 --> 00:11:56.000 So this is very much like, or is the concept behind white spaces and and repurposing spectrum for use for other uses and and Wi-fi is one possible use of that repurpose spectrum. 00:11:56.000 --> 00:12:05.000 And so this was the first time that was. This was done to a home, and we did a tower to home, and the research was in spectrum sharing. 00:12:05.000 --> 00:12:09.000 How can we coexist with TV diverse spectrum access? 00:12:09.000 --> 00:12:16.000 How can we use different bands from 500 Megahertz up to 5 gigahertz? 00:12:16.000 --> 00:12:26.000 And also I'd like to point out that this there's some generational, a progress here that this board here in the super Wi-fi trial was in this 10 years later. 00:12:26.000 --> 00:12:31.000 So it's multi-generations later from the from the original taps at PGA. 00:12:31.000 --> 00:12:39.000 And then the students designed a custom for an end. So this is the student Ryan Garra, who designed this custom front end. 00:12:39.000 --> 00:12:47.000 These are 2 of my students who deployed the antenna they didn't tell me in advance that they rented a bucket truck. 00:12:47.000 --> 00:12:47.000 Probably thinking that I would ask them about insurance and permissions and so on. 00:12:47.000 --> 00:12:55.000 So they just rented it, and then sent me this picture afterwards. 00:12:55.000 --> 00:13:03.000 So they were very proactive, ambitious students to make this happen, and so success again. 00:13:03.000 --> 00:13:06.000 Now we're in. We're in not only the paper newspaper, but by now the tech press has. 00:13:06.000 --> 00:13:20.000 As is catching on to these stories. It turned out that our first user, this is the one who had complained previously, and she's very happy now, as you can tell from the big smile with her super. 00:13:20.000 --> 00:13:28.000 Wi-fi service, and it turns out she's grandma, so that the tagline that the press used was Houston. 00:13:28.000 --> 00:13:32.000 Grandmother is nation's first super. Wi-fi user. 00:13:32.000 --> 00:14:02.000 And so I again the the the press release is followed by some interesting posts. 00:14:07.000 --> 00:14:14.000 The phone, waiting for him to say congratulations. But actually, he said, do you have a license for that transmission? 00:14:14.000 --> 00:14:13.000 And so, you know, you can ask, okay, what? 00:14:13.000 --> 00:14:21.000 The answer was, Yes, first of all, but you know the question, what? 00:14:21.000 --> 00:14:28.000 Why it get these questions right after this, and and and this was a tussle at the time between for spectrum sharing the incumbents on the spectrum, didn't want to share that spectrum. 00:14:28.000 --> 00:14:39.000 They were concerned about, would we properly use the right channel they were concerned about? Would we properly use the right channel? Would we spill over and disturb? 00:14:39.000 --> 00:14:55.000 And so one of the things that was critical for us, but also the broader research community was was to demonstrate that first of all, there's a value to share an unlicensed spectrum that we can be serving underserved communities that we can have large population. 00:14:55.000 --> 00:14:58.000 Densities, and we can make great use of the spectrum. 00:14:58.000 --> 00:15:05.000 But also that we can do so without interfering with with the incumbents. 00:15:05.000 --> 00:15:11.000 And so today, it finally fast forwarding to today, the Fcc. 00:15:11.000 --> 00:15:19.000 First of all, continues to open up spectrum unlicensed, and shared access, that we're making excellent progress, especially spectrum in in, in in the right direction. 00:15:19.000 --> 00:15:27.000 The spectrum, sharing capabilities are now in Wi-fi access points. 00:15:27.000 --> 00:15:45.000 For example, the 60 Gigahertz bands for Wi-fi have to coordinate with radar incumbents and the mechanisms that are used, or are are nearly carbon copies of what we had back in the in the UHF bands and also in 5G 00:15:45.000 --> 00:15:58.000 Deployments and Cprs using diverse spectrum, and just to give a little story, this student right here is is now CEO of Skylark Wireless that he co-founded. 00:15:58.000 --> 00:16:12.000 And so they're using diverse spectrum. 100 Megahertz that's UHF TV bands all the way up to what we use of Wi-fi 6 gigahertz and and using Massimo to make sure again, maybe these lessons. 00:16:12.000 --> 00:16:23.000 From these, from those frustrated with lower data rates that we've got a cover of the cell edges and get the high data rates and high high bandwidth. 00:16:23.000 --> 00:16:36.000 So what I'd like to do now is transition moving forward and look at our current work and some goals for us and the community at large. 00:16:36.000 --> 00:16:36.000 And so what will the next generation, wireless networks be and look like? 00:16:36.000 --> 00:16:41.000 And so some of the new capabilities that we can't do today, in the wireless. 00:16:41.000 --> 00:17:07.000 And and that's just combinations of doing high data rates and terabyte per second, combined with low latency for new applications, but also to do it while maintaining mobility for the clients robust and to outages security which I won't talk about today but 00:17:07.000 --> 00:17:09.000 That's another key component of wireless. 00:17:09.000 --> 00:17:18.000 And so like one of the key drivers, is to move up and spectrum that everything I talked about before was in the 6 Gigahertz and below. 00:17:18.000 --> 00:17:35.000 But if we start to get to millimeter wave and up to sub terra hurtz, in other words, 30 gigahertz up to say, 300 gigahertz, and then now our bandwidths become gigahertz versus 00:17:35.000 --> 00:17:46.000 Megahertz so a Wi-fi channel is 20, Megahertz, and and if we wi-fi in 60 gigahertz, this is 2 gigahertz so orders of integer wider bandwidth which means order of 00:17:46.000 --> 00:18:07.000 Magnitude, higher data rates as we go to lower wavelength millimeter, and instead of a centimeter or tens of centimeters, then our resolution for sensing also becomes increased resolution due to the lower wavelength, and we can now have high directive 00:18:07.000 --> 00:18:23.000 Beams. I'll show you some beams that enable us to serve many users in space. 00:18:23.000 --> 00:18:26.000 Simultaneously, for example, and this can drive new applications. 00:18:26.000 --> 00:18:20.000 So just a couple examples, we can have drone networks. 00:18:20.000 --> 00:18:29.000 This is one of our projects at rice about drone networks autonomously communicating and sensing. 00:18:29.000 --> 00:18:29.000 And so these drone networks can potentially have much higher data rates amongst each other. 00:18:29.000 --> 00:18:45.000 Higher resolution sensing, and then other examples of you know, virtual and augmented reality, that today, of course, there's there's wires. 00:18:45.000 --> 00:18:53.000 And if we had, the high density that you're seeing there of users, there's no way we could not only provide the data rate to what to one. 00:18:53.000 --> 00:18:56.000 But the high density. 00:18:56.000 --> 00:19:00.000 So I wrote tens of gigabits per second per meter square spatial density. 00:19:00.000 --> 00:19:00.000 So those are some of the things we can't do yet in wireless. 00:19:00.000 --> 00:19:10.000 And so I want to talk about some of the steps towards towards getting there. 00:19:10.000 --> 00:19:16.000 So to, yeah, get to the path of moving forward. 00:19:16.000 --> 00:19:28.000 I like to give a starting point of where we are with the standard and with the Wi-fi standard, and talk a little bit briefly about how that standard works, so you can see some of the challenges moving forward. 00:19:28.000 --> 00:19:35.000 So the latest standard for Wi-fi in millimeter wave is 8 or 2 11 a. Y. 00:19:35.000 --> 00:19:47.000 And it's a 60 year, Hertz, and if you'd like to learn more about it, this is a tutorial overview paper by my former student, and also 2 of the authors of the Standard from Intel, as well as myself. 