Explore 6G technology with MATLAB

Good afternoon, everyone. Welcome to this webinar on "Exploring 6G technology with MATLAB." And this is the agenda for today. In our webinar today, we will start with a brief introduction to 6G technology and its evolution. Then we'll talk about the new waveform candidates for 6G wireless.

Afterwards, we will discuss the end-to-end simulation scenario of base-band system, followed by the RF modeling and test and measurement aspects of communication systems. Then we will talk about one of the important key enabling technologies of wireless system called reconfigurable intelligent surfaces. Or we can call it as a intelligent reflecting surfaces.

Then we will discuss AI and machine learning algorithms for wireless communication systems, followed by how to accelerate the large and long-running simulation scenarios using multi-core computers and clusters. And at the end, during the conclusion, we will talk about the key takeaways and the summary of what we have learned today.

An introduction first. What is 6G? 6G is a next-generation wireless communication system, which is built on the strength of 5G. The vision for 6G includes significantly higher data rates, lower latency, massive connectivity, a greater reliability, and substantial connectivity, and so on and so forth.

While still largely in the research phase and not yet standardized or commercially available, the 6G aims to further transform the capabilities of the wireless networks. And that will definitely build upon the strength of 5G.

And recently, the International Telecommunication Union, ITU, published the 6G usage scenario. You can see in the slide, we are extending from the 5G IMT-2020 document scenarios. And we had enhanced mobile broadband system in 5G. Now it is called it as the immersive communication system. The massive machine-type communication is-- call it as the massive communication now. And the ultra-reliable low latency communication that is replaced by the reliable and low-latency communication. So these are the extension of 5G use cases and scenarios.

And what is new in 6G use case scenario? The main thing is the omnipresent connectivity, integrated AI and communications, integrated sensing and communications. Now, sustainability-- connecting the unconnected, the ubiquitous intelligence, security, privacy, and resilience-- these are the four overarching aspects of the 6G usage scenario. This is recently published by ITU, so we can call it as-- it is the extension of 5G.

The 6G technology is anticipated to unlock a host for new applications and use cases that could significantly impact various sectors-- the new applications, like the integrated AR, VR, XR, which is augmented reality, virtual reality, and extended reality, and physical experience, the advanced industrial automation using real-time robots, the smart cities and infrastructures using a wide area and micro connectivity, network generation-- the next-generation emerging services using digital twins, the autonomous vehicle, and smart transportation, high-resolution, holographic displays for mobile communications, and space exploration and connectivity, and much more.

These applications leverage 6G's higher data rates, almost instantaneous latency, enhanced reliability, and massive connectivity.

And this particular slide shows the likely development timeline of 6G specification between now and 2030. The 5G specified the IMT-2020 specification. We know that. Similarly, the ITU plans to create a an IMT-2030 specification. And it is likely that 6G specification will aim to satisfy this.

Some of the key milestones are going to be the requirement agreement at the end of 2026 and then possible decisions about the 3 and sub-terahertz spectrum at the end of 2027.

In terms of development of a technical specification, whether by 3GPP or some other body-- any other body, maybe-- emerging in the end of this decade, around 2027 to 2030. Obviously, the technical specification is still a long time away. Therefore, we want to support our customers to bridge the gap between now and the 6G specification while supporting the R&D efforts of 6G.

And there are a number of enabling technologies that keep coming up in research and discussions with our customers. I have summarized some of them here. We can call it as a key enabling technologies for 6G. There is a big interest in the community to use AI/ML and all parts of the 6G spec.

There is also a big push to use the higher frequencies where the larger bandwidth is available, such as sub-terahertz, even terahertz, space optical communications. But there are significant engineering challenges in these frequency bands. We'll have to address that.

Attractive technologies, such as reconfigurable intelligent surfaces-- IRS-- or intelligent reflecting surfaces-- IRS or RIS-- people used to call any of these names-- IRS and RIS-- are being proposed to overcome some of the challenges working at the higher-frequency bands.

It is very likely that the non-terrestrial network, the satellite communication, will be highly integrated into 6G to facilitate communication for all and everywhere. Technologies like cell-free massive MIMO are certainly very interesting for improving the network efficiency.

And of course, we are looking into the crystal ball to predict what are the other technologies that may be used. Something else will be the part of the 6G. You can please comment on the chat window. And if you are working on some other technology other than this list, will you please put it on the chat.

