UWB vs IMU vs Ultrasonic RTLS Compared | Marvelmind

0:00 Our colleagues, today I will be talking about indoor navigation positioning technologies overview—not about Marvelmind solution, not about our particular implementation of indoor navigation and positioning, but overall, overall overview. Why? Because we're being asked many times: what's your opinion about UWB? What's your opinion about ultrasonic? What's different between you and other solutions? So this pretty long presentation is the summary of our views on different technologies. Of course, with the focus on industrial applications, because we are mostly in industrial, and surely we are biased. But for this particular application, it's not so much about Marvelmind, but Marvelmind's point of view of other technologies, including Marvelmind.

0:57 Of course, what's the problem? The problem is obvious: GPS doesn't work indoors, and precision of GPS—five to ten meters—is not enough for the robot to move from point A to point B and not to kill anyone. There are many other solutions: ultra-wideband, UWB, Wi-Fi, audio meters, others. Many, but all of them have some serious limitations: usually precision, or price, or size, or power consumption. In one way, you know that you do need to know the location of your mobile object—robots, drones, people—in order to do something autonomous. I will be using much of the technology. You can later refer to this page to see what I meant. I will be talking about many types of indoor positioning.

1:54 Positioning methods—but not all of them. So for example, I just noticed that we even didn't cover the magnetic tapes or wires because they are somewhat used like GPS, but they are not here because, first of all, they are really old-fashioned, even though they're still widely used. And second, it's slightly different, slightly different terms of location. But yes, for some uses, they are widely used and they're almost simple. We are not covering them because they are pretty trivial. Now, about types of indoor positioning methods. Now, the first one is trilateration. Of course, our exercise-based methods are also till trilateration, but when I mean trilateration, I mostly use time-of-flight. So it's kind of equal in my terminology. What are those? Yes, obviously, trilateration—ultra-wideband, and of course, our technology, which is a

2:53 combination of a size clock synchronization between stationary and mobile beacons and the modem over radio, and precise distance measurement using ultrasonic pulses. So it's time-of-flight of ultrasonic. In the case of UWB, it provides bent. It's time-of-flight of very short pulses, so it's time-of-flight of radio. In the case of GPS, it's also time-of-flight of radio, but slightly different technologies. Let's say frequencies are exercise-based—widely available, known. But they are not designed for positioning per se. They're designed for something else, and the fact that you can measure the position, it's kind of an additional plus without much investment needed. And that's the biggest benefit. All these technologies like RSSI, Wi-Fi, ZigBee, they're designed for data transmission, not

3:51 for positioning. But yes, if you are able to measure our exercise or radio signal strength indicator or radio signal strength by measuring that, you can estimate. You know, not with high precision, but nevertheless, you can estimate the distance. So that's the huge assumption of exercise methods. There are many. I highlight some of them. The angulation again—the trilateration is calculating the position based on zero to three distances, and there crossing of those distances. So several circles are crossed in one point, and you see, okay, this is my location point. Degranulation is the same, but crossing of lines—so angles instead of distances. For trilateration, triangulation is great, sometimes easy to implement, but usually is not as precise as trilateration for

4:49 large distances because, for the end goal, usually you don't have such great precision of the angle as the distance. Mixed—I'm personally a great fan of all kind of fused methods and mixed like RSSI plus angulation. Ferrero is already an attempt to mix two things: RSSI-like exercise plus the direction. It gives you more precision or more data capabilities in general. But of course, at the expense. What is the expense? Again, it's more expensive and more complex. Odometry—great thing. Many TVs, many robots are using. Why not? Of course, it is prone to collection of errors, but for the short term, in a short distance, it is one of the greatest and very inexpensive options, particularly for

5:46 sensor fusion-based systems. Actually, they're inexpensive. I am using inertial measurement units made it possible, but pure inertial systems—virtually impossible. I'm not touching expensive laser-based and all those kind of interferometer-based different applications. No, I'm talking about robotics, drones, forklifts, people—again, industrial applications. We're using basic MEMS micro-mechanical IMU-based and purely inertial. IMU-based systems are not possible to make optical. Many, I'm covering only some of them, but in general, they're many. And the greatest benefit of optical is that they are becoming more and more capable with low investments, because the cameras are very cheap. The software in volume is

6:45 very cheap, and you can do great things with optical. Very promising. And there are several flavors of this that will be covered: QR codes, stargazers. So when you really, you know, the camera watches up and sees some peculiarities, optical flow. When the system is in peculiarities, but on the bottom, usually, and motion capture, which is mostly used in virtual reality—I'm not talking about cinema, but virtual reality, augmented reality, everything. OK, my favorite and all this recommending: sensor fusion. Sensor fusion—there are many, many, many options, because basically you can fuse anything with anything. And for your particular case, you may choose the best performance: IMU plus ultrasonic, or IMU plus odometer plus indoor GPS by Marvelmind. Many, many. So it's always depending on your

7:43 particular case. And there are many other—I call them other types and exotic then notice legs. Otic. So don't be offended. But many other life, I RFID, magnetic, magnetic tape, wires—many, many, many, many other. So they're depending on particular case, use case. So even RFID, which doesn't—everybody doesn't give you a location, but if you go through the gate and RFID marks that you went through the gate, it is effective on the location. Yes, it's very imprecise, but in many cases, this is sufficient. So industrial applications. OK, some, you know, high-level statement: no method or else is good for even Marvelmind is not as perfect, certainly. So it means that there are many requirements, and those requirements are usually contradicting: price, update rate, power consumption, size,

8:42 you know, line of sight, non-line of sight, external interference. Not many, many. I just list some of them. Update rate, outdoor, nests, IP67 or explosion protectors, price, Parkinson, wait, size, interferences, location, allocation, plus direction. GPS doesn't give you direction. You don't know where you are facing. And for robots or anything, gotten, almost, this is one of the most important element: where shall I drive, where shall I fly? You don't know you need to move. So it's not really to identify the direction. Notice that magnetometers are not available indoors. If I forgot to mention it, it's even deeper: we do not recommend to use magnetometers indoors ever. Sooner or later, you will face the issues, and as we all, you know, collapse all your thing. So use something like Marvelmind implementing a spare deacons. So you have two deacons on some base—20, 40, 60 centimeters base is

9:42 Already enough from our own mind to have very precise. If you decrease direction, use it. Don't use magnetometers, and unless you're really, really, really forced to—unless you have calibrations, unless you are mostly outdoor but not indoors—as soon as you move indoors a bit, ferromagnetic or electrical current or something, that's it. You don't have direction. And then sometimes you need not only positioning for your indoor positioning system, but also data flowing in and out from the system. So this is critical, because if you don't have it in your system, then you may have to deploy another communication system. There are many people asking, can I combine LoRa with you? Yeah, but why would you? Because we do have our own wireless, so you can use it for slow—