00:19:47.000 --> 00:19:55.000 And just to give a little sense of how Wi-fi works. 00:19:55.000 --> 00:19:57.000 So at 60 gear hurts, it starts with an antenna. 00:19:57.000 --> 00:20:12.000 Right? So we have some number of antennas, and they do bean steering, and so how they behe steer is put some different delays or phase offsets, and we'll call them weights for each of the antennas. 00:20:12.000 --> 00:20:15.000 So if all the weights are 0, then it just steers straight ahead. 00:20:15.000 --> 00:20:24.000 But if you wanna steer to the right or to the left, then you put a phase gradient, and you know positive or negative phase delays on each of those weights. 00:20:24.000 --> 00:20:35.000 And so each of those steering directions are quantized, and you call the sector, and then there are up to 128 sectors per device. 00:20:35.000 --> 00:20:41.000 So now you can squeeze the beam with down to down to several degrees, and the standard has some mechanisms to adapt the select and adopt the sectors in order to steer. 00:20:41.000 --> 00:20:47.000 So how do you steer the beam? And then the latest standard also it's my demo. 00:20:47.000 --> 00:20:54.000 I won't talk about that today, but you can do this. Multiple times with multiple are of chains for my bill. 00:20:54.000 --> 00:20:54.000 So how does the standard work just to give you a little sense of how to deal with directivity? 00:20:54.000 --> 00:21:04.000 Because next generation networks will be highly directive. And so it got to align the beings. 00:21:04.000 --> 00:21:06.000 And so this is the way that it's done in in 802 11 a. 00:21:06.000 --> 00:21:25.000 Why, it's called a sector level suite, and the idea is to transmit in each direction from we'll start with the access point here, and then the client is listening in as best as it can in all directions. 00:21:25.000 --> 00:21:28.000 So it's called quasi omnidirectional. 00:21:28.000 --> 00:21:32.000 So there's no truth. Such thing as I'm any directional. So it's doing its best to listen. 00:21:32.000 --> 00:21:33.000 And alright, and the sweep here transmits as many possible code books or beans as as the Ap. 00:21:33.000 --> 00:21:45.000 Has, and it does a short header in each of those to number the beams. 00:21:45.000 --> 00:21:56.000 So this is being number 17 or something. And then there's also a preamble, so that the the receiver can actually decode the number and know which beams the best, and then and then you repeat it in the other day. 00:21:56.000 --> 00:21:58.000 So now the client is going to do it. Sweep and the rest, and the Ap. 00:21:58.000 --> 00:22:15.000 Is now trying to listen in all directions, and that yields the best sectors and the best beings in on both sides. 00:22:15.000 --> 00:22:21.000 So we've it's a successful method for aligning the beams. 00:22:21.000 --> 00:22:32.000 And so it works in Wi-fi and it's the best beam pair until there's, for example, a mobility induced beam alignment. 00:22:32.000 --> 00:22:40.000 So we have here an Ap. Transmitting to a couple clients, and then, if the client rotates, then the client needs to re steer. 00:22:40.000 --> 00:22:47.000 If the client translates, then both the Ap. And the and the client need to re steer. 00:22:47.000 --> 00:22:48.000 So we're aligned, but we can lose the alignment with mobility. 00:22:48.000 --> 00:23:06.000 We can also lose the alignment with blockage. So if we have someone come in between the beans, then that is also blocked, and we would need to re steer in this case I'll show a wall that 16 year Hertz and higher frequencies can reflect give it a specular 00:23:06.000 --> 00:23:10.000 reflection off of the wall, and so it a flat surface. 00:23:10.000 --> 00:23:15.000 And so if there's a wall nearby, then we can restart the bee. 00:23:15.000 --> 00:23:19.000 Okay. Now, what's the problem in all this? Is this all solved? 00:23:19.000 --> 00:23:24.000 And now we're ready for highly directional beam forming at Miller. 00:23:24.000 --> 00:23:32.000 We need a wave, and beyond well, one of the challenges is that that when it's blocked or when it's misaligned while the trainings happening, it's a temporary outage. 00:23:32.000 --> 00:23:38.000 And so the there's no data happening when it's doing this sector sweep. 00:23:38.000 --> 00:23:46.000 So no data happening. This is fine. If it takes a you know, a nanosecond. 00:23:46.000 --> 00:23:52.000 But how long is it taking? And so this is just to give you an example of the time scales. 00:23:52.000 --> 00:23:50.000 So this green data here is the maximum transmission time for data. 00:23:50.000 --> 00:24:08.000 You know millisecond timescale, and then this entire sector level sweep on both sides can can be in order. Magnitude. More than that. 00:24:08.000 --> 00:24:11.000 So we'd like this to be the opposite. We'd like data to take this long. 00:24:11.000 --> 00:24:18.000 And and then the and the overhead to be this tiny little part in front of the data. 00:24:18.000 --> 00:24:19.000 That's what you'd love. And why is this? 00:24:19.000 --> 00:24:27.000 Well, we've got to have all the headers that I mentioned, and it's we've got to test the different directions. 00:24:27.000 --> 00:24:32.000 So if we're frequently retraining, then this the overhead can now start to dominate the air. 00:24:32.000 --> 00:24:39.000 Time. And so you know, when you get a wireless device. 00:24:39.000 --> 00:24:40.000 And you see this this data rate up to something up to Gigabit, or or hopefully in the future, we'll say terrib. 00:24:40.000 --> 00:24:57.000 It? That's the physical error rate. But if we spend so much time doing control, plane and configuration and overhead, then the throughput can be much, much less. 00:24:57.000 --> 00:25:02.000 And the problem is going to get worse. As the beams get more directive. 00:25:02.000 --> 00:25:23.000 So if our beam is this directive, then you can imagine the sweep taking much longer to try them all, and versus you know, today's beams with a smaller number and lower frequencies are more looking like this left one so perhaps not as many different pairs to 00:25:23.000 --> 00:25:30.000 Check. So the question, I wanna point you to is, can we rapidly re steer the beams? 00:25:30.000 --> 00:25:30.000 And so to give you one of the the designs that we've been working on. 00:25:30.000 --> 00:25:41.000 I wanna? First briefly review a new device above a 100 Gigahertz. 00:25:41.000 --> 00:25:50.000 Or also, I should say, in millimeter way, to subtract, and the device is called a leaky wave. 00:25:50.000 --> 00:25:57.000 Intel, and it gets the name from from the construction here, which which is 2 parallel plates. 00:25:57.000 --> 00:26:04.000 They're metal plates, and there are a slot on the side, and the wave leaks out the slide. 00:26:04.000 --> 00:26:14.000 Hence the name Link you have antenna, so if you put in an electromagnetic wave here, instead of just going straight through the place, it's going to leak out now. 00:26:14.000 --> 00:26:22.000 The interesting property in the property that we'll use is that the angle that it leaks out will call this file. 00:26:22.000 --> 00:26:23.000 It is a function of the frequency. It's very simple function. 00:26:23.000 --> 00:26:31.000 There's speed of light, there's the distance between the 2 plates, and there's the frequency. 00:26:31.000 --> 00:26:42.000 And so it's an easy calculation to say, well, if I put it in frequency, which which I angle, does it come out, and that properties angular dispersion. 00:26:42.000 --> 00:26:50.000 And I said, You know, put new antenna in quotes, because it's really from from a design from 1940, and they noted the same property that I just told you so. 00:26:50.000 --> 00:27:03.000 New new for us to be reusing it at higher frequencies, and I'll hopefully convince you that this is extremely valuable at higher frequencies. 