The spectrum is, of course, critical to any wireless communication system. And 6G connectivity will be able to operate in a wide range of frequency bands. This includes the current frequencies, which is FR1 and FR2. FR1 is in the range of 410 megahertz to 7.125 gigahertz.

And the 24 megahertz to the 56 gigahertz of millimeter wave frequency, which is FR2, and-- as well as the new proposed frequency ranges, which is from 7 to 24 gigahertz. It is a centimeter frequency range. And 90 to 300 gigahertz of frequency range, which is the terahertz frequency range. These are the new proposed frequency ranges.

And now we are introducing the Exploration Library, which is the extension of 5G Toolbox. If you have a 2024 version of MATLAB and the 5G Toolbox, you can download the 6G Exploration Library from the add-ons free of cost.

With this library, you can explore, model, simulate and test 6G waveform candidates and 6G technology. I'll just take a pause here. There will be a polling question. Kindly answer this polling question.

OK. Thank you for the overwhelming response here. I'll just go ahead. Waveform generation. And waveforms-- the increased subcarrier spacing beyond 5G numerology are one of the new waveform candidate for 6G. There are so many other announcement which is happening in this area. This is one of the key-- one of the key candidate for the waveform generation for 6G.

And Hexa-X deliverables defines the OFDM waveform with subcarrier spacing of 1.92 megahertz and a 3.84 megahertz, which is actually the extension of 960 kilohertz, and 480 kilohertz and 960khz from the 5G numerology. So the Hexa-X is proposing these two extended subcarrier spacing, which is 1.92 megahertz and 3.84 megahertz.

And here in MathWorks, we are providing a chance to explore higher than the subcarrier spacing, beyond 960 kilohertz. You can go from 1.92 megahertz and 3.84 megahertz, 7.68 megahertz, 15.36 megahertz, and so on and so forth. The OFDM system is the increased subcarrier spacing are shorter in time, which facilitates reduced latency, of course. A wider subcarrier spacing improves the tolerance to phase noise, which is more severe at the higher carrier frequencies.

And for example, configure a carrier with the subcarrier spacing of 3.84 megahertz and 60 resource blocks-- 60 PRBs-- and compute the sampling rate and transmission bandwidth. And the model uses 5G numerology rules to generate a waveform. And the same can be used for further analysis.

Here, you can see the tx waveform, the transmitter waveform, which is part of the 6G. But this is built on the 5G numerology rules.

And we have an example-- a shipping example-- on this particular use case. You can play around with that, which is already a part of the 6G Exploration Library. You can use it and explore the 6G waveform, which is developed on top of 5G numerology.

Orthogonal time frequency space modulation technique-- it is one of the new waveform candidate for 6G wireless communication system again, which is completely new, which is not the extension of OFDM or it is not built on top of 5G numerology, which is a new technique.

What gives modulation technique highlights its inter-carrier interference cancellation capabilities, as compared to the traditional OFDM modulation? In higher Doppler channels, the channel characteristics change rapidly, resulting in low channel coherent times. Coherent time is inversely proportional to the channel coefficient's variability. So the OFDM has been the modulation scheme of choice for various wireless systems for many years.

With OFDM, the higher Doppler channel environment requires frequent channel measurements and experience-- and which experiences ICI. The OFDM transmits data in time frequency domain-- the TF domain.

For the frequent channel measurement, we need more reference signal for the same. So the signaling overhead will be increased so that the room for data channels get reduced, which causes reduced throughput. This is one of the main challenges-- or one of the main drawbacks in OFDM

To overcome that, the OFDM modulation, which removes the need to have the frequently measuring channels-- because it transmits data in the Delay/Doppler domain. Due to its resilience in high Doppler channels, the OTFS is being considered as a modulation candidate for 5G and 6G as well.

To understand how Delay/Doppler domain transforms to the time frequency domain, you can relate the process of OFDM. But in this block diagram, the internal section of this diagram shows-- which is actually familiar with OFDM processing. OFDM been processing OFDM modulator, channel, and OFDM demodulator.

The Heisenberg Transform or Wigner Transform, or the generalized-- or the generalization of the OFDM modulator and OFDM demodulator-- the inner part of it.

First, the inverse symplectic finite Fourier Transform-- ISFFT-- modulates symbol from the Delay/Doppler domain to the time frequency domain. The combination of ISFFT and the Heisenberg Transform can also be mathematically represented using the inverse Zak Transform.

When it combines ISFFT and Heisenberg Transform, eliminate the IDFT and DFT pair, such that the operation simplifies to an IDFT across the Doppler axis. Similarly, the Wigner Transform and the symplectic finite Fourier Transform-- ISFFT-- can also be represented using the Zak Transform.