10:39 Speed data moving in and out. My rewind because you don't need LoRa. You don't need to invest in LoRa, LoRaWAN, or anything like that on top of Marvelmind. Some system diff supported by default, like Ultra-Wideband, for example. There's also data—even high-speed data—but some systems they don't have it. Now let's start with our RSSI systems. They are very popular. There are many, but all of them have more or less the same limitation. And the limitation stems from the fact that our RSSI systems are not designed for positioning. They are data transmission systems, and they make a huge assumption that if signal of the radio signal strength is, for example, minus 70 DBM—

11:35 Then my distance would be—and you can see, ideally it would be around 18 meters in this particular case. Oh sorry, 11 meters in this particular case. But those dots are examples of real measurements. With minus 70, minus 70 DBM, you measure strength of the radio signal, and then you are guessing the distance. It could be in some measurement 2 meters, in some other measurements 5 meters, in some other measurements 8 meters. Okay, if I measure minus 70, where am I? 2 meters, 5 meters, or 8 meters? You see, the range is from 2 meters to 8 meters. It's very, very, very difficult to build an RSSI-based system. So RSSI-based systems, to actually calculate—of course they do, but they do so mostly because of multiple beacons and very, very complex algorithms that somehow extract something out of this complete—

12:35 Mass of data from a single beacon estimate where could it be. And still, of course, the precision that you are getting from our RSSI-based system is not high. Now, to say at least—it's for Wi-Fi 2 to 5 meters. Also, sorry, for LoRa, 2 to 5 meters. For Wi-Fi, 5 to 10 meters, depending on whom you listen to. But of course, the greatest advantage of our RSSI-based system is if you have LoRa on your person, like your phone, you don't need to give the person anything. It's kind of free cheese. You don't need to install anything on the person. You can track something that the person already has, the phone, which is great for tracking people in the airport, in shopping malls, somewhere—

13:34 Basically people with their phone. If we're talking about industrial applications, the benefits are slightly degrading, because you need to install something on their forklift or something on the robot in order to track it. Not high expense, but still, it's not free cheese anymore. You need to install something. And of course, the greatest disadvantage is that the system is very sensitive. It means that inside the building, you have an electromagnetic field, and this electromagnetic field strongly depends on the environment—metal, anything that affects the electromagnetic field. And let's imagine you are inside the warehouse and nothing is changing between you and the beacons that are used currently by your system, by your mobile beacon or by your phone, in order to localize. Nothing in between is changing, but a forklift, a huge piece of metal, is moving somewhere aside. You know, on the right, five meters from you. RSSI at your point will be changing because RSSI at your point is not interference but a combination, let's say, of all fields from all the beacons—direct light, reflected light, or all ways. And it means that your location will be distorted by something which is not even—

14:31 In between you and the beacons which are handling you. And if there is something which is in between, of course, the strength of the signal may change drastically. And of course, it requires a lot of additional software. It requires a lot of additional capabilities, like automatic calibration and automatic, let's say, population of data from each of the beacons about all beacons, in order to share the data with all about beacons when the strength of the signal in a particular point is changing. So a lot, a lot, a lot of complex algorithms just to make the system work. It works, and there are many, many, many people who are selling LoRa. Absolutely great, wonderful for some applications—particularly non-industrial applications where not too much metal. Again, backwards shopping malls, museums—perfect. I would say nearly perfect, because you don't need high—

15:29 Precision. Two, three, five meters—what is it? Able? Okay, good enough. But not for industrial. And of course, our RSSI systems, like LoRa, they have several flavors to increase precision. And one of this is angle of arrival. Basically, instead of one antenna, you have multiple antennas that you use for estimation not only the estimate of the distance, which is very, very, very imprecise, but also the angle. And it improves significantly—up to three times, according to those guys, up to three times—you can improve the precision, which is great again. Great. But at the expense of complexity and, you know, having those beacons more expensive. IMU, IMU is obviously—

16:27 The holy grail. And you would want to have pure positioning based on IMU. That would be great if not for drift. Unfortunately, there's always drift. And by definition, there is drift. And the drift is such that you may not have, without additional things like sensor fusion, a purely IMU-based system that works for industrial applications. Anyhow, you know, plausible or anyhow working in real life. So I'm bringing some examples of sensor fusion that makes IMU they start working somehow usable. You can go to YouTube and see these wonderful things. Perfect. I love them. They are really, really, really impressive. But don't be fooled. These things work only when you put this on food. If you put these small—

17:24 Things on your shoulder, pocket or pocket or somewhere, it will not work, because it will not be able to detect the point when there's no movement. Because you must also deal with the drift of the IMU. At some point in time, and this system is detecting this by detecting the time when the foot is steady on the surface. At this moment, the system knows that the speed of the foot cannot be not equal to zero, and it basically cancels the drift every time when you step. But if you stay at one point, then the system will not be able to work, and the system will start drifting. If you are not working, but let's say using this mobile beacon—

18:23 things on your shoulder pocket or pocket or somewhere it will not work because it will not be able to detect the point when there's no movement because you must also the drift of the IMU at some point of time and this system is detecting this by detecting the time when the foot is steady on the surface at this moment the system knows that the speed of the food cannot be not equal to zero and it basically cancels the drift every time when you step but if you stay at one point then the system will not be able to work and the system will start drifting if you are not working but let's say using this mobile beacon

19:23 Whatever this device, if not on your food, then the system will not work. So this is lovely, great for video, but not really great for real industrial applications or any kind of applications which are not exactly like this—lifting, lifting, lifting. Marvelmind is using IMU fusion or Non-Inverse Architecture. Currently we don't yet support IA (Inverse Architecture). Sooner or later we will, but in Non-Inverse Architecture we do have it. So this is real video showing the screen capture of the real video showing the performance, how it works. The red dots of ultrasonic—ultrasonic is great because it gives you absolute position. What ultrasonic may have is jumps, obviously interference, non-line of sight situation, something like this. IMU, on the other

20:22 hand, it's also great for very short periods. IMU latency is actually a fraction of a second—one second, two seconds at most. We recommend not more than a quarter of a second without collaboration. With great collaboration, rather, you can increase it, kind of significantly increase, but still it's a few seconds. You cannot have an IMU-based system for minutes without canceling the drift because it will drift away far, far, far. So in our case, ultrasonic is linked to the system absolute location every quarter of a second—so four times a second. And IMU, we are using double integration of the accelerometer data. Okay, accelerometer, gyroscope, plus everything. So it's pretty, pretty complex. It's very complex, but we analyze it for a quarter of a second

21:21 or half of a second. IMU is good enough in order to stay within two centimeters. And after that, ultrasonic is coming and correcting that drift. And by combining all this data, you can get the best from both. We can get very, very high update rate from IMU—100 Hertz in our case, but it could be 200 Hertz based on more capable processors. So 100 Hertz update rate, latency of around 12–15 milliseconds. And at the same time, cancellation of the drift using ultrasonic every quarter of a second—so four times a second. So IMU-based RTLS—this is another system as part of fusion. Is great. IMU-based RTLS without any external system like UWB or like in case of Marvelmind, IMU plus ultrasonic. No pure IMU or real industry application will not work.