00:27:03.000 --> 00:27:03.000 So one of the uses and this predates our own work. 00:27:03.000 --> 00:27:07.000 But one of the uses was a filiki wave antenna was for beam. 00:27:07.000 --> 00:27:11.000 Steering so beam steering. It's more and more challenging with phased arrays. 00:27:11.000 --> 00:27:19.000 It'd start to go to 300 Gigahertz on above. 00:27:19.000 --> 00:27:26.000 And so this was the highest frequency beam steering, mechanism to date, so to truly steer Terra hurts beams. 00:27:26.000 --> 00:27:42.000 They were using liquid antennas and utilizing this property, that if you want to steer more towards broadside to the antenna, then simply lower the frequency of, and then, when you want to, or I'm sorry raise the frequency, and then as you want to go more towards broadside. 00:27:42.000 --> 00:27:52.000 Here, then then lower the frequency and the remarkable thing about this design is, is, its simplicity. It's one antenna beam sting. 00:27:52.000 --> 00:27:55.000 So normally we like. I showed you with the Wi-fi design. 00:27:55.000 --> 00:27:59.000 There's a phase array with this Myo to do. 00:27:59.000 --> 00:28:03.000 Steering, but so we've got an ability to steer with the leaky way of antenna. 00:28:03.000 --> 00:28:08.000 Now the question is, Oh, where do we steer? Where is the client? 00:28:08.000 --> 00:28:26.000 So our idea was what we call the Terra hurts Rainbow, and so the idea is that if we've got the angular dispersion property in a device, then we can take a high frequency pulse which in essence we can view as containing all frequencies so we're using 00:28:26.000 --> 00:28:40.000 A light analogy in this slide, where, if we send in a pause with with all frequencies here analogous to white light, then the leak wave antenna separates it according to direction, the the lowest frequency components in your broadside and higher frequency components near 00:28:40.000 --> 00:28:47.000 Parallel, and this emits in all directions, and then all the each direction has its own spectral signature. 00:28:47.000 --> 00:28:50.000 So in other words, this direction here is more towards the yellow component. 00:28:50.000 --> 00:28:59.000 So the receiver can identify its direction with respect to the transmitter base only on the frequency component. 00:28:59.000 --> 00:29:05.000 And so we call this one shot location discovery, because this only takes a nanosecond. 00:29:05.000 --> 00:29:11.000 This pulse. So remember before we're in milliseconds, orders of magnitude, slower to try to find the direction. 00:29:11.000 --> 00:29:12.000 And now we've got a single nanosecond pulse, and then all receivers. 00:29:12.000 --> 00:29:27.000 I'm showing one receiver here, but by knowing where you are on the rainbow, then you can immediately identify your direction with respect to the transfers. 00:29:27.000 --> 00:29:42.000 So the remarkable thing about this is the things that it doesn't have so it doesn't have the trial and error testing it doesn't even have phase information so there's no need to send a preamble that was one of the the problems with the prior methods is first had to send all those 00:29:42.000 --> 00:29:46.000 Preamble so you could start to use phase information. 00:29:46.000 --> 00:29:46.000 With training, and so on. And that's not needed here, and there's no array. 00:29:46.000 --> 00:29:52.000 It's again a single antenna. 00:29:52.000 --> 00:30:08.000 So how do we use this in practice? Well, Maxwell's equations can predict this peak emission angle curve given here, but it can also characterize how you fall off from the peak. 00:30:08.000 --> 00:30:16.000 So it's not that one frequency goes in exactly one direction, but one frequency has different intensities, in other in different directions. 00:30:16.000 --> 00:30:27.000 So you can start to view. If I'm at 30 degrees as a receiver, then I'll see the peak frequency around 300 gigahertz. 00:30:27.000 --> 00:30:33.000 But I'll also see some fall off plus and minus about a 100 Gigahertz. 00:30:33.000 --> 00:30:57.000 So I'll get this entire signature in space, depending on on my angle 00:30:57.000 --> 00:30:52.000 And when we move this to an experimental component that we we have here original theory predicted hemap. 00:30:52.000 --> 00:31:13.000 And then here is an experimental one, and so each angle now has its own spectral signal signature, and so we can use the original theoretical one, or we can enhance it with a with a data driven one that matches, our our antenna characteristics and you can see again, here if I take 00:31:13.000 --> 00:31:17.000 30 degrees is the same example we've got the same peak around 300 Gigahertz, and then we're we're falling off. 00:31:17.000 --> 00:31:26.000 But we've got some interesting other artifacts that a data driven model could could could exploit. 00:31:26.000 --> 00:31:33.000 But one is the model. Here only use the first transverse electric wave mode. 00:31:33.000 --> 00:31:41.000 But here you can see these other 2, 3 and 4, which could be also used to enhance the signature, and also you could. 00:31:41.000 --> 00:31:55.000 You can see how we get some spreading here at the very, at the very lowest angles that it's not in the m model due to the model making ideal assumptions about the leaky way of intent. 00:31:55.000 --> 00:32:04.000 For example, that the parallel plates are perfect conductors, and infinitely thin, and so on. 00:32:04.000 --> 00:32:13.000 So, how does all this work? So we did some experiments, and and this was my former Phd. Student. 00:32:13.000 --> 00:32:18.000 Yasmin, and and and her thesis work and the experiments show. 00:32:18.000 --> 00:32:24.000 If we have a measured angle versus an estimated angle, and then ideally, we want to be on the red line that we've got this this center region where we're just extreme high accuracy. 00:32:24.000 --> 00:32:31.000 And and that's the sweet spot in these curves here. 00:32:31.000 --> 00:32:35.000 So that's as we're going here, and we get a nice differentiation. 00:32:35.000 --> 00:32:53.000 And then we get a little more error as we start to go up sharp here and sharp here, where we start to deviate some from from the ideal due to the those those curves being harder differentiate at those points. 00:32:53.000 --> 00:32:58.000 So we're less than 5 degrees error in over 80% of the cases. 00:32:58.000 --> 00:33:18.000 And again the key about all of this being in a single shot, one of the remarkable inhabitants to this method, that it wasn't by my own group, but I wanted to share was use of this method for radar, and so radar sweeps and all the directions 00:33:18.000 --> 00:33:24.000 mechanically shown here or electronically, and gets a signal back. 00:33:24.000 --> 00:33:27.000 And to show an example of the rainbow used for radar. 00:33:27.000 --> 00:33:30.000 You. You can really imagine how this is. Gonna happen if you've got a rainbow type of transmission and then you reflect off of it. 00:33:30.000 --> 00:33:40.000 Object, then, to know where the object is is simply a matter of saying Well, which color did I get back? If I got back? 00:33:40.000 --> 00:33:46.000 You know, right on the edge of green and yellow. Then obviously, you're at this angle. 00:33:46.000 --> 00:34:07.000 And so this is Middlemen's group, and at Bron University, and they called it Terra Hurz radar, with a leaky wave antenna, and the remarkable thing about this experiment is is the scale here. 00:34:07.000 --> 00:34:05.000 So the blue is where the object really was. It's a metal object that they're moving around. 00:34:05.000 --> 00:34:17.000 And the red is their sme, and and if you look at the scale now, we're here in millimeters. 00:34:17.000 --> 00:34:17.000 So there able to tell you where this object is. Millimeter scale. 00:34:17.000 --> 00:34:28.000 Oh, accuracy! A degree or so, an angular accuracy! 