ON Semiconductor Corporation (ON) Earnings Call Transcript & Summary

Hello, everyone, and welcome to today's technology webinar on how ON Semiconductor autonomy mobile robot solutions overcome technical design challenges brought to you by ON Semi. I'm Kyli Miller, and I'll be your moderator today. During this webinar, you will learn about the ON Semi autonomous mobile robot, which is a comprehensive robotic solution signed with highly effective and innovative products from ON Semi. At the end of the webinar, we'll be holding a Q&A session to answer any questions you may have. During the webinar, you can type your questions into the chat box on your right, and you will answer them at the end of the webinar. I will also share the slide deck of the webinar with you during the Q&A session, and this webinar will be recorded and posted onsemi.com. You'll be notified via e-mail when the recording is available. Now let's meet today's presenter. Theo Kersjes works at ON Semi as the industrial business and solutions development with over 25 years international experience in product management, marketing and applications in the semiconductor industry. Theo's currently responsible for Industrial business development and solutions in the Advanced Solutions group. Now let's get started with the webinar.

Thank you, Kyli, for that wonderful introduction, and everybody, welcome to this presentation on autonomous mobile robots. The agenda for today will be to go over some examples and also stories about robots and how they progress in their wisdom of becoming more integrated in the environment and in our solutions. And we go focus in into the autonomous mobile robot that ON Semi created to highlight our technologies in vision systems and the different sensor technologies as well as motion, power lighting and communication. You see here on the bottom right, this picture about the progression of robots. I initially thought actually to put a different picture there, which was a variation on the kitten, the little kitten that looks into the mirror and then sees a big lion so it has a different self-perception. If you look at robots, I thought, well, let's put this first robot in front of the mirror and then it will see this an evolution and it is very advanced robots. But the point what that picture would make in this context would be we as humans don't see how intelligent the role of this that we're dealing with. For instance, in the car space, you see that with automotive driving solutions that you see a car on the road, you have no idea how advanced it is with its technologies of self-driving. Just last month, there was this example of in San Francisco, where the self-driving taxi had an accident and it's actually a very interesting accident to look at because another car hit a pedestrian and that pedestrian fell in front of the self-driving taxi. Self-driving taxi stopped, as it should be, safely, everything was fine. But then after a minute, it decided that it would be safer for its occupations -- occupied to be at the side of the road. So it drove about 100 feet or 30 feet to the side of the road, dragging the pedestrian with it that what you say accident. And the conclusion of this was that at this point, the permit for that self-driving solution is stopped. To highlight the complexity of these systems and what they need to look at. If you think of the same scenario where you're driving the car, you get your whole family in the car, spouse next to you, 3 kids at the background and back seat. And the same scenario, you just barely stop and this pedestrian is in front of your car. And all of a sudden, you check your back mirror and you see a freight liner coming down the hill and somewhere in San Francisco or a cable car that derailed. And you just do the same maneuver, you go about 100 feet to the side, save your family, save the pedestrian that you dragged along instead of making a coalition. So these are not easy decisions these systems have to make. And in automotive, they are very obvious. What happens there in the industry, we have more safety regulations as well, and they're not that obviously not public, these kind of examples. So we go quickly back then to our industrial automotive solutions. And here, you see the evolution of our industrial numbering and we are currently in the Industry 4.0. Industry 4.0 is dealing with concept as automation, artificial intelligence, connected devices, big data, data analytics and more. Big Data, by the way, the example that I read that really brought that home for me was a solution where they had collect the data, video data out of, again, automotive, self-driving cars or cars in general that we're driving on the road. Some of it we found out that after a certain time, the video collected basically had the whole city mapped out. So all the roads in the city were visited in a certain time when looking at these video materials, and they came up with a system to look at potholes in the road. So they say unevenness in the road and then correlated that with sensors in the car motion sensors to find the biggest potholes and they created a automatically created a list of the first 5 worst potholes in the city that should be fixed. So this is an example of where you have big data, you collect the data, you don't know yet what to do with it. But this was a really nice solution and I think a better one than every time when I research a product and buy it, did I get the rest of the month recommendations to continue buying the same product. So that's Industry 4.0. Industry 5.0, in our new emerging phase where you see humans work alongside artificial technology and AI-powered robotics, and this is coupled with a more human-centric focus as well as improved focus on sustainability. So what does that mean? All these industries have concepts of an enabler, a technology that enables this next step as well as social economic