22:19 Trilateration, obviously. Since Marvelmind indoor GPS is based on this, this is the most favorite and the most basic and I would say the most robust, simply because the system is designed for positioning. Already mentioned: trilateration is not equal to triangulation. It's not a big deal, but trilateration is cross-section or circles from known points. In this case, Marvelmind beacons—it could be GPS. The same story, you know, the distance based on time-of-flight of something, a radio wave or like ultrasound. Cross-section would be a point, but it

23:19 never crosses at one point. So you always have some spot instead of a point. Again, triangulation could be by Loran, angulation, or whatever. Quad, whatever thing. But trilateration is the most known term. But all this is—is a cross-section is a crossing point between three or two or more circles. In case of GPS, radio waves, in this band, and ultra-wideband, their different flavors. So some guys are using all, and some guys, you know, shorter, half a part of this band. Divide the band, use the higher precision you may get. In case of Marvelmind, we developed our own system. And the

24:14 system consists of ultrasound and radio. So Marvelmind is not an ultrasound navigation system. It is not a radio-based system. It's ultrasound plus radio. So we have a very precise clock synchronization between all the beacons using radio, and very, very precise distance measurement between the mobile beacons and stationary beacons. So we belong to trilateration-type systems. Precise timing and synchronization, as mentioned, in case of Marvelmind is solved by continuous synchronization between all the elements and the control of the model. Yes, the same is done using atomic clocks. Okay, our beacons are significantly less expensive than via satellites, but we could theoretically install instead

25:14 of—or use the system instead of continuous synchronization. We could install like rubidium clocks, or whatever atomic clocks. Why it's not done? Because you always measure with some precision. You never measure exactly, exactly, to the last digit. So now is interference, abstractions. In order to solve it, redundancy is the best option. Like in our case, three stationary beacons is enough to calculate the position in 3D. We recommend three plus one. So if one of their beacons is blocked, the remaining three will still serve you. You can use n plus one or n plus one redundancy. We can use any redundant. Effectively, have two full overlapping submaps of the same area. And the system then will choose between

26:12 their data from one submap or another submap. It's really complex logic, but in majority of cases it is able to recover the data and give you the best information, especially in mobile mode. When the mode is stationary, it's a bit more complex because sometimes it's nearly impossible, even in terms of physics, to recover if you hide somewhere behind the wall or behind somewhere. It's really impossible, even in physics, to find out where you are. And this is another my favorite topic because, of course, we are all the time compared to UWB—Ultra-Wideband. And ultra-wideband is absolutely great technology. Even though we are compared and we consider ultra-wideband as competition, but never realize we totally respect to

27:10 the right band. And frankly speaking, if ultra-wideband were available some years ago when we were developing our own system, we wouldn't do that because we needed it for robotics. It was not available. So we had to develop our own system, which eventually became more precise than ultra-wideband. But of course, each system has a limitation. And one of the strongest limitations of our system is a requirement for a line of sight or line of hearing. But let's go a bit deeper into this. And my point is that all precise, real-time location system RTLS must have line of sight. Now let's check: if you are inside a wooden house, radio wave from GPS satellites may go through wood. Is it line of sight? Now, okay, it's not line of sight for you, but it is line of sight

28:09 for radio wave because it can go through not-too-thick wood. The same story: when people asking about ultra-wideband—is it non-line of sight system or line of sight system? No, ultra-wideband is also. When it's precise, it's also a line of sight system. Simply, their obstacle in front of ultra-wideband wave is a radio-transparent, for those waves. But if you go to the real industrial applications—not the office way I'm sitting right now with radio-transparent walls—but the real industrial applications, thick concrete walls, thick brick walls, metal all around. Obviously, radio wave cannot go through those obstacles either without complete blocking or

29:09 without huge distortion. So complete blocking is obvious—if the signal cannot go through, then it cannot go through. There is nothing to discuss. But let's discuss a bit the intermediate case when the radio wave is going through. First, but second, you need to have high precision, like our case: centimeter-level precision. Imagine beyond your wall is 40 centimeters thick, which is very easy to be—it could be even thicker, but let's imagine 40 centimeters thick. Assumption: for any time-of-flight based systems like ultrawideband, GPS, or Marvelmind indoor GPS is

30:05 that you do know speed of light. In case of GPS and ultrawideband, it is speed of light of radio wave. In case of Marvelmind, it is speed of light of sound. Yes, we know speed of light and Einstein showed that this is—this is well-known figure. But where is it a well-known figure? In vacuum, it is a well-known figure in air, which is very, very close to vacuum. So don't pay attention to the difference. But if you go to material which is significantly different from air or vacuum, then you have electrical—like a term, EMU or whatever. Now you have these two parameters: magnetic and

31:03 electrical properties of your wall. Let's put in this well. I will not go to terminology, and those are significantly different from air, significantly, which means that the speed of light inside your wall, it's not anymore 3 × 10^8 or whatever, 300 million meters per second. No, it's completely different and usually much smaller. So it means that inside your 40-centimeters wall, your effective length would be not 40, but unknown—you, unknown to you, a multiple point of view. Now, first of all, you don't know the properties. Second, you usually don't even know the thickness because again, your system doesn't have this data. If it has, you can try to

32:00 combine it, but it will hugely place the system. It's virtually impossible, and I don't know anyone using this. An active second: when it goes, as you see it, it will always be refracted, so the direction of the wave will change. Third: wave will be open. So you have one nice falling wave. As the result, you will have one wave coming through, another wave reflecting and also coming through, but with some delay. So instead of a nice pulse, you will have a train of pulses with different amplitudes, with different delays, unknown to you. What is it leading to? It is leading that you cannot actually take into account all this data and to make a precise estimate. You somehow

32:59 magically made a radio system capable of one-centimeter-level precision. But when it goes through the wall, you have 40 centimeters of physical uncertainty, which is not—yes, it could be in terms of radio wave propagation, 40 centimeters, 50, 60 centimeters. It could be anything. And these 40 centimeters of radio wave inside this is not 40 centimeters in the air or in vacuum. It's longer. So it means that you have some tens of centimeters of uncertainty. Even if the signal coming outside you are able to detect, in many cases you are not, because ultrawideband or GPS usually have a very, very big signal in order not to distort or disturb other systems, not to interfere with other systems. So you usually don't