00:34:28.000 --> 00:34:32.000 And again, with a single antenna, and they were able to do real-time computations and so on. 00:34:32.000 --> 00:34:52.000 So this is what I was alluding to. With this the the fusion of communication sensing so with devices, the future devices that it won't be desktop inferences, but it will now get down to millimeter scale inferences, so really exciting times for the 00:34:52.000 --> 00:34:56.000 Future, of fusing communication, sensing, and my last point about leaky wave antennas is, it is? 00:34:56.000 --> 00:35:03.000 If you're if you're a skeptic thinking, you'll never see 2 parallel plates sticking out of your iphone. 00:35:03.000 --> 00:35:18.000 I will agree with that. And so one of our colleagues and colleagues, Customs and Gupta Princeton, built a circuit to realize. 00:35:18.000 --> 00:35:36.000 He called it terrifts prison, and he this is only a year after the prior work on that where he built a circuit to do threed localization, and also not only that, but we use a 100 gigahertz up to 900 gigahertz so we weren't being 00:35:36.000 --> 00:35:42.000 Careful with how much spectrum you use, but in a real implementation you'd have restrictions of the band. 00:35:42.000 --> 00:35:50.000 And so he use a smaller bandwidth, and you'd have restrictions of the band. And so he use a smaller bandwidth, and it works over a smaller bandwidth, and it works over a smaller bandwidth. You don't need the whole 9 gigahertz to realize this function. 00:35:50.000 --> 00:36:09.000 So that's F. So the takeaway on on on this part of the talk, and on sensing and communication, is the the old way of doing it was, if phased, array, unique id for each being that you've got a number. 00:36:09.000 --> 00:36:24.000 All these beans, put a header on it, do trial and error, testing with each header and the scale of all this is micro microsecond and if you want to know the location, then you do traditionally rate processing with new devices like the leaky wave 00:36:24.000 --> 00:36:31.000 Antenna, and new physics, or are really, you can say it's old physics, but physics is only now relevant. 00:36:31.000 --> 00:36:34.000 That actually angular dispersion is there at lower frequencies. 00:36:34.000 --> 00:36:41.000 But it's such a trivial effect that we don't even bother modeling and are mentioning it papers. 00:36:41.000 --> 00:36:43.000 But now it's front and center, with wide bandwidth. 00:36:43.000 --> 00:36:49.000 And so now with the single antenna, we can realize different spectral signatures. 00:36:49.000 --> 00:36:53.000 Do all angles at once, and be done with with localization in the nanosecond scale, and then also apply this. Not just. 00:36:53.000 --> 00:37:03.000 Where we were doing it for purposes of communication. But you can also do it for purposes of environmental sensing. 00:37:03.000 --> 00:37:10.000 And so I showed you the fast, high resolution radar not to illustrate that. 00:37:10.000 --> 00:37:13.000 Okay, so the for the final part of the talk I wanted to talk about a scenario. 00:37:13.000 --> 00:37:30.000 Go back to this blockage scenario. So I gave you a scenario previously when I talked about beamstering and reeering, and I said, Well, there's an obstacle. 00:37:30.000 --> 00:37:39.000 But I made us get a little bit lucky in that scenario, and I said, there happens to be a wall right here, and so we're going to to create a specular non line of sight path over using that wall, right? 00:37:39.000 --> 00:37:48.000 So I gave us a lucky scenario. 00:37:48.000 --> 00:37:58.000 Now I want to address the case of suppose there's no wall, suppose we're blocked, and and we don't have a good 9 of sight path available. 00:37:58.000 --> 00:38:06.000 It is a game over, is it? Just forget terabyte per second, and that's it. 00:38:06.000 --> 00:38:13.000 So the objective that I wanna talk about today is, can we curve the beam around the obstacle? 00:38:13.000 --> 00:38:24.000 And you're probably thinking no way, you know laser beams they don't curve right flashlights they don't give curve beams so surely it's not possible to curb it. 00:38:24.000 --> 00:38:28.000 Being. So, let's ask. Let's ask the theoreticians. 00:38:28.000 --> 00:38:35.000 Let's see, can we do a a curved beam, and will our friend over here, Maxwell, who's scowling at us? 00:38:35.000 --> 00:38:36.000 Will he object and say that's going to violate my equations, and therefore it won't propagate. 00:38:36.000 --> 00:38:49.000 Well. Fortunately, a few decades ago the theoretical physicist told us that there's a being called an Erie beam, and I'll tell you why it got it. 00:38:49.000 --> 00:38:52.000 That name that has a quadratic path. So we can have a curve being, and, moreover, it satisfies Maxwell's equations. 00:38:52.000 --> 00:39:05.000 Solution to hell, voltes, equations that satisfies all the key equations for the wave to propagate. 00:39:05.000 --> 00:39:09.000 And so the question that I'd like to turn to is well, the wave exists. 00:39:09.000 --> 00:39:14.000 So in theory. So how could we generate one to curve around the obstacle? 00:39:14.000 --> 00:39:18.000 That's our goal. Ultimate goal curve around the obstacle. 00:39:18.000 --> 00:39:23.000 So the way it works is is it's all about the launch of the way. 00:39:23.000 --> 00:39:37.000 So we have to launch the way with an initial spatial profile and a spatial profile that this I'll promise this will be the only equation I'll give, and this is, we'll call this 5 X and this is a spatial profile of of the waves launch. 00:39:37.000 --> 00:39:54.000 At position X. So X. Here is just like antenna array, or different points in space, where you launch the wave, and so a profile means just simply what's the amplitude and phase of the wave, you're launching. So think of it. 00:39:54.000 --> 00:40:01.000 As launching many little mini waves, and each of these Mini waves has a gain in phase given by this function. 00:40:01.000 --> 00:40:04.000 Now, if 5 x was simple, and I'll show you a simple one in the next slide. 00:40:04.000 --> 00:40:05.000 Then, okay, we can understand what it's doing and what. 00:40:05.000 --> 00:40:15.000 But unfortunately this profile is given an area function. Hence the Erie beams name, and this is a function you probably never seen before, so I'll plot it. 00:40:15.000 --> 00:40:22.000 And this is it here in blue. So that's this. 00:40:22.000 --> 00:40:28.000 This blue function is the 5 x, so it's an area function with an exponential modulating. 00:40:28.000 --> 00:40:31.000 It. And so this function goes up and down. It's oscillating. 00:40:31.000 --> 00:40:36.000 It's not linear, and that it peaks here, and then it goes down to 0 on the right side. 00:40:36.000 --> 00:40:43.000 So the theory says, if you initially launch a wave with this profile, then you get this on the right. 00:40:43.000 --> 00:40:52.000 So what is this? So this here, this is the same X axis as this is, and the the wave is propagating up. 00:40:52.000 --> 00:40:57.000 And so down here at Zoom, equals. 0 is where we launch the wave. 00:40:57.000 --> 00:41:01.000 So the heat here the peak is the same as the peak here. 00:41:01.000 --> 00:41:07.000 So you see, this is the highest here at X, a little less than one, as the most intense. 00:41:07.000 --> 00:41:12.000 The wave is most intense here, right there. Cause we put an amplitude of one there. 00:41:12.000 --> 00:41:15.000 Now we, if we launch the wave there, what does it do? 00:41:15.000 --> 00:41:15.000 It starts curving in space. So it's curving in space. 00:41:15.000 --> 00:41:21.000 That's the the red part is is the highest intensity. 00:41:21.000 --> 00:41:27.000 Part of the way, and there's nothing here, and then there's some interference pattern on that side. 00:41:27.000 --> 00:41:27.000 So the point is, if we can launch this, then we get this. 00:41:27.000 --> 00:41:38.000 That's what the theory tells us. All right. So let me give you a little context of these 5 x functions. 