impact. For instance, in Industry 3.0, we had the computer coming up, and that allowed us to do -- make a new step in the industry scenarios. But the social side effects or social things that are dealt with in that era as well are, for instance, offshoring. So in the Industry 5.0, this focus is, and I'm now quoting an important distinction between 4.0 approaches the use of technology to provide prosperity beyond job and growth while respecting the production limits of the planet. So pretty lofty goal, and this falls into this sustainability aspect. And if any of you have a really good example that brings this home like some of the examples I listed so far, I'd like to hear that. So please share that with me because that would be good to see what those examples could be. So current robots in industry, we see here an overview. And in the bottom left, the traditional large industrial robot that can carry a car. That's a system that is not safe around people. It's not collaborating with people. It's even shielded from people. And so we're not going to focus on those type of robots. Next to it, you see a new version of that same theme, a system that is not targeted to have people interact with it. This is the new high picking systems for grocery stores. We have hundreds of robots driving over a grid some of the robots fill in the grid, certain buckets with toothpaste or other items and other robots would pick one or 2 out of these and fulfill orders for customers and put that in other buckets and then they're shipped out. There was an interesting presentation on these kind of systems, and they even argue that that's not really 100 robots. It's one big mind, one big robot or entity in the cloud that controls all these different sensors and motors. But the power levels of these motors that are used, they're better we operated as well are similar to the mobile robots that we will talk about further in this webinar. Further, you see here on the top, the 3 traditional robots that are used in various industries. And again, they don't have special considerations to be safe around people or autonomously drive. Obviously, it's not what they're doing. But the concept of autonomous mobile -- mobility and being a collaborate robot prehuman that gives the flexibility that you now see where this robots and autonomous mobile robots gain traction in the industry. So we're talking about these bottom 3 robots here. The cobots, collaborate robots, as well as autonomous mobile platforms that could be on wheels or walking and they have certain characteristics to be safe around people and be flexible in their behavior. So then here, a plethora of use cases. In the middle, you see a list of industries. We get warehouse, agricultural, medical, commercial, smart retail solutions. And then to the right of that, here you see various things these robots do. So they deliver the [indiscernible], notation, packing and inventory management and other controls. In this overview, I listed the systems -- the power levels from left to right, so to the right increase in their power level. And my focus would be for this webinar because we use specific power, DC-DC power solutions and motor control solutions at solutions, about 1 kilowatt of power to each motor and about 100 volts or less for the solution. So this would exclude the top tractor here and as well as this autonomous forklift solution. But On Semi does have a full set of products for higher power levels, which includes our silicon carbide MOSFETs, retraction inverters and as well as onboard charges and other solutions. So what do all these robots have in common? That's kind of the theme of this discussion, can we come to a common base that these robots need to perform or to be autonomously driving around a safely around people. And then in addition to that, they will perform some kind of task. So next slide. Here, I replaced a couple of robots with some others, and I highlighted this in yellow. So we see the 4-legged robot coming up. It has a use case where there's, of course, stairs or very uneven terrain. And this kind of highlights one of the good things about autonomous mobile robots in the industry is that we have more control over their environment. If they're in a warehouse, they're mostly on a level floor, lighting conditions can be controlled as well as the speed is way lower than in -- for automotive driving solutions. So we have some benefits there. Of course, quickly, then we use those to make solutions with cheaper compute or less compute power. And so that's then a drawback for the cost considerations. So let's look at the main adoption criteria for these type of robots in the industry. And the 3 main criterias I want to focus on is fast implementation these robots need to be ease of use, quick to set up and use intuitive -- with intuitive behavior. And this kind of comes down to the process of I show you what I want you to do. Maybe I grab anti-factor of the robot and set way points or follow me, I set waypoints and then I say to the robot to repeat. So it doesn't require any software programming. It should be an intuitive process. And of course, certainly to do that, you require sensors, you require video image sensors, LiDAR ultrasonic sensors and create sensor fusion techniques, and this will then enable you to control the software for this easy use of these robot interactions. Secondly, these robots have to be flexible. So the same type of robot or the same robot you have, you have to -- can do different applications or different -- is used in your shop at different times for different functions. So this is one of the differences I also want to highlight to the robots I showed in the beginning, the traditional industrial or robot this new hive system. When you create such a system, then you have to create your whole factory automation profile and plan out the whole activities of this. And when everything