33:58 have their generosity, and the strength of the signal is not to be able to detect after it is a tornado coming up. Line of sight is the must for size industrial real-time location systems. And we always recommend not to try to build non-line of sight unless you absolutely know that it will be, again, a transparent, and unless you absolutely know it will be a simple thing: radio-transparent walls, like in many offices. But you go to industrial warehouse pallets, metal shelves, metal objects, large objects. They create shadows or completely non-transparent objects for the signal, and that's it. It's done. So this is why in our case, even basic wood or glass would be non-transparent. But we are not even trying to make

34:58 non-line-of-sight systems, and we do not recommend even radio-based systems to be non-line of sight. Our recommendation: if you want robust industrial application, always use line-of-sight topology. Okay, but what to do in real-life environments because non-line of sight is everywhere? Shelves, objects, people, doors, glasses, metal. Okay, there is no simple solution. And the solution is basically a combination of different solutions. Now, first of all: proper network planning and beacon placement. This is why we usually recommend to place beacons high on the wall or high on the ceiling. And the mobile beacon must always be placed high on your forklift, high on the person, high on the drone. Why? It's very basic: because the

35:57 chances that you will have obstruction between the stationary beacon and the mobile will be their smallest. It will be the least. Second: redundancy. Use submaps, use more stationary beacons. So don't make the system non-redundant. If one beacon out of three is blocked, that's it. You will not have 3D. You will have broken 3D. And our recommendation is, of course, use N+1 or M plus M or M+1. No, planar 3D models is the same. Basically, use redundancy, use of trilateration, proper network planning, use proper placement of the beacons. That's there the strongest, the least expensive option, and it's always depending on your requirement. If you have large open area like basketball

36:56 court, that's it. I mean, it's easy to do. You place beacons at the top. We give each player a Mini-RX. In our case, put beacons on their shoulder, that's it. It works. But if you go to the warehouse, there are many shelves. Don't try to go the signal through the shelves. You need to cover each aisle separately with submaps on each aisle. This is why, when we are getting RFIs and we are asked, "Okay, how much would it be? Or how many beacons do I need to install for this?" we always ask for their floor plan. And if you have photos, just understand: because warehouses are different. Some warehouses don't have shelves. Then you need signal differently, fewer beacons to cover. But the majority they do have shelves. Okay, but then what are we tracking? 2D, 3D, drones, people? So we need all this information. And I would say any system

37:56 would need information before even a budgetary calculation is given. Another option: if you are not able to make line of sight, a line of hearing, like in our case, use sensor fusion, which is perfect until you recover. Because, for example, you have a shadow from something, but this shadow is small—a column or some minor object—so it creates a shadow of, whatever, one meter. Okay, fine. Use odometry for one meter. Odometry is perfect. It doesn't yet accumulate error too much. So odometry is driving you. And then when your forklift or the robot is out of the shadow, you use sensor fusion: odometry plus indoor GPS. You recover and cancel the error that odometry

38:54 accumulated and you are fine to go. Once again, sensor fusion is always the best option. It could be many—it always depends on your particular application. You can use odometer for robot C, but you cannot use a diameter for drones. But you can use some obstacle for drones sometimes. Optical malfunction, okay, using their GPS. So combination sensor fusion is the best, the best option. Automated passenger GPS. I am U plus ignore GPS the world. Sometimes you can tolerate reduction. It must be additional feature we provide. The raw data to you—not only location material data—and you can get our location data quality signal that you are getting from our data and lower distances, and you can detect that, okay? It was continuous, continuous, continuous reduction of some distance, then continuous increase of another distance, and then some sudden jump on the third beacon, okay? You detect this, and in your external algorithm you say, "Okay, I will"

39:54 be using X and Y. I'll be not zette for the moment because I cannot trust that anymore. It jumps. My model doesn't assume that my, my, my whatever for Cle jumps up. It's physically impossible. I will fix and freeze the set for the moment. It I know that it will give me some error in that, and which could be translated in error in XY. But if you do it smartly, the error is usually not so great. So for temporarily reduction from 3D to 2D, you can tolerate obstruction of one of the beacons. So it's one of the options. In some cases you say, "Okay, I will tolerate a lack of taking now, whatever toilet. You don't want to take any one in the toilet. Fine, there's no one else or there's no another way." So you track in your corridor, then the

40:52 person disappears, okay? For one minute, then the person is back. You are fine. So once again, it's always about your particular application. Line-of-sight is a must. So the best option is making line of sight. Is not possible, then you can increase slightly the complexity and recover. But in some cases you say, "Okay, for me it's not economically viable to fill in and cover, and I'm okay to have some black spots." Okay, it's economical—not only technical decision, but economical path. Technical ultra-wideband. Now again, we are not in the position to comment Martian ultra-wideband. The first of all, we are not as familiar with ultra-wideband as with our lovely indoor GPS. Second, of course, with the red band is one of the nearest competitors. So again, for me it would be not in the right position. But point is that there are many flavor so filter wideband, and let me once again stress ultra event is a

41:49 great technology, great technology. What you need to remember is ultra-wideband. About ultra-wideband is the following: first of all, they certainly do not recommend non-line of sight for industrial application because sooner or later you will heat the problem. Second, ultra read that tags are usually very power efficient, and we are reading like whatever half a year and one month, three years on the one battery. That's perfect. But at the same time, the station, the beacons, or base stations, whatever depending on the manufacturer, usually is very time-consuming, also very power consuming. So they do not run on the batteries and base. Now experience, very many customers particularly interested in battery power supply for the station beacons, so

42:46 sometimes provide electricity to the beacon is more costly than the beacon itself. So this is why Billy is great, and that it's also battery powered on the station and beacons. My real mind is kind of in the middle. We usually recommend to run their fixed power supply for everything because it gives you peace of mind. But yes, we can quite easily get weeks for embedded battery and for the external battery—one year, two year there, okay? Fine. When Astor complications, simply you have large battery next to you. It works with us. It works with Billy. It certainly doesn't work with ultimate them because usually the base stations will turn red and require external power supply, pretty high power consumption. So just notice, be fooled with is everything to go, go, dip it to the details.

43:46 Lighters, ladders are perfect. Weathers a great. But lighters I designed for obstacle avoidance and detection. Lighters are not designed for positioning. Yes, I'm aware. Right, lighters have been used for AGVs for a long time, and they used for cars today, etc. But lighters are very, very complex. Lighters are very, very costly, and if you sacrifice in complexity and expense, then you have lighters with even less performance. So it means that your chances to be lost in terms of position even higher on the right. There are some truths we obviously have got from the internet. Sorry for not doing link. Yeah, but this would lighters sees in the messy environment of the real

44:44 warehouse or the real, real environment. It's pretty messy. So our short recommendation is: unless you are really not limited with my own money, unless you are not limited on size—because lighters are usually larger and more power consuming than other types of beacons—do not use lighters for positioning. Because sooner or later you will struggle. You will have the situation when lighters are prone to, let's say, misposition yourself. Lighters as the element of a sense of fusion system. Miss odometer with indoor GPS is something else. With external, let's say, visual or with optical, a great. They're good. But again, then matter of complexity and price and all those kind of things. Lighters are great for obstacle

45:41 avoidance and detection, not so great for positioning. My favorite, one of my favorites: robust and simple—like as simple as possible and as robust as possible. QR codes plus I am you pass odometer. That's a fusion. Great. How the system works? You have to our codes or any other types of visual things. They change told either on your floor usually or sometimes from the ceiling or things around. And you have Connor's inexpensive, and they are washing either one part code and it's of what they format or many, many, many other copycats available in the market on the floor. And between those are codes which are inside the warehouse every one meter or every two meters or something. The robot is moving based on odometer.