00:41:38.000 --> 00:41:44.000 And so those of you who are doing wireless probably haven't seen the area function before. 00:41:44.000 --> 00:41:48.000 But you know about being foreign, so let me give you a quick fee for it. 00:41:48.000 --> 00:41:51.000 So what if you just wanted to be it? The no blockage. 00:41:51.000 --> 00:41:56.000 Just steer the beam glass in a certain direction. Well, what is the function? Phi. X. 00:41:56.000 --> 00:42:12.000 Look like there. Well, then, the amplitude on all antennas is constant, and the phase, remember, those were the weights that I gave when I was describing it, to to 11 ax the weights all they are are different delays across the antennas and so to steer a beam all you do is a 00:42:12.000 --> 00:42:18.000 Linear delay, profile, and so if you have a linear phase, profile then you'll steer the beam. 00:42:18.000 --> 00:42:25.000 If this is 0, the beam goes straight, and if it's and then you can just control the slope and that controls the direction that you steer to. 00:42:25.000 --> 00:42:28.000 So this is what we're doing today. We're doing line line right? 00:42:28.000 --> 00:42:33.000 And now we're saying, Aye, Bean, let's replace this line with this blue function and then let's replace the phase. 00:42:33.000 --> 00:42:40.000 Concent gradient with something that's going positive and negative 00:42:40.000 --> 00:42:46.000 Okay, so how do we launch this way? So we're, I said, at every point in X, we wanna launch a way with a particular profile. 00:42:46.000 --> 00:42:53.000 So if we're at this point in X, we want to transmit the maximum amplitude. 00:42:53.000 --> 00:42:56.000 No phase offset, because this is one that's 0. 00:42:56.000 --> 00:42:59.000 And that's just a little to the left of the center. 00:42:59.000 --> 00:43:07.000 And then over here, for example, these are both 0. So transmit nothing over there but we've got to discretize it and radiate ways. 00:43:07.000 --> 00:43:11.000 And so we wanna radiate away from each of those points in space as many as we can. 00:43:11.000 --> 00:43:18.000 Now suppose we said, use the antenna array. We already have antenna array, and Wi-fi take those whatever 8 antennas and approximate the area weight. 00:43:18.000 --> 00:43:29.000 Well asking to do that is an essence asking to approximate this function with 8 points. 00:43:29.000 --> 00:43:27.000 And so you would probably think well with 8 points. That function will. 00:43:27.000 --> 00:43:41.000 That will not be a good approximation. We need a much higher resolution, approximation to this function to realize curvature. 00:43:41.000 --> 00:43:53.000 And so our approach is to use a medicine and a Meta service is a 2D structure, and it's it's got sub wavelength. 00:43:53.000 --> 00:43:57.000 Meta atoms that are going to to help us shape this this wave profile. 00:43:57.000 --> 00:44:00.000 So the amplitude and the phase of this profile. 00:44:00.000 --> 00:44:04.000 So let me show you how we did that. 00:44:04.000 --> 00:44:11.000 So first of all, the building block of A, of a Meta service is a meta atom, so it's the radiating element. 00:44:11.000 --> 00:44:15.000 And if you like, just think of this as an antenna rather than a dipole. 00:44:15.000 --> 00:44:20.000 This is a a circular structure that metallic and radiates. 00:44:20.000 --> 00:44:37.000 Now this particular one is called a split ring resonator, because it's a ring, and it's got a split in it, and the parameters that are going to control the amplitude and phase of polarity are the 3 parameters i'm showing here 00:44:37.000 --> 00:44:41.000 What's the radius of it? What's the opening angle? 00:44:41.000 --> 00:44:47.000 And what's the orientation angle? So what we're going to do is we're going to radiate these. 00:44:47.000 --> 00:44:50.000 Split ring resonators with with a normal weight. 00:44:50.000 --> 00:44:53.000 We're just gonna blast electromagnetic energy at these. 00:44:53.000 --> 00:44:53.000 But then we're going to rotate them in a particular way. 00:44:53.000 --> 00:45:02.000 So that they radiate out with the amplitude phase, and flurry that we want. 00:45:02.000 --> 00:45:05.000 So think about this as a atom by atom configuration. 00:45:05.000 --> 00:45:12.000 We're laying all this out to make all these atoms work together to give us a function by X. 00:45:12.000 --> 00:45:22.000 So to put it all together, we're gonna come back to our function in blue that is going to allow us to to create the air being. 00:45:22.000 --> 00:45:30.000 And then for each point in space, we're gonna select the Meta atom that gives us the amplitude shown up here that so this is one that's the highest amplitude, and that's a 0 phase. 00:45:30.000 --> 00:45:43.000 And then here we want to lower amplitude down here less less than minus one and also negative phase, because it's negative. 00:45:43.000 --> 00:45:54.000 So notice, this points down. So this one's pointing down this one's pointing up, and then here, if it's horizontal, then the opening sort of only aligned, then this is our attenuation. 00:45:54.000 --> 00:45:56.000 So just take these align them up in columns. 00:45:56.000 --> 00:46:06.000 So these are 5 x, so there's x, and then each column here is 8 a discretization of X, so we're basically took the function to scrutize it. 00:46:06.000 --> 00:46:06.000 And put it on a Meta service. And now, wonderful! The question is, how do we build this? 00:46:06.000 --> 00:46:20.000 How do we radiate? Can we actually transmit in these bands? 00:46:20.000 --> 00:46:34.000 So there are 2 methods to fabricate these meta surfaces that that I'd like to talk about, and I'm gonna show you an experimental result, using the one on the left. 00:46:34.000 --> 00:46:39.000 But I wanna talk about 2 different classes of Meta services 00:46:39.000 --> 00:47:02.000 So on the left we have a hot, stamped, static Meta service, and this is a this is developed by the middleman lab that I mentioned before, and the idea there is to to take a meta-surface layout like this and print it on normal office laser 00:47:02.000 --> 00:47:09.000 Printer paper, and then once you have that laser printer paper to take that along with a metallic sheet that's metallic powder. 00:47:09.000 --> 00:47:26.000 This is the same metallic sheets that you can buy in any craft store and and run that through a heating language, and at the output of the laminator the metal will bond to the toner from the printout. 00:47:26.000 --> 00:47:32.000 And so this is a office supply printed meta-surface that realizes the function that I just showed you. 00:47:32.000 --> 00:47:40.000 So we lay out an an area profile on this particular metaurur using this method just rapid prototyping, we can do it in minutes, and it's inexpensive. 00:47:40.000 --> 00:47:47.000 Everything is just normal office supplies. So really simple to get medica services. 00:47:47.000 --> 00:47:54.000 Now the downside is once we print it, we can't change it. 00:47:54.000 --> 00:47:57.000 So if you wanted to say Well, now, we're curving to the right. 00:47:57.000 --> 00:47:59.000 What if the obstacle moves and we need to curve to the left? 00:47:59.000 --> 00:48:09.000 Right, then you need to reconfigure the Meta surface and we obviously can't reconfigure something that's printed. 00:48:09.000 --> 00:48:25.000 So how do you reconfigure a service? So this is a clever design by the Patrick Sin, Gupta group in Princeton, and the idea here is to take a metallic ring and the short and open different segments of 00:48:25.000 --> 00:48:31.000 The ring. So here you wanted to ring, and you wanted this part open here, you wanna ring with that part open. 00:48:31.000 --> 00:48:29.000 So how do you dynamically do that? You can't physically rotate the metal. 