comes together, then you can operate. With these mobile robots, you could get one and maybe every Christmas or every new year, get another one and deploy it to where you want it to onboard it and use it. Another example there, what often here and uses that I think I could take the robot arm you see here in the image with the base and drive that to a CNC machine. And then teach it or train it to load and unload products, run it overnight and come back in the morning and push this robot out of the way of the CNC machine and then my assumption is my intuition is the CNC machine is safe for me to operate. So there cannot be any wireless communication between this robot and the CNC machine to start and stop it. This cobot anti-factor will get a 3D printed finger mounted on it and it will operate a CNC machine like everybody else in the shop. So pushing the start and stop buttons of the CNC machines so I have an intuitive behavior, a safe behavior that allows me not to use checkout and log out procedures and be more flexible in my setups. So for both fast implementation and flexibility requirements, of course, these add cost. And so another requirement is to be very cost efficient and have short payback scenarios. And today, we see those short payback scenarios because these robots can be used in the night shift or in a weekend to flexibly jump into packetizing pelletizing products or even allow us to run smaller production runs that are now profitable because you use that robot. And this will lead end user adoption, which I think is good because it also leads to technology advancements and add more sensors to get even more flexible ability in these robots systems and even faster implementation behavior. So it's a really a self-fulfilling prophecy in that way, making the robot smarter and then thinking back to the mirror, these boxes become very advanced and smart systems. So what have they all in common? And they all need to be safe and then around humans and drive around and then do some kind of path planning or people and obstacle avoidance. And so for the programmers almost, they have a slam simultaneous localization and mapping algorithm, which again relies on sensors that are in these systems. And of course, more obvious as well, they have batteries because they're all battery powered. And so they have battery recharging algorithms and solutions as well. So then finally, we will come to the ON Semi autonomous mobile robot demonstrated that we created. Here you see an explosion of that system. And what we did here is use existing development platforms we have for our individual products. And then combine these and into this platform that became an autonomous mobile robot that allows us to now look at interaction between these systems to do sensor fusion for instance, or other applications and/or advanced, motor drive solutions and other things. And it's a flexible thing. We use dim reels, we can easily upgrade this system and even look at the digital twin in the omni verse that I will look at and show at the end as well. So the system is in here, of course, there's a compute unit. We use the MVDR, Jetson Orin platform quite a powerful computer. But again, we were not cost estimate -- cost sensitive for that component because we do want to see what it is able to allow us to do. We have the Jetson nanos as well that operate the same system for now, but with sensor fusion applications and 360 cameras and other things, it -- you can quite use some power. So besides that compute power, all the other components are from ON Semi, and there is power and charging system communication system, motion, sensors and lighting. Then on the bottom here, you see the robot that we created. So each of these systems have more detailed products in them and list them out here. So for the motion system, we have a BLDC motor drive solution that's, of course, motor controller, gate drivers, MOSFETs, power regulators. I list here some of them in blue, and they come also back later on when I go to the individual platform. Then for power and charging, we do battery charging, battery monitoring, e-fuses, eco switches, turn monitor, DC-DC conversions. For lighting, we have LED control and drivers. Some of them are listed here. The compute unit, we talked about it and we at Jetson Oren, and we use a Ross Docker container with all our software in it, so we can easily move it from one platform to another and do a lot of experimental work integrating sensors and looking at drivers for sensors as well. Then for sensor solutions, of course, our image sensors LiDAR, rotary positioning sensor, ultrasonic sensor and communication, we have our 10-based TNS and BLE solution. So here, you see the -- we have a separate information on how we created this AMR. It's readily available. There's nothing in there that ON Semi -- we're not in the business of creating AMR. So we're just interested in how these systems interact and how we can make better products and solutions for our customers using at the overall system. And so we have the bottom with mechanicals, the picture in the bottom shows how we start. This is a frame you can order from a company and then have some dim reels in it and put our -- all our development platform in it. We demonstrated that this year on several different trade shows where above the AMR, there is a cobot robotic arm kind of hanging and it simulates one of those robots that has a cobot arm attached to it, but then the bottom section would drive under out of it revealing all the content of what's inside this base that makes it an autonomous mobile robot. So we have a landing page on our web page with block diagram going specifically to individual products from ON Semi. And then we have these development kits that I referenced in -- for each of these products, and those are the ones I want to go over at this point. So what brings to the AMR, we're going to add