Wireless Positioning and Localization Design with MATLAB

Hello, everyone. My name is Houman Zarrinkoub. I'm the Product Manager of all the wireless communications products here at MathWorks. And I'm joined by my colleague and friend, Uvaraj, for this MathWorks seminar, which is entitled "Wireless Positioning and Localization Design with MATLAB." Uvaraj, please tell us more about yourself.

Hello, everyone. My name is Uvaraj Natarajan, and I'm from the Application Engineering team working in MathWorks. My primary focus area is on the wireless communication-related technologies, such as 5G, LTE, WLAN, Bluetooth, satellite communication, and so on. Nice to meet you all today.

Great. Let's go with the agenda of the talk today. First, we're going to start with some introductory remarks regarding positioning and localization technologies and methodologies.

And essentially, we're going to present today's talk in four sections-- the cellular approach to positioning localization, including 5G and LTE, the Wi-Fi systems and networks, the Bluetooth innovations-- and the finally, the ultra-wideband technology-- and finally, summarize what we have learned.

Let's go over some introductory notes. There are many applications of positioning, localization, and ranging in wireless communication of today. Positioning, essentially, is known as the mechanism for determining the position of different objects in space.

Localization is the take in those coordinates where things are and placing them on, essentially, maps and the reference to an origin. And ranging is the act of finding the distance where meaningful communication is possible.

All these three types of applications are found in different contexts-- in outdoor navigation, that we use every day, when we are trying to go somewhere using our cars. The indoor navigation is emerging, for shopping and other proximity-based applications.

There is medical and e-Health applications, asset and freight tracking, and managing the spectrum and transmitting based on the location of devices. And finally, ultra reliable and low-delay applications in Industry 4.0.

Now, we all know about the global satellite navigation systems, in the case of North America GPS. It's one of the most widely-used technologies for us. And it's ubiquitous in outdoor positioning and navigation.

But note that the typical accuracy of GPS GNSS seems to be-- for example, for smartphones, is about two-to-five meters. And if you look at the latest results for high-end dual-frequency applications, maybe a few centimeters.

But there are some challenges that remain. The indoor applications suffer from low signal strength. And that is a place where we should get better performance there. And there is a lower accuracy for precise positioning. That's why the wireless positioning standard and technologies have emerged to augment what is possible with GNSS and GPS standard.

There's a categorization here based on radio access network-- RAT dependent versus independent positioning. If the technology uses cellular networks-- in this case, LTE and 5G-- that's RAT dependent.

Whereas if the technology, like Bluetooth, Wi-Fi, UWB, doesn't use the radio access, then we call RAT independent technologies. No matter what technologies are used, there are some common methodologies that you see, that forms the basis of positioning and localization.

I've categorized them in four segments here for you. Some class of methodology are based on power or signal strength. For example, Bluetooth uses receive signal strength indicator, RSSI. Many approaches in Wi-Fi, LTE, 5G are time based, for example, time of arrival in Wi-Fi and time difference of arrival-- TDOA-- in 5G and LTE, as well as round time trip.

Now there are some that are space based. Bluetooth users angle of arrival and angle of departure, and that's space based. And there is a new emerging technology called fingerprinting, that you use in Wi-Fi standards.

Now, each of these methodologies and standard technology offer a trade-off. Nothing is perfect. So I here have gathered some useful information based on multiple sources, that paints a picture for you of each technology and how they differ in terms of comparison regarding positioning accuracy, the effective range, and power consumption.

Know that LTE and 5G below six gigahertz, they essentially have a lower accuracy, in terms of meters. Whereas the 5G millimeter wave release 17, for example, is promising less-than-one-meter accuracy.

Note that Bluetooth Low Energy BLE, or UWB, are notoriously designed to provide high accuracy. For example, UWB boasts of being in tens of centimeters. And the same is Bluetooth Low Energy-- definitely less than meters. Wi-Fi, on the other hand, is in a few meters-- 10 meters and so on. And of course, GPS, as you saw, in also two-to-five meters.

Now if you look at the effective range, what is essentially the optimal range that these methodologies are possible. Again, the LTE and sub 6G-- 5G-- they allow you to be outdoor and indoor with hundreds of meters. Whereas the BLE and UWB are intended for tens of meters. So is Wi-Fi. So it's mostly indoor.