00:48:29.000 --> 00:48:43.000 So his idea was, well, we'll create dynamic opens and shorts, and that will enable us to to to emulate that we had rotated this in a dynamic way. 00:48:43.000 --> 00:48:52.000 So upcoming. We, we will have some programmable meta-service results. 00:48:52.000 --> 00:48:58.000 But for now the first step is to show that we can do it with with the static meta surface. 00:48:58.000 --> 00:49:05.000 So I wanted to show you one results along these lines. 00:49:05.000 --> 00:49:09.000 And so first of all, the setup we've got, it's a small scale experiment. 00:49:09.000 --> 00:49:18.000 So the axis of the propagation distance. Here these are centimeters instead of meters, and why is it so small? 00:49:18.000 --> 00:49:18.000 It's a limit of our in lab method. 00:49:18.000 --> 00:49:31.000 Here we have a sub micro watt transmitter. So a real commercial system would be you know, tens or hundreds of millil lots for communication. 00:49:31.000 --> 00:49:40.000 This is a time domain, spectroscopy, transmitter, and so for spectroscopy, there in the sub micronaut regime, so it's got very low power. 00:49:40.000 --> 00:49:47.000 But nonetheless we can do this at a smaller scale, and we've hot stammed a Meta service. 00:49:47.000 --> 00:49:59.000 It's transmissive. So we we illuminate it with with the pulse from the system, and then the meta-surface design at 150 gigahertz. 00:49:59.000 --> 00:50:01.000 And then we measure the output of the beam. After going through our meta surface. 00:50:01.000 --> 00:50:08.000 And so here, so again, this is the same X that we've been using the entire time with our function. 00:50:08.000 --> 00:50:17.000 So we have you have to view that the meta-surface is right here. 00:50:17.000 --> 00:50:27.000 This is Z, where it's propagating. So the meta surfaces is at Z equals 0 spanning spanning the X equal minus 3 to 5. 00:50:27.000 --> 00:50:32.000 And so the electromagnetic wave goes through the Meta service. 00:50:32.000 --> 00:50:34.000 And then this is what happens to it. It has a curve profile. 00:50:34.000 --> 00:50:44.000 So here's the heat map of it. So this is showing the strongest intensity point as a function of Z, and you can see this. 00:50:44.000 --> 00:50:52.000 This red line is the fitted trajectory. So we're fitting the maximum of the the beam as it goes forward in space. 00:50:52.000 --> 00:51:00.000 So it's not just going off to the right. So if we're beam steering, then it would follow straight to Directory like the right. 00:51:00.000 --> 00:51:12.000 So ideally. And this this is work coming up, we'll be able to put a obstacle right there, and we'll be able to show you if you use straight beam doesn't matter how much power you give it. 00:51:12.000 --> 00:51:18.000 It's not gonna blast through the object, and it'll be blocked. 00:51:18.000 --> 00:51:27.000 And yet we will be able to curve around it and get energy up here, and of course we'll get energy farther and farther as we've put more power on this as well. 00:51:27.000 --> 00:51:31.000 It's also, you know, the theory will a great guide here? 00:51:31.000 --> 00:51:36.000 The theory told us we would be curving on this black curve, and we're curving on this red curve. 00:51:36.000 --> 00:51:43.000 We thought it was a pretty good match because of some of the things that were doing as approximations. 00:51:43.000 --> 00:51:47.000 First. We're always spatially sampling this function. 00:51:47.000 --> 00:51:49.000 Remember, I said 80 antennas would be lousy. 00:51:49.000 --> 00:51:56.000 We've got many more, maybe 1,000, by 1,000 different Meta atoms. 00:51:56.000 --> 00:52:04.000 But we're still sampling in space and the Meta atoms are imperfect because they're printed with a normal laser printer. 00:52:04.000 --> 00:52:08.000 The metal on the printer. Is this bonded metallic powder? 00:52:08.000 --> 00:52:12.000 It's the theory assumed. It's a perfect conductor, and so on. 00:52:12.000 --> 00:52:22.000 So we'll we'll declare victory for for the curvature and and and and that can give us opportunity instead of the straight line of sight. 00:52:22.000 --> 00:52:23.000 Pat, so future work, we'll call them airy network. 00:52:23.000 --> 00:52:35.000 So with, you know. Generating a beam is is a important first step, and you know, Hi Fis. 00:52:35.000 --> 00:52:38.000 All around the lab. But there's a lot more work to be done. 00:52:38.000 --> 00:52:38.000 First of all, at scale experiments, including the obstacle. 00:52:38.000 --> 00:53:02.000 So we've got to get real size transmissions to tens of meters, and then real size obstacles, people, size, obstacles, furniture, size, obstacles, and those experiments are are coming up soon for us on the control plane you know how do you AIM the 00:53:02.000 --> 00:53:07.000 Curveball. So you know, I like the metaphor of this, the soccer player. 00:53:07.000 --> 00:53:11.000 And you know, if I can't kick like this. 00:53:11.000 --> 00:53:16.000 And so it just seems like magic, just just like light bending seemed like magic. 00:53:16.000 --> 00:53:22.000 Are having a curve. Trajectory seems like magic, but it's all in the launch, so it's all he does here. 00:53:22.000 --> 00:53:27.000 There's no manipulation afterwards. Right? And think of our meta service the same way. 00:53:27.000 --> 00:53:31.000 It was all about that 5 x function. How do you launch the beans? 00:53:31.000 --> 00:53:31.000 And then, after that launch, it's gonna go where you programmed it to go. 00:53:31.000 --> 00:53:45.000 Now that's a wonderful, powerful functionality. But you know those of us in networking know that if you have a a function, new functionality, you have to have a control plane to control that functionality. 00:53:45.000 --> 00:54:01.000 And so control playing questions are, you know, where is this target, you know, is Aming for the upper corner of of the net right? But he's blocked, you know. 00:54:01.000 --> 00:54:04.000 There's some blockers here, and he can maybe partially see where he's going. 00:54:04.000 --> 00:54:13.000 Maybe can't even see it. Maybe he just kind of knows where it is what's the best curvature profile and steering angle that you know. 00:54:13.000 --> 00:54:20.000 Do you wanna try to curve way out as much as you can, or just go, you know, super close, with less curvature. 00:54:20.000 --> 00:54:24.000 You curve last curve right? Lots of possibilities. 00:54:24.000 --> 00:54:24.000 And then, how do we program the Meta surface to realize these functions? 00:54:24.000 --> 00:54:37.000 That once we get to programmable Meta service we'll have a different set of imperfections that will be a discretization. 00:54:37.000 --> 00:54:47.000 Just like I talked about code books back and Wi-fi will have Meta service code books and will have to do this within the constraints of the of the codebook. 00:54:47.000 --> 00:54:52.000 And then data playing. Yeah, we've got to send information right? 00:54:52.000 --> 00:54:56.000 The eye and size is, how do we get the information from the transmitter to the receiver? 00:54:56.000 --> 00:55:02.000 And I was showing results for in essence a single tone, so if we want to do high data rates, we have to use wider bandwidth. 00:55:02.000 --> 00:55:08.000 And so as we use wider bandwidths, the curvature will be a little bit different. 00:55:08.000 --> 00:55:13.000 And then how do we? How do we compensate for that? 00:55:13.000 --> 00:55:24.000 How do we analyze that? So the data playing part will also be a challenge moving forward, and the I'm showing here a a paper. 00:55:24.000 --> 00:55:24.000 The 3 of us just wrote it on my web page. If you're interested, it doesn't have our experimental results yet. 00:55:24.000 --> 00:55:48.000 But some of the theory behind these types of beings, and also others so some of the upcoming challenges, just to just to wrap up that I see in wireless, as we have new antenna and meta-surface capabilities, this really 00:55:48.000 --> 00:55:55.000 Changes, the the Foundational building block. It's something radically new at the core of wireless. 