ultrasonic sensors. They were not in there until now. We integrate the and we'll have this demo early next year for CES. We'll have this system there. We integrate LiDAR, and there will be a different compute system from D3 that have NVIDIA Jetson carrier boards with our camera image sensors, which is power over correctional so you can do 360 camera views. In this system, we also have a rotational sensor positioning sensor solution. So this monitors the wheel location and that can be used in the motor drive algorithm to specifically without EMF sensor feedback, sensor-less motor control, but using this rotational sensor. And it can also be used in the cobot for the location of where the arm is in its space with different sensors. So I have a slide on that detailing more details in a minute. I want to jump to this bottom product from one of our customers that uses our sensor. They created a the 360 3D vision system with 1 camera. So normally, you need stereo vision or multiple cameras to look at an object to be able to calculate debt. And this company uses special mirror. It's, I would say, coffee filter upside down mirror and which has these facets all it. And if you hold an object in front of it, you see reflections on each of these bridges. And the camera picks that up, the camera looks up and there's another mirror on the top that then shows this picture. And with that, you get 3 or 4 angles, 3 or 4 different pictures to your destination object and you can make the calculations, as you know, the shape of this filter coffee filter mirror. And so you can do debt sensing with 1 camera, interesting product. They use our AR 135, which is a 30 megapixel camera, higher resolution to be able to pick up these or segment the image basically for each of these different pictures you have. Then we have our e-com USB 3.0 camera which is a global shutter camera, which is a good technology for autonomous mobile robots. Global shutter is when you take all the pixels in your image and process them in one go. So you need some buffer to store these and then offload them from the camera into your computer. The other approach is a rolling shutter, we do it per line, you have the smaller buffer requirements. And I -- each of them have some advantages that I'd like to illustrate now and for instance, rolling shutter tends to have higher dynamic range. So if you have a system where you have less control over your environment. So again, in the warehouse, we probably have control of the lighting in our intelligent retail solution, you have control of the lighting. If you go in agricultural or outside driving, you might not have that or you, for sure, don't have that. And some cameras that don't have a lot of contrast or a lot of dynamic range cannot resolve the big contrast in the image on the left. So when you look outside of the tunnel, you see bright light, cannot go and see what's outside there. And on the right, you see our solution where we can resolve that, we can look outside and see the object outside of the tunnel. So that can be very important for specific applications. So I want to mention that solution, which is the AR0822 and again, it's listed here, as you can see for machine vision, robotics and the other solutions. So then looking at sensors to do depth sensing, specifically for robotics applications. You can, of course, on the left here, we use stereo 2 cameras, vision system, and that gives you probably the highest resolution of separate images on separate cameras, but it does complicate sometimes how these are mounted and calibrated together to each other. And so there's some complexity around that, but a very good system to use. Then of course, there's LIDAR. We have time-of-flight solutions for direct and indirect time of life methods. And this gives us very good LiDAR feedback, and I'll show you some slides in the next slide some information on that. And then we have our ultrasonic solutions and that also can be used for this depth sensing, and we can combine these sensor fusion applications to get even higher resolutions. So the LiDAR applications. I just wanted to highlight is you can have 1D, 2D and 3D applications there. We see a lot of 2D applications in the warehouse where you have maybe an older robot driving around or be static and do some functions you want to create a safety zone around that. You can have a LiDAR like this mounted on the walls and people enter that area can detect that and then slow down the robot or even if they get too close stop it, so it makes it safer. The ones we're talking about in our autonomous model driving solutions are the 3D LiDARs and you see here also use case samples for that. So then the inductive additional rotational sensor solution we have, which is depicted here on the right. If I compare that to the 2 other main solutions used in at the moment, which is a magnetic encoder and optical encoder this solution, deductive encoder, has several good benefits for detecting rotational movements and I lists these here on top. So they have less -- lower sensitivity to vibration and contamination. There's a low component count only or have a slide on that robustness. So it's a magnetic fields are not influencing the results and temperature is also a very -- it's very stable over the temperature range and it's low weight. So if we look a little bit closer at that solution. You can see that it consists of 2 PCBs. One is the rotor PCB. It's kind of dark but they look from this side, almost the same. So let's look at the lighter one at the bottom. You see here a course and a fine inductive trace on the PCB. So the PCB has no components on this side for the rotor. It just has these inductive traces and that's it. Then we have a state PCB that has our solution mounted on the bottom here and that also has these inductive patents on it that then give us plus or minus 50 seconds accuracy, but if this is a 38-millimeter