And also, power consumption-- notice that Bluetooth and UWB are the lowest signature in terms of power consumption, whereas the cellular technology has more profile of power. And Wi-Fi is Something similar to it..

Now, having gone through such remarks, let's go one by one and look at what we have invested in MathWorks Tools for positioning, localization, and ranging in various standards that we support. As you know, the flagship product for 5G modeling and simulation and testing in MathWorks Tools is 5G Toolbox.

Now, in 5G Toolbox, we have recently added a couple of 5G NR positioning examples-- reference designs for you to learn about. One of them is the NR positioning reference signal definition. Will go through that in a second. And one is using the PRS to actually perform positioning. And as we see, it's based on time difference of arrival technology.

Let's go look at them a little bit further. Now, 5G NR positioning reference signal is a signal defined by 5G standard, that allows you to configure the signal needed to achieve accurate positioning. So it can configure the time frequency aspects of this signal. Notice that this methodology requires synchronization.

So multiple cells transmit PRSs in a coordinated manner. And note, for example, there's a pattern of transmission, you see at the bottom, that's expressed in the example, that shows when some g Node-B's are transmitting, some are muted. And the muting of relevant PRS transmission in some cells is useful to avoid interference from adjacent cells and make the act of positioning more accurate. So that is all the signaling that you see, described as an example to make the PRS possible in 5G.

Now, using that reference-signal definitions, there is a wonderful demo in 5G Toolbox that allows you to perform 5G NR positioning. Now, this demo and this example uses the reference signal timing differences-- RSTD-- otherwise known as time difference of arrival measurements.

So essentially each UE, or each handheld device computes or measures the delay difference, between the transmissions received from each g Node-B. And computes these hyperbolic loci, or locuses, of where the great difference seems to be constant.

Now then, we can use multilateration, or trilateration, by intersecting all these different hyperbolas to pinpoint the location. And notice that it requires synchronized transmission of PRS from all g Node-B's.

Now, I would like to invite my friend Uvaraj. Why don't you to tell us more about how this 5G NR positioning using PRS is implemented in 5G Toolbox?

Thank you so much, Houman. Now let's see a quick demo of apportioning and localization example. This is a part of the 5G Toolbox. So let's quickly go to the MATLAB. So I'm using the MATLAB R2022a. And to go to the Documentation window, type Doc 5G in your command window, which will take you to the Documentation page of the 5G Toolbox.

So here you will see a lot of materials related to 5G-- getting started and the details regarding all the channel signals, and various examples as well. So the example part will give you a list of pre-built examples, which you can use for your day-to-day R&D activities. And these are readily available for you to develop your system with the pre-built examples. And hence, it reduces your development turnaround time message.

Now let's quickly go into the positioning-based example, which is available in the 5G section. So we have two examples here-- one on the positioning reference signal, which is actually generating the PRS signal from the 5G-based PRS signal. And the other one on the NR Positioning using the PRS, which is actually using the PRS to get the position of the user.

So let's open this example. So this specific example show how to calculate the position of the user within the network of g Node-B's, which is available in the complete telecom network. And this example uses the observed time difference of arrival-- or OTDOA positioning approach-- to estimate the position of the user.

So the OTDOA approach is one of the downlink-based repositioning technique. And this technique uses the reference signal timing difference, RSTD, or the time difference of arrival measurement to perform the multilateration or the trilateration by using the theory of hyperbolas.

So let's see in this example how this is implemented and how we calculate the position of the user. So we start with calculating or defining the configurations of the users.

And so we define the number of frames, we need to differ-- we need to simulate-- and the position of the user as such and the number of g Node-B's we are interested. So here we have given five.

And this will generate five g Node-B's and plot it in the graph. So we have a UE surrounded by five different g Node-B's in a in a random fashion positioned in the graph.

So we proceed further with the configuration of the carrier and the 3GPP based PRS configuration. For this, again, we use an IP address config object, which is readily available as a part of the 5G Toolbox.

And we overwrite the configuration with the data, which we want to be configured. We also configure the PDSCH node, to transmit the data along with the PRS signal.

So we also configure the path and the other channel configurations, which has to be introduced between the signal, which is transmitted from the GNB to the UE And finally, we end up generating the PRS. So once the configuration is done, we generate the positioning reference signal and the PDSCH resource

So we create another resource grid, which is actually the dummy resource grid. And then, we populate the PRS signal, which is generated using the 5G Toolbox, and populate for all eNBs which we have configured.