46:40 Odometer is perfect. It accumulates error, but usually not sufficient in order to be lost over the one or two distance—one or two meters distance between the car codes. And when it comes to the QR code, it corrects the position such way that error of odometer is not accumulated. Robust, perfect sensor fusion system, very reliable, very great. Recommended. Obviously recommended. There obviously some other issues with it. My QR codes can be broke, and we're now too lost, partially covered. But in general approach is great. Robust, very inexpensive, and recommended. And this why many people are using it. Good. Visual, visual is also great and coming not without issues. Like any other system, but in many, many applications it's a very

47:38 good solution. There are several flavors with visual stargazing and landmark navigation. Basically, the camera something which is peculiar around like light, like dots, like star, and its positioning itself against it. Since this is an group type position, which means that if your star is far or if you object is far, and when you move by whatever half a meter, your end goal changed just a little, then you cannot have—you cannot have high precision. That's the biggest disadvantage of any triangulation or angle-based system, and visual position belongs to that because it doesn't usually have time of flight capabilities. The newest cameras with external light, they have it. So it may coming, but here I'm talking about pure

48:37 Inexpensive cameras that are able to calculate the angle to the object. By having either multiple objects with different positions, you calculate the angle and then you calculate the cost. So this is your position. You have multiple cameras on you, and they calculate the angle to their one known dot. By knowing the distance, you can calculate the relative position. In one way or another, it depends on their other object, it depends on the distances to other objects. Of course, it's prone to lighting, to gain, to dirty cameras—many, many, many things. But still, it's a great solution because it's inexpensive and it's progressing very fast because chips on the drones, on the robots—they're capable to crunch a huge amount of data. And what is—

49:36 Ten years ago was a very, very difficult task. Today, it's done by very expensive drones. QR codes are very robust, so it basically helps the system not to guess but to detect. QR codes are your known dots, which tell you not only that this is a dot, but what kind of dot. Okay, this is a dot in this direction, and you can hide much information. Let's say, embed much information in a QR code. So by combination of QR code and visual position, you can do many, many nice things inside and outside. Now, as you see, this is the difference. You may have a camera or several cameras on your object like a robot or like your VR camera, and nothing outside the camera itself. The camera detects the external—

50:35 World and calculates the position against that external world, particularly if the camera knows the map and land. It needs to know outside, in. Different. So you don't have anything on your person, on your robot. It's very similar to motion capture. So a motion capture system—you don't have any cameras on your person. You simply have reflective markers on your person, and you have many, many, many cameras around. Usually, those cameras are expensive. If you think about motion capture, because the requirements are so very low latency. A regular camera gives you 30 frames per second, 60 frames per second. Motion capture gives you 200, 400 frames per second. So obviously, they're not standard cameras. Second, you need to install those—

51:34 Cameras pretty densely, like every 2 meters, every 3 meters, if you want centimeter or even sometimes millimeter-level precision. Because if you have cameras 20 meters aside, they will not be able to detect your end, or say, angle to you the sufficient precision. This way, in case of RTLS, let's say time-based, you can have very precise tracking even on high distances. But in case of angle-based or triangulation, you really need to place those cameras very closely and densely. And if they are cheap, that's great. If they're very expensive—sorry to say—optical flow, sorry, motion capture systems are usually very, very expensive, running sometimes to tens of thousands for a water or tens of—

52:29 Thousands of dollars per unit. Now, I already mentioned cinema, and we are using motion capture because it's proven technology, high precision technology, but costly. Quality depends on the lighting, depends on the distance, depends on dirtiness, and many, many other things. But some visual systems are great for indoor applications and indoor applications and can be recommended. What is not a trade location update rate is one of the elements of the system precision, price, size—yes. But location update rate. If you have one drone flying for inspection, it's one story. But if you have hundreds of workers inside your warehouse or sample plant, is completed—sorry. You may want to track all of them at the same time. High update rate can be a very, very challenging task for—

53:26 Some types of indoor positioning systems. It could be done as easily for 400 people as for one person, or for another system. So it really depends on that gate. Actually, it really depends on the task. In case of Marvelmind, we know about this. So this is why we have two architectures: Non-Inverse Architecture and Inverse Architecture. In Non-Inverse Architecture, the beacon on your forklift or on your drone is emitting ultrasound. And this is why the system is very robust, and they're used with drones, which are noisy. But at the same time, only one beacon can emit at the same time. So it means that if one drone—that's okay. If you have two drones, then the update rate per drone would be the update rate per system divided by two, by the number of drones. If you have one, two, five, it's still okay because we will have still 16, 8, 4, or—

54:25 Whatever number of Hertz, which is okay. But if you have hundreds, or let's say 100 drones flying, then Non-Inverse Architecture will not help you because the update rate per drone will be just low. We have Inverse Architecture, which is great because Inverse Architecture works like this: stationary beacons are making ultrasound, and the mobile beacon receives the ultrasound. This way, it's all of them receiving at the same time from the same stationary beacons. They can receive 200 mobile beacons easily without update rate reduction. But since those beacons are receiving ultrasound, you cannot place them on noisy objects. In my own mind—I'm talking about other systems now. So in case of Marvelmind, because the mobile beacon will be receiving ultrasound, which is great for forklifts, very, very—

55:24 Great for people, perfect for robots. But not so perfect for drones, which are noisy—not only in audible noise but in ultrasound noise. So it means that the noise from the propellers of the drone will be blocking the signal, and the range will be greatly reduced. Not 30 meters like we usually say, not even twenty, ten—maybe it can be as low as 5 meters or even less. So it's very much dependent on the drone. It very much depends on the relative position. Many, many things. To make it short, currently we do not recommend Inverse Architecture for drones. We recommend it for people, mostly for robots, for forklifts. Power supply. Power supply is important. Sometimes it's as important as other parameters like price. Because battery—