sensor. And it's an absolute encoder, so really important for robotic arms when you start up the system, you want to know exactly where the robot is. You don't want to spin the motor first a couple of times because then the robot is moving. So you want to keep and have this absolute value of how the rotation is and the position of the robot is. So we get up to 6,000 RPM, we get full accuracy of this 50 arc seconds, and then it will degrade, and we have all kind of information on that, but the sensor will work up to 45,000 RPMs. So all the very demanding applications can use this. Here's a the one-pager on that specific product. It's the NCS32100. Then we have the motion portion of the AMR. For this, we have 3 different motor boards. One is based on our ecoSpin, that's the ECS640A. It's an integrated motor solution with a controller with the gate drivers and A2Ds to completely drive MOSFETs on the outside. And we have this set up or all these 3 boards are set up in the same way. So we're running in our AMR 48-volt battery system. So we have these systems run at 24 folds. And so they all have -- they are configured in that range by using specific MOSFETs for gate drivers in the other 2 examples. So the other 2 examples are discrete solutions that use our discrete gate drivers and MOSFETs. And one platform has a very high advanced controller, sidings, module that you can use here and you can do additional FPGA system that you can use to make very advanced control algorithms. And you can think about if a robot drives up a hill, and it sees that from its video, so it's another way of sensor fusion or it goes down here or you hit a wall, you could get very high current inverses and you could mitigate that or do something special so your battery life is longer with an algorithm in the processor and this platform allows you to create that. And lastly, we have a platform that we use in power tools, but that's, again, giving us about 600 watts of power to a motor. And so we also use that in this motor drive in the AR mark, this uses our RSL10 and RSL15 Bluetooth chips as the controller and then a discrete outside gate driver MOSFETs. So in the power tool environment that allows us with this system to do Where is my Power Tool. It has a Bluetooth link. So it can be asset tracking solutions that we have implemented on that system as well as if you hit your drill start to drill your vacuum system automatically starts at that point. Because they are all battery operated, you want to minimize the time things run, we have longer efficiency use for these type of systems. So for Power Solutions, we are the AC/DC power supply that gives us full valves out, 600-watt for charging the system and a DC-DC part that we use for all, creating all the different voltage levels in the robot. So the MD system needs 20 volts, the motors we run at 24, the lighting system all can create several different power levels that you need within the system and then also have e-fuses to protect each of those voltage reals. So if you look at battery operation, the subsystem, then of course, you have to talk about battery life and at the end of the slides, I have a little bit of information on the NVIDIA Omniverse and having a digital twin and simulations. One of the simulations you can do is if you run in an environment, how long does your battery last and if that -- depending on that, you can do or implement something like opportunity charging, if that is giving you enough energy to get you through the shift of those robots. You can, of course, add more robots that are stand buy when others run out or in the bottom picture, you see here a solution where you change the battery between shifts and that's also done and then to deep charging of the batteries separately from your run cycles. But in general, a payload, cargo less velocity and distance driver you can have some kind of a formula where you can then simulate behavior and get really good numbers for estimation of how long your batteries should last and what kind of system you need. Then on our lining solution, we use our automotive taillight solution. This has -- is 3 lighting options, red light which is normally brake light. We use it for signaling the backside of the robot and white light for the front side and we have a turn sign indicators, so left, right and for this AMR. So this is signaling to its environment. You see this type of robot with these type of wheels, they can move sideways, they can move forward and backward. There's no real front or back to it. So we signal with the lighting, if it's white light, you know it's coming towards you. If it's a red light, it's driving away. And so it's for the environment signaling what robot is doing. The same for the RGB light driver, we included this is on the side of the robot and that has different colors. So when it's charging, it's blue. When it's active, the robot can be red or -- and if it's passive or off it can be green. So again, signaling to its environment. Now this solution also allows something which is called VLC, variable light control, which allows the robot to pick up again the light or dim it slightly, which is not visible by the human eye, but other robots can pick up on that and they can use that as another communication channel. And this is -- we have demos where we just send music over that type of link. So it's a nice additional feature that some of these line solutions provide. And then lastly, I have here the communication link. So there is a wireless unit in this robot, which is a telemetry sensor node based on the RSL10 which has various different sensors, air conditioning of air pollution, air quality, humidity, temperature, et cetera, move, acceleration. And this system is battery operated, very long battery life solution. And it's just stuck on the AMR. So there's no physical other power connections. The use case for that is also in your industry