And then, we also generate the data. So nrPDSCH data, and populate the same in all of the locations wherever the PRS signal is not located.

So the final chart looks like this, where we have the positioning reference signal for all the g Node-B's which we have configured. And the data-- PDSCH data, which is populated for all digital piece.

So once the resource grid is ready, we modulate the signal. So we use the NR way of DM Modulate function-- which gives the transmit waveform-- the IQ samples of the transmit waveform.

So once they transmit waveform is ready, we add the channel imperfections on top of it. So we add the path loss. We add the delays. We add attenuation to the signal, to create it into a generic or kind of real-world environment. On top of this, you also can add various other channel models on top of it. Like 5G based CDL OR TDL channel models, or any of the channel models on top of it.

Coming to the receiver part-- so once we are done with the transmission of the signal, we receive the signal and the receiver site, and then do the estimation of the timing. So NR timing estimate function help you to get the timing estimate of the signal and get the correlation of the signal in the receiver side. So the output of that gives us the g Node-B1, which is highly correlated, and the g Node-B2 and g Node-B4, again, have good correlation.

So once we get the timing estimate, now we get the RSTD values, to estimate the UE position using the OTDOA technique. So here, once we get the RSTD values for each individual piece, we calculate the hyperbolic equation and then calculate the distance of the user Equipment from the g Node-B.

So we also use the function get estimated UE position, which is actually an inbuilt function to get the estimated portion of the UE. So the algorithm runs, and data is calculated-- the UE position-- to be in this particular location, which is almost close to the location where we place the UE in the graph. And there is a small error. So the system also calculate the position estimation error.

And this changes based on the channel models and the other imperfections which you had. So this is the final graph, which you see where, where you have the g Node-B's and the position of the user.

So the actual position and the estimated position are very close to each other. And this helps you to identify the position of the user by using the OTDOA technique.

So to summarize this example shows the OTDOA-based UE position estimation in a two-dimensional array. And on top of this example, you'll be able to add various UEs-- various UE positions, or multiple g Node-B's located in a different position in the graph, add more channel models, and find out the strength of the algorithm positioning algorithm-- which you build by your own.

Thank you so much. And back to you, Houman.

Thank you, Uvaraj. We also have an LTE product-- LTE Toolbox-- that models the LTE, or 4G cellular standard. In that LTE Toolbox, we have one example related to-- same concept-- using PRS as defined in LTE standard for time difference of arrival positioning. And that one is also available for you, if you are working on 4G systems.

How about the positioning localization in the Wi-Fi, or wireless LAN context? As you know, the Wireless LAN Toolbox is our product that models all different standards that belong to the 802.11 Wi-Fi family of standards. The Wi-Fi 802.11az is known as the next-generation positioning standard.

And we have two examples in the Wireless LAN Toolbox that allow you to examine how accurate you can do Wi-Fi position. One of them is called three-dimensional indoor positioning with 8 to 11 a z fingerprinting and using deep learning networks.

So it's a combination of using fingerprinting, or computing the received power at a grid inside-- the indoor environment. And then using deep learning to discriminate, as you are in the room, what is the best match for you in that context. That's a wonderful combination of techniques that uses deep learning to make the measurements even more accurate.

And the example that we're going to go through is equal to 802.11az positioning using super-resolution time of arrival estimation-- TOA. And let's go and learn more about that.

This example is based on time of arrival, as defined in the IEEE 802.11az standard. It estimates the time of arrival by using multiple-skill classification, super-resolution approach, and then estimates the two-dimensional position of a station by using, again, trilateration.

Now, Uvaraj, it would be nice if you can show us in more detail about this demo.

Thank you so much, Houman. So let's see an example, which is based out of WLAN standard. This is actually IAAA 802.11az, which is termed as the next-generation positioning standard. So let's quickly start with the documentation of WLAN. Type doc wlan in the Command window, which will help you to take directly to the documentation, or the Help page of the WLAN Toolbox.

So here, you will see a lot of information regarding the basics of WLAN Toolbox. MATLAB support all the WLAN features and functionalities for various flavors of the WLAN standards-- in the physical-layer modeling, or MAC modeling, or even signal transmission reception, end to end link-level simulations-- system-level simulations, distance measurement, and so on. So there are various use cases, which can be performed using the WLAN Toolbox.