56:24 Battery life time—how long? Let's go a bit deeper. How long? What station beacon? Mobile beacon? Is it replaceable, not replaceable, chargeable? Is it normal conditions? Is it external temperature range? Many, many other parameters required. So our recommendation is ask questions. Ask questions. Ask, ask questions. Other guys—it totally—once again—depends on your requirement. If you are able to provide power supply for stationary beacons, that's perfect. Peace of mind. You don't need to worry about that. Then you need to worry only about your mobile beacons. Then, what is your mobile beacon? If you have a forklift—okay, no problem. Again, you basically supply your mobile beacon from the electrical circuitry of your forklift. That's it. You are done. No issues with the battery at all. But sometimes you cannot have your stationary—

57:21 Beacons on external electricity. For example, you deployed their navigation system or positioning system for your mobile team fixing something under the bridge. There is no electricity. You need battery. And then it's important. Then you need to ask questions like: What's the update rate? How many mobile objects? What are the temperatures? What's the battery lifetime? Better lifetime with the update rate—it's always coming. Because, for example, when you see a five-year battery lifetime, what are the conditions? What is the update rate? What's the range? What are the temperatures around? And also, is it replaceable battery or not replaceable? Sometimes, you know, beacons or texts—they don't even have an option to replace the battery. In our case, it's charging. But sometimes you cannot even replace the battery. You need to throw—

58:19 Away, the complete tag also station versus mobile. So it also depends on the application. Sometimes it's very easy to charge them about because, for example, you have a shift and you want to charge every shift, and there's no problem. So because there is charging for your light or charging for vehicles, but sometimes it's an issue. And the same with stationary beacons—sometimes, okay, there in the world there is electricity, no problem to cover. But if you go outside, it may be an issue. Or you don't have the permit or terminals are expensive, or something is more expensive than the battery, then the battery is an option. We do provide all the options: embedded battery, rechargeable—it's always rechargeable. In case of Marvelmind, external power bank, external battery, or external power.

59:17 Converter. All the options are available, so just choose and ask. Also, technologies are not equal in this regard. Ultrasonic is great in terms of stationary as mobile. It's more power-consuming, but also great, so it provides very long-time battery. Ultra-wideband is perfect for mobile beacons but terrible for stationary. You do need for ultra-wideband power supply for station beacons, at least as far as we know based on information available on the internet. So UWB is somewhere in the middle between ultrasonic and Bluetooth in this regard. So yes, you may have one year or two years, better a lifetime for both mobile and stationary beacons, depending on the battery size. Larger battery—why not? Industrial applications, usually you are not very limited with the size of station.

60:15 Beacons. Installed there, it's there. You know, once a year you come, charge it for one year. Also, sorry, for one day, and then another year is there. That's it. And between this you use the embedded battery, which is inside. It can last for days to weeks, depending on the update rate. In one day you return back, you put back the battery, and it works there for another year. And then, of course, you need to calculate what is less expensive for you: have external battery and maintenance once a year, have external battery and maintenance every two years, or have external power supply and nobody else at all. So it's all about money and your particular case, location, and direction. Now, as mentioned in the beginning, we do not recommend to use magnetometer or compass inside. Soon or later you will.

61:15 Have a huge issue with this, simply because there are a lot of ferromagnetic and magnetic materials indoors, and it will always distort—again, sooner or later—distort your magnetic compass and magnetic readings, and you will not be able to know the direction. The recommended solution is you have mobile beacons on your object, your shelf, your reality, robot, drone—it doesn't matter. The sufficient base, and then you know they're facing there, the angle. What is the vision base? It depends on the technology. If you have really precise navigation system like Marvelmind, then you have centimeters precision. Then, what is two-centimeter precision? It means that this dot is not its spot with two-centimeters radius, and this is spot with two-centimeter radius. If you have.

62:12 Whatever twenty centimeters base, then your worst-case scenario is that your spot is here and the nut is here. So this is your angle. And another worst case is here and here. This is your angle, and this is the difference between these two lines would be your error. Okay? In case of ultra-wideband, that wouldn't be possible with this, is no—simply because ultra-wideband is not as precise, and you have not two to sixty meters but ten to thirty centimeters. But it will travel bent as well. Simply you would need to have larger base, like one-meter base, for example, is perfect for that. You have also sixty-meter level precision, but unfortunately, it doesn't work indoors. Okay, there is video. Let's not watch.

63:08 Video for now where we are, because people asking where we are, where we are. No, we aren't of this pyramid in terms of precision, because there are many, many, many, many providers. Publication Wi-Fi—why? Oh, it's really easy to do because you can get ready-to-use hardware models. You can download the software, or, let's say, use the low-level software, and then you need to bring on the application. This word—why?—there are hundreds of years, but use Wi-Fi? We don't go there, no, because it's well covered. Ultra-wideband, much, much, much better technology in terms of precision and capability to the industrial applications. Still relatively low barriers because you can buy.

64:06 Ready-to-use chipsets now. Lovely takeaway from some other players. And then, of course, you are using the same approach with getting the low-level software from the same chipset provider. Your task is it's more complex—obviously more complex than Wi-Fi positioning—but still you get a very usable system relatively easily. And then it's all about higher-level applications. It's all about higher-level features, like again your product, final product, like protection against moisture, like price, like battery, like many, many other. And in the case of UWB, in the case of Marvelmind, we were not able to afford at the time—obviously, this precision was not enough to us, and ultra-wideband was not available years back when we developed. We developed our own architectures and our own protocols.

65:04 Our own hardware or software—low level, high level, and everything. So this way we were able to achieve such great precision with such a low price. Now, Marvelmind. A few words—I will be speeding up because it's already very, very long. And you will see our system is off-the-shelf ready using the navigation system, which is based on ultrasonic beacons united by radio in license-free band. This is example based on Super-Beacons. You have four stationary beacons, one mobile beacon, and one modem. Give the stationary beacon, in case of—or in case of Super-Beacon—can be stationary, can be mobile. You simply click and use it. And there are the distance. The location is measured using ultrasonic pulses between the stationary beacon, mobile beacon, and modem. But let me stress again.

66:02 Our system is based on radio for very precise clock synchronization and data exchange and ultrasonic for very precise distance measurement. So it's combination of two technologies, not one, not on transonic, not radio, but two. Now, as mentioned earlier, our system is very inexpensive. So you can—you can compare. We are very proud of this: two-centimeter precision and still very, very affordable. Now, as the result, as you see, we have customers in many, many, many different countries and at the same time, absolutely different types of applications. What are they? All kind of robots, all kind of drones for inspection, and all kind of virtual reality and augmented reality. You know, customers are coming saying robots are great, everything is perfect, but can you do the same for real environment? With real environment, forklift carts, all kind of mobile objects—sure. You have told them about.

67:01 Beacon on your forklift, and you know everything about the forklift—path, average speed, running time, peeling, not speeding, broken acceleration—everything, everything is recorded, everything is stored, and ready for analytics. Some analytics is embedded inside, but we also provide you the access to this raw data, and you can build effectively billions of different options for the external analytics, for improved mobile asset utilization. The same people are coming to us and say, okay, great, but can you do the same with people? Sure. This is why this moment helmet: it has embedded navigation, it has IMU, and it has options to connect external things. What are those things? Emergency button, or vibro, all kind of.