00:55:55.000 --> 00:56:02.000 And so once that basic building block changes everything else above, it changes as well. 00:56:02.000 --> 00:56:09.000 So how should the network be designed? An architect it? What does the network look like? 00:56:09.000 --> 00:56:09.000 And and it's layout optimization and management. 00:56:09.000 --> 00:56:16.000 So how do we dynamically we don't always want to curve a beam. 00:56:16.000 --> 00:56:19.000 So when do we even invoke those features to curve teams? 00:56:19.000 --> 00:56:22.000 And how do we know what the right criteria is? 00:56:22.000 --> 00:56:27.000 A second thing I'd like to talk about is I ended. 00:56:27.000 --> 00:56:35.000 The talk with these higher frequencies, but we'll still always be using the lower frequencies. 00:56:35.000 --> 00:56:49.000 And so those lower frequencies give us coverage and penetration that we just can't get with the higher frequencies which give us these wonderful data rates and sensing so future networks will fuse all of these it won't be a choice one or the other we will use them. 00:56:49.000 --> 00:57:01.000 Together. And so that gives a new opportunity for network architectures, design, analysis of sensing is it's this wide. 00:57:01.000 --> 00:57:04.000 A spectrum is this unprecedented? We've never had it that way. 00:57:04.000 --> 00:57:07.000 Think about Wi-fi. We have 2.4 5 gigs. 00:57:07.000 --> 00:57:12.000 We get a factor of 2, right? Here, we're talking about several orders of magnitude. 00:57:12.000 --> 00:57:20.000 I mean, it's just astounding. The diversity of the spectrum that we'll have moving forward the components. 00:57:20.000 --> 00:57:35.000 So our circuits, colleagues are, and system designers, giving us programmable low cost energy, efficient components that has to happen in in parallel to to have the next generation devices and then, lastly, 2 points. 00:57:35.000 --> 00:57:36.000 I didn't bring out in the discussion today, but I think are really important. 00:57:36.000 --> 00:57:44.000 Moving forward, and I wanna bring it out explicitly. 00:57:44.000 --> 00:57:50.000 One is, how powerful the physical models are at higher frequencies, that is it. 00:57:50.000 --> 00:58:08.000 Some of the physical models don't work at lower frequencies, but there's so much scattering and reflection off of every device that it, going back to say a model of a wave is just not helpful versus here you saw with the leaky wave antenna and you saw with 00:58:08.000 --> 00:58:24.000 the, airy being that the physical model I mean this is straight out of, you know Maxwell's equation that those are providing tremendous guidance, and so that guidance helps us not only as designers, but I think it will help us in the field as well that when we go 00:58:24.000 --> 00:58:23.000 to sensing and learning, and trying to understand the environment rather than be pure. 00:58:23.000 --> 00:58:35.000 Black black box that we can use that physical models to help to help with those inferences. 00:58:35.000 --> 00:58:56.000 And lastly, new applications, new use cases. So as we get to higher and higher data rates, that working with real scenarios, real devices, just like I talked about with technology for all in the beginning that taking technology to the ultimate end users is a super valuable feedback because it helps 00:58:56.000 --> 00:59:00.000 Optimize, what's really important to some of those users. 00:59:00.000 --> 00:59:04.000 And and that can also foster the next generation of designs. 00:59:04.000 --> 00:59:04.000 So to summarize, we started with technology for all. 00:59:04.000 --> 00:59:18.000 And the the Tower, and the diverse spectrum multi hopping, and then got to the one shot path, discovery and the tariffs, rainbow. 00:59:18.000 --> 00:59:28.000 And then I ended up showing you how we could take a meta-surface, put in an airy profile on it, and then generate a being that that curves in space. 00:59:28.000 --> 00:59:33.000 So, I'll end by acknowledging my Phd. 00:59:33.000 --> 00:59:32.000 Students, all alumni. This talk since I went back 20 years. 00:59:33.000 --> 00:59:38.000 That's it's got a excellent group of former students and end current students as well, who helped with the work. 00:59:38.000 --> 00:59:46.000 I showed that's ongoing. But all of them are on my group page, and then collaborators at rice and beyond. 00:59:46.000 --> 00:59:51.000 I mentioned papers and work from Dan Middleman, Joseph, Jornette, and Kazakhstan, Gupta. 00:59:51.000 --> 01:00:01.000 So the the papers that I cited from them are on their respective sites, sponsors. 01:00:01.000 --> 01:00:23.000 The army is funding some of our work and security, which I didn't mention today as well as some of the work in some Terahertz, Cisco and Intel have been long time sponsors of as well, and they've been a great input of not only understanding where 01:00:23.000 --> 01:00:19.000 Standards and industry are going, but giving a chance for us. 01:00:19.000 --> 01:00:31.000 The community to influence where they're going. And then, lastly, and and of course, most importantly, National Science Foundation. 01:00:31.000 --> 01:00:48.000 And so I again acknowledge and thank you all, not only for the invitation today and the chance to talk to you all, but also for the support that's just every slide on this presentation has had an assess support in it. 01:00:48.000 --> 01:00:57.000 So I thank you all for that, and with that I will conclude, and I'll be very happy to take a questions and comments 01:00:57.000 --> 01:01:10.000 Thank you, Edward, for a wonderful talk in demonstrating that you can bend light, at least at the right frequencies so the floor is open for questions and answers just a reminder to folks. 01:01:10.000 --> 01:01:24.000 We welcome all questions, and you can just put your questions in the, and we can go from there and while we're waiting, Edward, let me just start with at least one on the description. 01:01:24.000 --> 01:01:34.000 The meta surfaces and using them as reflectors or ways to get around the blockage. 01:01:34.000 --> 01:01:34.000 And the how do you see that scaling? Are there difficulties there? 01:01:34.000 --> 01:01:43.000 You talked about like, what if I have to get around a piece of furniture, or you know a car or something you might be this blocked by a larger object. 01:01:43.000 --> 01:01:48.000 How do you see that working. 01:01:48.000 --> 01:01:58.000 Right, yeah, I mean, that's great question. And I would say, it's a research question for us. 01:01:58.000 --> 01:02:07.000 I mean I on on the math side the math scales so that there's not we won't be breaking part of the beam as we try to bend more and more. 01:02:07.000 --> 01:02:24.000 Some of the things that that that are needed. If you wanted to really wide curvature, for example, is it, you know, one of the one of the physical limits is this, you know this this ultimate width here? Right? 01:02:24.000 --> 01:02:29.000 So you know, we're showing, you know, 8 cm. 01:02:29.000 --> 01:02:46.000 So if you want to keep bending and bending, then you've got to get larger with. So if there's a scenario such as a car where you wanted to, you know, get radar or different sensors bending around objects, then you could potentially use the larger surface of the 01:02:46.000 --> 01:02:48.000 Car to do things like that. But if you got a small device, then eventually you're going to be limited by this aperture here. 01:02:48.000 --> 01:03:00.000 As for how much spending you can have, but for sure, that's one of the challenges moving forward is to explore. 01:03:00.000 --> 01:03:11.000 You how we can overcome those, those those practical limits of transit power budget, and then aperture size of the transmitter 01:03:11.000 --> 01:03:20.000 Yeah, thank you. We have another question, what is the decay, rate or power loss of the parabolic weight? 