And there are various inbuilt examples. If you click on the Examples tab, you will find a list of examples, which will help you to make your work easy. So currently, we are going to focus on the 802.111az positioning using super-resolution time of arrival estimation example.

So this example uses 802.11az for estimating the position of the user. So this example, again, uses three access points and one station. And you'll be able to simulate multiple access point more than this, as well. So this example show how to estimate the position of the station in a multipart environment by using time of arrival positioning algorithm, as defined by a IEEE 802.1az Wi-Fi standard.

This example estimate the time of arrival by using the basic algorithm, which is multiple signal classification super-resolution approach. And then estimate the two-dimensional position of this station, by using the technique of trilateration.

Let's quickly see the complete flow of the transmitter and receiver. So we have a transmitter, which generate the HE waveform. And then be passed the waveform into the channel, which introduces the delay. We also have IEEE 802.11-based TGax multi-part channel model. And we also add AWGN on top of it.

Again, on the receiver side, you have the packet detection, timing synchronization, frequency correction, and the de-modulation of the signal-- and finally, the channel estimation-- so HE-LTF channel estimation module will help in that.

And after the packet reception, we do calculate the distance from the station to AP-- and finally, calculate the position of the user. So distance ranging is calculated mainly based out of uplink time of departure and time of arrival-based techniques.

So we have a station, which is transmitting at time t1. And that's received in the AP at time t2. There is a delay while calculating the distance using the MUSIC algorithm. And then the downlink transmission, again, happens at t3. Reception happens at t4. Again, MUSIC algorithm is used to estimate the distance.

So here the round-trip time is calculated, and the distance between the station and the AP is calculated using this particular information. And we also have elaborated diagram, which illustrates the distance ranging process of the MUSIC algorithm. So you are free to use your own algorithm implement for the time of arrival estimation and so on.

Now, going through the simulation parameters, there are various iterations, which we run the complete simulation. And depending on the number of iterations you run the simulation, you will see the result in a different fashion. And we also define various SINR ranges and number of access points. At least three access points are needed for the trilateration process.

And IEEE 802.11az gives flexibility to have multiple-channel bandwidth and, again, multiple delay profile configurations. You're also having the flexibility to select number of transmit and receive antenna as per your hardware configuration-- so of the simulation configurations. You can also add repetitions on top of it.

So on top of that, you'll also be able to do channel configurations. So WLAN Toolbox support the IEEE 802.11-based channel models. So we use the TGax channel model with Model B delay profile. You also have various other choices to make here.

And the ones the channel model is defined-- so we also have the ranging measurement, which is the final simulation, which we are going to do here. So here, as a part of the ranging measurement, we run the complete simulation for various SINR points-- and the number of iterations, which we configure.

So the more the iterations, the more the simulation accuracy would be. So we run the complete simulation for various iterations. And we have the waveform generation. Add the signal delay to the waveform.

Add the channel model on top of that AWGN, and then the complete receiver operation can be performed. So this is the complete cycle, which runs for various SINR points. To calculate the range, we use the algorithm, as we saw in the figure above .

So finally, the algorithm calculates the distance and the ranging mean absolute error is also calculated, which is actually in a very less than a meter kind of range, for various SINR points.

So the next sequence is the trilateration process, trilatering the location of the STA in two dimension, by using the distance estimate. Then, calculate the position using root mean square for each iteration, by using this station position estimator. So finally, we display the RMS positioning error again, which is actually less than meters in terms of various SINR points.

So you also have flexibility to add more SINR points on top of it, or more backchannel conditions, delay profiles can be added. And you can see the performance of your positioning algorithm with that kind of environment.

Finally, we come up with a figure where you have the APs trilaterated, to find the position of the station. So this is actually a node, which is run at 35 dB SINR. But you increase or decrease the SINR and see various imperfections added on top of it. And the accuracy changes.

So this is with respect to the WLAN-based example. And there are various other resources which we have for you to explore further later on. And the references to the standards are also given as a part of this example. So hope you enjoy. Thank you so much. Back to you, Houman.

Thank you, Uvaraj. Next, look at the positioning and localization approach in a Bluetooth standard. And recently, in 2022a release of MATLAB, we have introduced the Bluetooth Toolbox. It's a Toolbox that enables you to simulate, analyze, and test Bluetooth communication systems and networks.