67:59 Alarms. White alarms or any kind of alarms. Again, sensing, so it means that the person with a hard helmet entering the gear fencing area, there will be alarm for the person, it will be alarm for the boss, and everything is recorded. For what? For safety and productivity. What are the cases? Many, many, many different: underground, mining, metal, construction, manufacturing, dangerous oil, gas, refineries, agriculture everywhere where you want to increase productivity by knowing who is doing what, where, and why, and improve safety. If the end things happening is much, much better than, you know, solve those bad accidents. As mentioned, always we have two different architectures: Non-Inverse Architecture and Inverse Architecture, known in the circuit. Texture is very robust, and it was developed first, and none of us architecture works this way.

68:56 You have stationary beacons which you install on their walls or ceilings. Okay, in this case 25 meters, but it's for Mini-RX and okay, every 30 meters or close. Those beacons are controlled on the radio by the modem. So the modem is working and talking to each of the beacons over radio and synchronizing the beacons and controlling the beacons. So the modem is the smallest element physically, but logically is the most important element. It's responsible for size, clock synchronization. It's responsible for all controlling and distributing of the tasks between all elements. And it's possible because it's not only a router, but this modem to the external world. So none of their applications. I needed to run the systems.

69:54 You don't need it instantly. We do this initially. Don't need the cloud. You don't need Marvelmind servers, nothing like this, in order to run the system. If you want to be absolutely closed and not like any location data outside, you're absolutely free to go. You can do this without system because the modem with complete map and function without anything else. Modem is sufficient. You just power up the modem. The system is there. The modem knows the map, all the beacons know the locations, and you know the locations of all my bio beacons. Beacons know beacons know all of its location. And when the mobile beacon can know location of all beacons for collision avoidance, because you know position of all your robots or all robots know position of each other. The map and building of the map consists of elements. First, you need to provide location in case of Mini-RX.

70:54 Of each of the stationary beacons. The beacons hardware version 4.9, Super-Beacons, or Industrial Super-Beacons, they calculate the distances between each other by themselves. Let me stress: you don't need to calibrate anything. You physically put the beacons inside your building, they have line of sight. You put the system into the mode of building. There are some maps of beacons. And what is left for you to do is just to confirm: okay, I see that distance is all right. The system is helping you to estimate that, and you just confirm it and freeze the submap. So our Non-Inverse Architecture and then works also with many beacons. Support self-building the map. No calibration needed, no manual entering needed. Mini-RX is...

71:52 A limitation. In this case, it's small and it's not able to emit the sound, unlike beacon hardware version 4.9 or like Super-Beacon or unlike the ultra Super-Beacon. This way, you do need to enter the distance between them. Distances between them, all of them, or location the amount on the ball ceiling, and they're measuring the distance to them about beacons using ultrasonic pulses. Because the modem basically is controlling everyone, and everyone knows when the mobile beacon is emitting ultrasound. By measuring the time of flight, it's able to calculate their distance. By knowing the distance from this to this, from this to this, from this to this, the data is sent back to the modem, and the modem is calculating the position, and then it is distributed back to the mobile beacons. That's it. The system is proposed to be connected with many.

72:50 External devices, robots, drones, anything. We are using mobile or say virtual UART, regular UART, then pins, SPI, I2C. Its of industrial application. Additionally, there is RS-485 for industrial applications. You can get the data either from the badge beacon directly, to example for the robots or drones, or you can get the data from modem method. And the robot knows its current location. And also, the modem knows the location of any robot or drone or any person inside the system. Many of them. Badness. They have IMU on board. Sometimes accelerometer plus gyroscope, large compass. Sometimes only accelerometer and gyroscope. Everything is at V.D. Obviously, and you can get the location data up to 25 Hertz. But ultimately it depends on.

73:50 The distance. For larger maps, it's rather close to 8 Hertz. For mid-sized, 16. For very small maps, if you met meters, society, you can get up to 25 Hertz with IMU fusion. In Non-Inverse Architecture, you can have 100 Hertz with 15 milliseconds delay. In regular systems, the latency would be trade divided by their 1 divided by the rate. So if you have 8 Hertz, you can expect around 125, 150 milliseconds of latency, unless you use the averaging. Averaging gives you higher precision and high confidence and higher, let's say all more robustness to jams and all kind of interference, but it gives you latency. And it depends on what kind of hero templayer using or what kind of averaging, because they have many.

74:48 So you can increase the latency up to let's say one-half seconds. And instead of 2 centimeters, you may have less than one centimeter precision. So much better precision and much more robust, but for the expense of latency. So it's up to you. You choose. There is a trade-off. In the Inverse Architecture, it's significantly more complex in terms of internal structure, but for you as user, it's nearly the same. So the difference is that in this case, mobile beacon is emitting ultrasound, and stationary beacons are receiving. And in this case, about beacon is receiving ultrasound, and stationary beacon is transmitting. But as you see, stationary beacon starts meeting different frequencies: nineteen, twenty-five, thirty-one, thirty-seven, forty-five players. We have currently five frequencies. It means that if you want a basic submap or basic map consists of once a map, and you simply need to choose a couple 25.

75:46 31 or 25 and 19. It's a must for Inverse Architecture that stationary beacons have different ultrasonic frequencies. It's a must because otherwise the about beacon will not be able to distinguish ultrasonic signal coming from this, so from that. But if you have larger maps consisting of, let's say, tens of beacons or hundreds of beacons, obviously you need to repeat repeating sequences because we have only five. And I'm any method. And this is one of the complexities of Mini-RX. Currently, we do provide remote guiding guidance for indoor architecture deployment and the indoor architecture calculations. But for known and respected texture, you can do this easily because I have so-called placement manual. Go to downloads page and use a placement manual. But for another circuit, actually, recommend for.

76:45 Complex map, we do, we do for you the calculations. Otherwise, the cities are the same. Otherwise, the capacity of both in Russia collection another circuit X are the same. So currently we support 150 beacons, mobile and stationary combined, overall. Recommendation: Non-Inverse Architecture is great for something noisy like drones or for some reason you cannot have a noise near them about beacon. And University texture is perfectly suitable. First of all, for people tracking, because usually don't want something, you know, transonic ticking on you. It's okay for robot. It's.

77:44 Perfect for the wrong. It's okay for forklift, but usually people prefer that be sticking on them. So Inverse Architecture is your choice, and ultrasonic is far—five meters, 20 meters from, and nothing is sticking. Even though you don't hear usual ultrasound, but you can hear in a small thick formula showing beacons, especially if it's very quiet area like museum. And the next one: what a care-about until now, our usual cases—warehouses, robots, drones, people—but we have multiple, multiple, multiple cases for long tail. I will not go too deep into this, but those cases sometimes are very, very useful for very special applications. Like one of these days, so you want to warn users: the large.