So on the whole, that input is inverse Zak Transform, and-- I mean the transmitter-- it's ISF-- I mean inverse Zak Transform. At the receiver, it is Zak Transform. So that forms the OTFS modulation. It is one of the flavors of OTFS here.

And you look for the OTFS example in the 6G Exploration Library. But this particular example is a part of the Communication System Toolbox. To run this example, configure the simulation parameter with the higher adopter environments, and run the simulation.

And you can compare the constellation against OFDM, that OFDM has a strong residual of ICI because it involves higher Dopplers. So where does OTFS compare? It compensates for the Doppler.

OTFS is a promising modulation scheme to mitigate the effect of high Doppler in the time-varying channel environment? This example introduces the concept of Delay/Doppler domain, the effect of the mobile channel on the transmitted symbol in that domain, and how to modulate and demodulate OTFS symbol.

Transmitting the same data over the same channel using OTFS and OFDM and observing the block error rate, after a simple channel estimation and equalization source, the OTFS effectively combats inter-carrier interference, as compared to OFDM. And we have a shipping example for that as a part of the Communication Toolbox. You can check it out.

And link-levels simulation-- end-to-end simulation on the physical AL level-- for that, MathWorks has a long history of developing channel models, and we have created a number of models which is suitable for 6G exploration.

We provide 3GPP, CDL and TDL channel models which are capable of modeling up to 100 gigahertz of carriers, as per the specification-- as per 3GPP specification, TR 38.901. In addition, with the TDL and CDL, we have the high-speed train-- HST channel model as well.

And on top of it, MathWorks invested a lot of time on modeling raytracing channel models for a number of releases, now, providing multiple methods and adding capabilities.

So that we know that these channel models-- the 5G proposed channel models-- TDL and CDL channel models-- are empirical channel models, but the raytracing channel model is the real-time channel model, we can say. It's the realistic channel model, the 3D channel model.

And we have a reference design for 6G link-level simulation, which is a part of 6G exploration library. And it shows how to measure the throughput of a pre-6G link. It is based on the 5G, but allows you to explore larger bandwidth and subcarrier spacing beyond 5G.

You can use more than 275 resource blocks and subcarrier spacing bigger than 960 kilohertz. This we have discussed in our earlier slides itself to use the same set of configuration parameters in this example as well and run the simulation.

And in this example, the link-level simulation of a PDSCH channel is considered. And you can reuse the same thing for other channel also-- PDSCH channel as well.

And here, in this particular example, the DL-SCH transport channel coding has been considered-- the multiple code words depending upon the number of layers, the PDSCH model demodulation reference signal demod generation, and variable subcarrier spacing starting from 15 kilohertz till 15.36 megahertz.

And it uses both the normal CP and extended CP as well. And here, we have considered the 3GPP 5G channel models, TDL and CDL channel models. And we haven't used any new channel model-- raytracing channel model or something else. For this particular example, we have used the same 5G channel models.

And the PDSCH band precoding and singular value decomposition-- the technique has been used. The perfect and practical channel estimation-- the perfect channel estimation is the one which is directly reused from the channel model. The impulse response-- the channel model is directly used. And the practical channel model-- I mean channel estimation technique is-- and use the channel estimation technique based on LS estimate or-- LMS estimates.

And of course, the HARQ operation with up to 38 processes have been used for this particular example, and so on and so forth. There are a lot of other parameters, as well, for this. And we can measure that PDSCH throughput of a pre-link-- pre-6G link. So here, with higher number of subcarriers, higher number of base, what is the maximum throughput we can achieve? This is being analyzed.

And there is a live example for this as well. And it is part of the 6G Exploration Library. You can access this and try it out.

And measurements and RF modeling-- how to explore the impact of hardware impairments at sub-terahertz frequencies on a 6G candidate waveform. The hardware impairments--- we know that there are very big list of hardware impairments, like phase noise, power amplifier non-linearities, and filter that limit the spectral emissions outside the channel bandwidth.

So here, in this particular examples, we have considered these three-- phase noise, filter, and power amplifier non-linearities. And you can introduce the other impairments as well. Other than these three, you can introduce the phase offset, I2 imbalance, frequency offsets, time offsets, all those stuff.

And this reference design measures the adjacent channel leakage ratio, or adjacent channel power ratio measurement, and the error vector magnitude, EVM, measurements. We that these two are the important transmitter figure of merit of any transmitter system.