Within the Toolbox, we also have included two essentially localization, or positioning examples. One of them is known as Bluetooth LE positioning, by using direction finding. And that is the model that uses the angle-of-arrival departure. And one of them is Bluetooth early-direction finding for tracking node positions.

Now, you can do node positions using angulation techniques. Triangulation is fine. And as you can see, if you know the angle of departure from multiple Bluetooth beacon, and the hubs relative to a Bluetooth device, we can triangulate and find the location of that.

So of course, in a Bluetooth context, the signals, wavefrom generation, and receivers are included. And for the purpose of calculating node position, there a constant tone extension is added to the data. And that constant tone is used for your angulation techniques.

Let's look at the more detail. So the angle of the arrival context, on the top, notice that the VRE transmitter transmits the wavefront in the direction of the receiver. And the receiver has an array of antennas.

So when you have an array of antennas, you are essentially using angle-of-arrival estimation. And in the context of where the transmitter is MIMO and has multiple antennas in an array-- and the BLE receiver is single antenna-- you're talking about angle of departure.

Now notice that in order to find the angle by which transmission is done, or the angle between transmitter and receiver in both of the context, you have to know the phase difference-- in this case, that constant tone. The phase difference in the constant tone provides that. You have to know the lambda, which relates to frequency of the tone. And you have to know the separation, or the distance between antennas in the antenna array. That's known as D.

Knowing that, you can find out the angle of departure and arrivals and perform triangulation to pinpoint the position. What is that? There are multiple techniques to localize the nodes.

You can calculate position with the antenna array and AoA and AoD processing. And it can use either 1D or 2D arrays, as specified by the Bluetooth standard.

And you can assess positioning accuracy based on various SNR levels. So as noise increases the performance, of course, degrades. And you can optionally use Kalman filtering to increase the accuracy. So these are all discussed in a Bluetooth Toolbox, as is released this year as a standalone Product .

Now let's look at, finally, the ultra wideband approach to localization and ranging. So notice that our Communication Toolbox, our base products for wireless communication, has a library-- an add-on library known as Communication Toolbox Library for Zigbee and UWB.

If you go to MATLAB, and if you have Communication Toolbox, just to go to Add-on Explorer, as you can see here, by clicking on Add-On. And just type you UWB, or Zigbee, and you download the library that you see here.

Now there are two ranging and localization demos available in the free add on that comes for Zigbee and UWB, with Communication Toolbox, as you can see here. Readability ranging using IEEE 802.15az, and UWB localization using the same standard.

So the example is known as one-way ranging, or time difference of arrival example. It presents the UW approach of localization, but transmissions are either ennobling or in downlink. In downlink, the synchronized nodes periodically transmit broadcast messages to the device, with a known time offset, right?

And the time difference of arrival-- difference between these things-- are used to form, again, that hyperbolic surface of each pair of synchronized nodes. And then the intersection of all these hyperbolic surfaces gives the location estimate. It's very similar to the approach we saw in the 5G case.

Finally, if you want to learn more about all these different approaches that we have, notice that we have 5G, LTE, wireless LAN, satellite communication, and Bluetooth-standard-based products. And you can look at their product pages at MathWorks.com/products, and you can find them.

If you want to get access to UWB, or ultra wideband positioning and localization and ranging functionality, you've got to look at Communications Toolbox and download, for free, the library for Zigbee and UWB. And if you want to see the Wireless Communication Solution page, you get a chance to see the overall solutions we're providing in the wireless communications Space.

To summarize, positioning, localization, and ranging have many applications-- indoor and outdoor wireless communication. Multiple technologies, including GPS, 5G, LTE, Wi-Fi, Bluetooth, and UWB, have developed positioning techniques with different profiles of accuracy, complexity, and reach. And MATLAB, its wireless toolboxes provide algorithm, simulation, modeling, and analysis for these methodologies.

And using the open MATLAB code in these tools in MATLAB, you gain hands-on insight in every aspect of signal generation, measurement, analysis, and estimation of position and ranges.

And then not only, you can seed actual code-- MATLAB code-- describing these positioning localization algorithms. You can edit them, modify them, upgrade them, and be part of developing the systems of future, and optimize your performance.

With that, on behalf of myself and my friend Uvaraj, we want to thank you. And very soon, we will open the floor to the questions that you may have. Thank you very much.