01:03:20.000 --> 01:03:26.000 Yeah, great question. And we're we had a meeting about that yesterday where we were looking to do that calculation, because it's you can't just plug in. 01:03:26.000 --> 01:03:34.000 You know the freeze equation and say, Well, yeah, here's the. 01:03:34.000 --> 01:03:37.000 So, you know, we'll just do the path loss. 01:03:37.000 --> 01:03:40.000 But it is computable. So we have not done that yet. 01:03:40.000 --> 01:03:50.000 But that is something that is is is very important, and and we will do. 01:03:50.000 --> 01:04:04.000 And we're hoping it will come to some nice close form results where we're able to start using it'll come to some nice close form results where we're able to start using it, using it, say, in a link budget analysis just the same way, we would regularly 01:04:04.000 --> 01:04:06.000 use like a free sequence, and say, I lost a few dB. 01:04:06.000 --> 01:04:10.000 Here due to, you know, an object and a few dB. 01:04:10.000 --> 01:04:13.000 Here due to the receiver, and so on, and that, and that here we could set. 01:04:13.000 --> 01:04:22.000 Okay, if my total distance over the parabola is this, then I can do the calculations. 01:04:22.000 --> 01:04:30.000 And so we will for sure be giving some some good analysis of that coming up 01:04:30.000 --> 01:04:48.000 Yeah, thanks. Next question. The overhead using traditional phased arrays is large, which is eliminated and using Mickey wave antennas, what about the data using a single Mickey wave intel 01:04:48.000 --> 01:04:55.000 Yeah. So the so the leaky way of antenna, you know, we if I go back here? 01:04:55.000 --> 01:04:58.000 Yeah. So we didn't for our League. 01:04:58.000 --> 01:04:58.000 We have antenna, we, you know, we didn't send data over it. 01:04:58.000 --> 01:05:01.000 So we were doing the direction finding and didn't send data. 01:05:01.000 --> 01:05:16.000 So so one of the key things about the efficiency of coupling into the leaky wave intent for us, and then the total power. 01:05:16.000 --> 01:05:18.000 So, as I mentioned, our total power is very low, micro scale. 01:05:18.000 --> 01:05:25.000 And so, and the efficiency of coupling in to out of the parallel plates was something we didn't optimized because it was more of a direction finding thing that data thing. 01:05:25.000 --> 01:05:37.000 But but there are others who are who are using leaky wave antennas for data, and then that becomes front and center for them. 01:05:37.000 --> 01:05:47.000 You've got a couple in efficiently and you've got to emit at tens of milliwatt scales, if not 100 Millil Watts, you know, rather than what we're doing with microwa. 01:05:47.000 --> 01:05:46.000 So I'll leave that to the to the Cmos implementations. 01:05:46.000 --> 01:06:01.000 This this particular paper was about localization, but I believe that there's some other papers here showing the use of it for data. 01:06:01.000 --> 01:06:20.000 And so they are able to get their activity gained, although I don't have a specific number for you today. But for sure there's the possibility of getting the hydroactivity gains that we like face a race for even with the leaky event. Test 01:06:20.000 --> 01:06:28.000 Okay, thank you. Over what distances might the leaky way, the antenna technology work 01:06:28.000 --> 01:06:28.000 Yeah, so this is not a restricted distances. 01:06:28.000 --> 01:06:40.000 In any in any more severe way than than any other, intent. 01:06:40.000 --> 01:06:44.000 So once you get the beam forming game, you know. 01:06:44.000 --> 01:06:48.000 Now you actually can do a normal link budget analysis. So once you select your frequency band, then that frequency is going to propagate in that direction. 01:06:48.000 --> 01:06:58.000 Yeah, it. With path loss, according to freest equation. 01:06:58.000 --> 01:07:04.000 So from that, you're not gonna be distance constraint by the leaky way of antenna. 01:07:04.000 --> 01:07:08.000 In fact, one of the let's see if I can find the original patent here. 01:07:08.000 --> 01:07:09.000 When I when I yeah, this guy. So they were originally using this for for aircraft. 01:07:09.000 --> 01:07:15.000 So they wanted to point it up and see what what the relative angle was. 01:07:15.000 --> 01:07:24.000 Of the aircraft, and you could analyze this signal based on that. 01:07:24.000 --> 01:07:29.000 So this was being used for long distances, air to ground. 01:07:29.000 --> 01:07:41.000 So these are definitely able to realize the beam forming gains that your custom to, with traditional methods like phase array. 01:07:41.000 --> 01:07:50.000 Okay, yeah, makes sense. Yeah, so one more question in kind of a high-level question. 01:07:50.000 --> 01:07:59.000 I guess is we hear a loud about 6 G, and so do these technologies and ideas play into the next generation. 01:07:59.000 --> 01:08:03.000 And how, maybe when. 01:08:03.000 --> 01:08:09.000 Yeah. So for sure. I mean, I think 6 g, wi-fi 8, you know. 01:08:09.000 --> 01:08:14.000 I think these will be where you'll see frequencies. 01:08:14.000 --> 01:08:29.000 60 years, 100 Gigahertz, beyond taking off, and one of the wonderful things, too, about about standards, especially your Wi-fi, that I know the best. 01:08:29.000 --> 01:08:48.000 But is that they leave so many aspects of it that the vendors can do on their own. In other words, all the vendors are allowed to innovate on top of the standard, so they leave it flexible enough and so what you can end up with 01:08:48.000 --> 01:08:58.000 Is a standard that even if it has a say, a sector suite that that devices can still add on additional methods in order to. 01:08:58.000 --> 01:09:03.000 In order to do that process better so, let me give a specific example. 01:09:03.000 --> 01:09:13.000 The standard doesn't tell you what type of intended use you can use. 01:09:13.000 --> 01:09:10.000 You can use a, you can switch among sector intensities. You could use a phased rate and so on. 01:09:10.000 --> 01:09:27.000 So none of that is prescribing the standard. So you can potentially do things on top of the standard and so on. So none of that is prescribing the standard. 01:09:27.000 --> 01:09:29.000 So you can potentially do things on top of the standard and use it as a vendor. Specificspecific enhancement to the standard. 01:09:29.000 --> 01:09:39.000 But broadly, the higher frequencies are very much a part of 6 g. 01:09:39.000 --> 01:09:44.000 And that is because of the data rates is really my first slides. 01:09:44.000 --> 01:09:52.000 I moved to moving forward the data rates, the wide bandwidth, and the sensing resolution join sensing and communication. 01:09:52.000 --> 01:09:58.000 Those are all 6 G. You know. Wi-fi, 8 sort of technologies that we want. 01:09:58.000 --> 01:10:10.000 So utilizing. So when we get to those frequencies, and that directivity, then we're very much going to want different methods for fast localization. 01:10:10.000 --> 01:10:17.000 And you know, curving means, and so on. So absolutely. The context of this, I would say, is very much 16 01:10:17.000 --> 01:10:22.000 Okay, thank you. That's kind of the end of our questions. 01:10:22.000 --> 01:10:28.000 For now in the we also got menu thumbs up and thank you for the presentation. 01:10:28.000 --> 01:10:32.000 So again. Thank you so much as we've seen. 01:10:32.000 --> 01:10:36.000 The presentation will be, has been recorded and will be available. 01:10:36.000 --> 01:10:40.000 So if you missed some of all that, you'll be able to replay it. 01:10:40.000 --> 01:10:49.000 And just reminder for and I said that there is an office hours this afternoon at 3 30, and so bring further questions and discussions. 01:10:49.000 --> 01:11:01.000 There. So I think with that we will conclude this seminar and thank you again, Edward, for wonderful presentation.