And through EVM demodulation error, other frequency offset, phase offset, phase noise, and other transmitter figure of merit can be analyzed. And at the same time, when the TCL measured or the ACPR is measured, we can analyze the spectral mask emissions, out-of-band emissions, channel power measurements, and so on and so forth.

What I mean to say here is we are not limited to only the EVM measurement and ACPR measurements. With these two measurements, we can perform all other independent measurements as well-- the other transmitter figure of merit measurements as well.

And we have a shipping example of the same. And we have designed with phase noise, lowpass filter, and power amplifiers. You can introduce the other impairments and you can check the ACPR and EVM metrics.

Our next-- the reconfigurable intelligence surfaces-- RIS. For the last several years, the reconfigurable intelligent surface, RIS, also known as the IRS, has emerged to be the key enabling technology of 6G.

Since the appearance of system, antenna arrays have become a critical part of both base station as well as the user equipment. The millimeter wave bands adopted the 5G system-- make it possible to pack larger arrays in terms of number of elements into a tight space. In return, the beamforming gain provided-- there are arrays-- there are the list of arrays-- help to compensate the severe propagation losses in these micro-- I mean millimeter wave bands so that the desired data rate and the capacity can be achieved.

But here, the RIS is a passive planar array formed by many elements, antenna elements. It is important to highlight that the surface is passive, thus keeping the cost low, as there is no need to include the RF circuitry in the surface. However, each element can adjust phases of the reflected signal.

So a controlled-- I mean, this one-- this particular controller could coherently combine them by adjusting the phase at each element. Currently, there are two approaches to form such a surface. The first approach is to use the reflectarray with a phase shifter for each element. This works a lot traditional passive phased arrays.

An alternative approach is to build the surface with a metal surface. By modifying the metamaterial characteristics, the reflected signal from a given element can be modified. Compared to the reflect-- reflectarray, the elements in the metasurface can be much closer to each other, which makes the entire aperture more compact.

And here, we can see the RIS system model, which has a transmitter with the pre-coded W, and in between the RIS elements and a receiver. From the transmitter to the RIS element, there is a channel we can consider a serial channel here. The same way, from RIS elements to the receiver, there is a serial channel.

The mathematical model is y is equal to h theta G ws plus n, where h is the channel coefficient between the RIS element to the receiver. And G is the channel matrix between the transmitter and the RIS elements. And theta is the diagonal elements of beta i e power j theta i. And w is just the precoding matrix. And this is the transmitted signal plus n, which is GN, of course.

And here, this example-- as we discussed before, between the transmitter and receiver, there is a serial channel. And between the RIS equipment to the equipment, there is another CDL channel. This stochastic channel modeling is being considered here.

And when we run the simulation, when the RIS is disabled completely, you can see the completely scattered constellation. When that RIS is enabled, you can see the way better constellation, as compared to previous. But with this constellation, you can regenerate your data, but with the scattered one, it's completely lost.

Yeah, we have an example for this, a shipping example on this. You can please refer it. Now I'll pass it over to one of my other colleague. He will talk about AI for wireless and other stuff. Thank you.

Thanks, Yeah, you can move to the next slide. So Damon was introducing us-- AI is playing a crucial role, primarily on enabling technology for 6G workflows. So today, we thought of giving you a brief overview about the different steps in AI workflow and how MathWorks, as a company, is enabling AI plus wireless applications. can you move to the next slide? Yeah.

So just to give a context, oftentimes we consider AI modeling as a single step, but it is a small piece in the overall ecosystem. It has a lot of steps associated with this. So it ranging from the data preparation, AI modeling, validation of the AI modeling to the system. Particularly in this case, it can be a wireless system where you can able to bring the trained AI models into the simulation and validation framework. And then, finally, we'll take it to the field deployment.

Of course, these are the integrated steps, and there are-- a lot of iteration comes into play for modeling and validation. But in general practice, a lot of steps can be automated in MATLAB. That's one of the USP that we bring to the table when you are starting your journey with the AI workflows.

And moreover, if you carefully notice on the interoperability piece, on the AI modeling bucket, primarily, even if you have already started the work in other frameworks, such as Python, TensorFlow, Keras, or PyTorch models, you can able to bring it into MATLAB, and you can able to perform the simulation and testing or even the deployment workflows.

And how we are supporting the domain engineers like you is primarily, we have a number of apps to support the complete end-to-end workflows. Let me give you a quick overview about the AI apps. Can you just go to the next slide?