78:44 ATV is approaching. In this case, you do not offer the full area. You do not provide coverage like for very, very long way of this large ATV. So GV itself is the map, so you install the station beacons on your large ATV. So effectively, the whole map is moving from a DD point of view. This is a station beacon because nothing is moving, but from the point of external world, how map is moving? So it's kind of moving station and beacon. So it's not extreme around, but this kind of thing. What are the benefits? The benefits is that you have your moving, and when a person is in dangerous area, the person will be alarmed—visual, vibration, sound, anything you wish, depending on the options. Every

79:40 time when something is closed then your area, so give testing area. So this is kind of moving your sensing area, and this could be up to 30 meters radius from your ATV or from anything moving. The same on the construction site: here you can come by all these, so you may have stationed a big, and you can define gear fencing over area. In this area, you can deploy a regular inking system on side. You can put my bio beacon, or let's say not, I'm about beacon, but the whole system on. You can, so it means that the crane will be moving like in the next one, and you'll

80:39 be moving, and the crane will provide. Oh, let's see, have the moving your dancing zone, not the crane, but will be moving. And where'd it is, they will be danger zone in for safety. So site like construction site is usually very complex area, and sometimes it requires multiple different approaches: a regular map, moving gear fencing zone, multiple floors. So tactus give us description, and we would be able to provide you with a solution as well as these options. It could be a regular beacon for very quick installation—a regular beacon with external power supply—or it could be a fixed beacons with sticks power supply.

81:34 If it's inside another special case, you don't need to worry about all the areas. You need to worry about areas under the crane or under the tip, under the cargo. And in this case, you place the beacons, or actually, stationed the beacons, and they built the map of station beacons right on the tip. And every time when the crane is moving, there, your dancing zone is moving along with the crane, and you provide the safety for the workers—not just general safety because it's not relevant, but you provide your gifting zone only where it's really dangerous. Another example, another example: you don't want

82:31 to cover with the beacons the full building, okay? Back to this. I was interrupted. For example, in the case when you don't want to cover the whole building with the system but because it may be too expensive for you, you may have what's covered—it's very special, but we can do this as well. For example, you want to track your cleaning workers so that they do perform the tasks. What does mean? So it means that if the worker must clean every 15 minutes, I can easily detect with the system that the person was at least inside this building now, inside the spot where the cleaning must be done in 50 miles, not like the person comes at the end of the working day marks everything—okay, I was there, there, there, there—but the system is obviously taking to

83:31 the person who was not there during this time. Not tech with our system that the person actually did the job, but at least you can be sure that the person spends whatever 5-10 minutes, and you hope that it was cleaning during this time. There are many options like this, and it means that based on our system, you would be able to deploy not only a regular precise and allegation system with blanket coverage, but some special cases. Well, today we support a single Modem system covering large distances up to a few hundred meters, and this distance is defined by the radio. So the beacons must not be farther the 30 meters, so your mobile beacon must be within three

84:28 meters from two or more station beacons. But in terms of radio, the coverage is much, much larger. With full-size antennas, it's a few hundred meters—now, let's say, four hundred meters—which means that from the Modem, and you have a tunnel for instance, a large assembly plant, you stole the Modem in the middle with full-size antenna. You can have one hundred meters this direction and four hundred meters this direction. I think, for example, this Industrial Super-Beacon with full-size antenna. But what to do if there is no coverage or there is no direct visibility, even for radio, and the distance is more like long-term tunnels? Okay, in this case, we will be supporting—not yet, but it's coming pretty soon—already so-called multi-Modem systems. Multi-Modem system is basically the same single Modem system but with one layer additionally organic. Currently, we have

85:27 Modem, which is responsible for the map. Map is consisting of a maps. It's a map is either 1 or 2 or 3 or up to 4 beacons, and all together, you will have 250 beacons per submap or map, or 250 the maps per map—that's what we currently have. What is coming: this multi-Modem system is effectively the same, but you will be able to have up to 20 Super-Modems, which will be under control of so-called Super-Super-Modem. The system will be able, incoming—let's say, 2-3 months—would be able to support a few thousands of details, and it will support their applications when the radio coverage from Super-Modem, the beacons, and not be provided. Another

86:26 floor or larger distance, or for some other reasons, so in this case, there will be another Super-Beacon. Those Super-Beacons, they connect to each other, not for synchronization—where we need our own very, very precise clock synchronization over our own radio—but for date exchange and location data streaming. Soft milliseconds is usually not an issue. This way, regular Wi-Fi is OK, a net is OK, LTE is OK, and private LTE is at a, but this is for data, not for synchronization. And a Super-Modem, and then the Super-Modem is basically a single point where you look at all this map as a one large map. Don't see all this complexity for you. It's just one, one, one, one large map which is together.

87:23 By the Super-Beacon, Super-Modem already mentioned, we have a very, very large variety of different applications—from virtual reality to drones, from drones to forklifts, from forklifts to humans, from industrial humans to sport. So this way we have multiple types of beacons. Some of them are only receiving ultrasound, some of them are only transmitting ultrasound, some of them are universal—transmitting and receiving—and some of them are more protected in terms of moisture and even exposure. So we can have everything, whatever explosion-protected. And of course the price is different, sizes are different, power consumption is slightly different.

88:20 Different, but majority of them have batteries, except for industrial, and we have battery-less options. And of course, most of them can be used in the same architecture. Like, for example, you can build very diverse, non-linear circuit texture consisting of stationary beacon, Mini-RX, stationary beacon, Super-Beacon, stationary beacon, Industrial-RX, and the mobile beacon, Mini-TX, and Super-Beacon, Industrial Super. We can do that if you wish, so, but usually, of course, we recommend to have less—let's say two or three beacons. Finally, summary: no method suits all needs. Choose yours. It totally depends on your application. This certainly can recommend Marvelmind, but even Marvelmind is not perfect for all locations. Choose yours. Sometimes ultrasound is better.

89:18 Sometimes UWB is easier choice. Sometimes sensor fusion is an option. LiDAR could be an option. So it totally depends on the applications. Specifically designed for positioning systems like Marvelmind or GPS, ultra-wideband or GPS, is better for positioning in majority of cases than other—let's say—not specifically designed for positioning standards or types of systems like Bluetooth or Wi-Fi or LoRa. And my favorite: sensor fusion is the best if economically viable and technically feasible. You can always get the best result out of this. As usual, thank you very much for your time. If you have questions, don't hesitate and contact us. Message to inform. I remain calm. Watch demos. We publish quite often new demos and new tutorials, and this tutorial will be published as well, and we will be happy to see you enjoying our system. Thank you very much.