So here is a snapshot of a multiple apps. Primarily, what I'm showing here is two different apps. One is focused on development of deep learning workflows or a framework where you can simply interact with the frameworks and you can able to build your network, which means you don't want to worry about the syntactical error or you don't want to learn a new programming language to build AI kind of models.

Primarily, domain engineers like you, the experts in wireless-- kind of subsystems, you can able to easily develop the AI models just by interaction with these apps, and you can able to get started with it. There are multiple apps available in the frameworks to support the AI workflows.

So we are not also stopping the programming enthusiast. If you're already a programming enthusiast and you want to have a better control and you want to understand the inner line code, whatever you are seeing in the screen, you can able to do equivalent in the programming workflows as well. So ideally, you can able to develop, train, model both in UI-based environment as well as programming construct. We leave it to the choice of you by which you can able to achieve the quicker results-- you can able to decide and select it. So in a way, we are enabling AI with wireless. Can you just move to the next slide?

So just to give a context, briefly touched upon the Wireless Waveform Generator and different set of examples and things like that. And I just showed a couple of apps with the AI workflows. End of the day, you try, at your end, to make sure that like you can able to leverage the AI workflows for your 6G development.

With this case, we have both like data generation or a system-level simulation in the MATLAB framework and Simulink framework. Equally, we support the AI development as well. And moreover, we also support the interoperability by means you can able to bring the readily available models that you already developed in Python into MATLAB for system-level testing as well as deployment workflows. Next slide?

Yeah, so just to give a context, we are having a multiple examples focusing on AI wireless to get started for you. Similar to the wireless examples which talked to you, here are some of the examples in a broader area, starting from spectrum sensing all the way to transceiver design and digital PA, power amplifier, kind of design and so on and so forth, where we have a complete end-to-end framework starting from developing the data from the Waveform Generator or bringing the data from your software-defined radios, and then getting started with the pre-processing model development, testing, and deployment workflows. So in a nutshell, it covers the complete end-to-end workflow. Can you just move to the next slide?

As coming to this all this AI-driven model, or in a nutshell, even these wireless simulations-- it is computationally intensive. You take care of-- or you work in multiple matrix manipulations and so on and so forth. So in order to leverage the multi-core capabilities and GPU available at your perusal, there is a dedicated workflow in MATLAB, which we call as a parallel computing workflows. Can you just move to the next one?

So here are some of the examples or customer testimonials, where they leverage the computing workflow in MATLAB and accelerated their wireless development. And equally, we support this for wireless plus AI development as well. So how we are leveraging is there are multiple ways you can able to use it. For example, let us assume that you have a laptop or desktop, which has a multi-core capability, or a GPU machine. You can able to seamlessly bring it as a compute node, and you can able to accelerate your overall development process.

And there are cases where you find you have dedicated clusters, remote clusters, like kind of a high-performance cluster machine that is available at your end, and you want to offload the simulation or training to those machines. Definitely, that is also possible. And even you can able to leverage cloud directly to accelerate this kind of process.

And one interesting factor is there is-- no specialized programming is required to perform the parallel computing workflow. Most of the time, for MATLAB apps and things like that, it's automatically enabled for parallel computing workflows. Provided that you have respective toolboxes at your end, you can able to run the simulation in a parallel mode.

Just to give a sneak peek, what are the toolboxes that we support out of box, apart from the wireless one? I'm just going to the next slide where you can able to see the list of toolboxes that is supported for you to get started with. This is very simple subset. There are a lot of other toolboxes also supported. For example, signal processing, toolbox, communication workflows, image processing, or deep learning, machine learning, so on and so forth is heavily supported on parallel computing workflow.

In a way that you don't want to worry about time and effort required to run the simulation or train a model, you can able to simply build the code and simulate or offload your simulation to your high-performance clusters that is available to you. With that, I'm handing over to for the conclusion of workflows.

And let us conclude today's session. And in today's webinar, we have talked about the introduction to 6G Exploration Library and the new candidate waveforms with respect to 6G, which is built-- one is built upon the 5G numerologies. Another one is the OTFS. And then we discussed about the link-level simulation for a pre-6G waveform.

And then we discussed about the RF impairments, the various RF impairments in the hardware, the RF hardware, for sub-terahertz and millimeter wave frequencies for the 6G waveform. And we discussed IRS.

And then we talked about the and the machine learning technique for wireless communication systems. And followed by-- we discussed about how-- accelerating simulation for large and long-running simulations using multi-core computers and clusters. And thank you so much for attending today's session.