when you have machines that you don't want to power them down and go into the machine and add IoT features or Internet features or preemptive maintenance features and have downtime on your system and have PLC that you have the program and a C code of Phyton or whatever the machine is, how it is. Some even have not known electronics, an old cutting machine. This node can still provide information on that machine because it can sense, for instance, if the cutting blade gets old, and you can do some -- we have AI examples on this note as well. So temperature is well in here, so it can do preemptive maintenance on the robot itself. So you can see if there's elevated temperatures or other conditions in there. It can signal safety areas for humans. If you, for instance, have a gas leak or some other activity happening as well as other robots in its environment where you can get information from this node. And then we have a new -- the new standard based 10-based T1S with the NCN26010, which is a 2 wire solution for internal -- in the robot communication between the central processor and all the other units. So how that works is particularly in robotic arms, you have quite a distance to go and also in a larger robotic model platforms you have this distance. And this 10-based T1S is 2-wire Ethernet to the edge solutions. So it's full Ethernet, no conversions in the part, no gateways, no switches, 10 megabits per second. It's multi-drop so you can just have little steps to each of the different nodes you want to connect and it's single pair Ethernet. So a very exciting technology for hooking up these systems and making platforms out of them. So now I see -- I lost my -- oh, there it is. Next slide. So the standard provides 8 nodes at 25 meters, while in these robots we're talking about, we're not going to go 25 meters, so we can enhance the number of notes because is really on the signal-to-noise ratio that you create with the stops that you will have to adjust. And so our product also has a ENI function, enhanced noise immunity, which then if you want to go longer distances, if you 40 notes at 25 meters or even longer distances with reduced notes. The product brief for that product is listed here. And some of the application, this is used. We've talked about robotics but it's also in industrial cabinets, so fixed infrastructure where you replace all the different wires. It could be [indiscernible] and all these things with this one technology and then everything is Ethernet-based till the edge, and you see other examples as well. So finally, the Omniverse. So NVIDIA has created a very nice system where you can have your robot, your digital twin of your robot to see our robot inside a virtual environment, and you can create synthetic data in this environment. So obstacles or other things that the robot has to navigate. So you can train it on various different environments, and that's really good. The light bulb went on for me when I saw the example of this robot this 4-legged little robot here that they created that could not walk and they created a simulation environment, which stares up and stare down, spun 2,000 of these robots, let or run overnight and in the morning downloaded the algorithm it created into the robot and it walked, just stand up and could walk stairs and follow people. Another good example is the bottom, where you see in a virtual environment. There is an AMR underneath this pellet that drives that around. You see here what the image sensor or the camera on the AMR would see in this virtual environment. And it uses everything for the -- from the vision system, except for actually the image sensor because the data is virtually created. But this way, you can use your drivers that you use, if you have some parallax algorithm or you have debt measurements algorithms, you can all test those out in this virtual environment and do your path planning solution. So I think with that, I'm at the end of this overview presentation, and it's time for a break for myself and my friend here. Thank you so much.

Wonderful. All right. So thank you for the excellent presentation. We have received a number of questions, so we'll jump right in. If you would still like to submit any questions, just type them into the chat box -- but to get started, Theo I hope you enjoyed your break. All right. How important is camera 3D field of view with AMRs?

It's very important. And particularly if these devices can drive sideways with the wheels we are using. So it's depending on the motion as well that they can perform. And so that's why we have also multiple cameras to have 360 view, almost like your car, where you can just see surrounding around the robot. And for individual cameras, and it's important as well. What you have to consider then is that you have a calibration algorithm to correct for [ parallaxing ] if you want to look at, for instance, ArUco markers, so barcodes that they don't get to distorted by the lens that you use in front of your camera. But very important topic, and there's a lot of information on that, that we have that we did share.

Wonderful. All right. Next question. Just curious, is there a reason to use both LiDAR and ultrasonic sensors at the same time?

Yes. And image sensors as well. So this is what we call sensor fusion. So you get different inputs from different sensors. You combine these to even get a more accurate termination of what is in your surrounding and where the distance between your different objects, yes.

Wonderful. All right. This one is a couple of different questions. First one, what are the main challenges or problems you are facing with the current 2D or 3D vision sensor stack? Are you actively searching for new technologies to solve those corner cases? And lastly, have you heard of and considered using the light code photonics 3D direct time-of-flight camera?

So to the last point, no, I have not considered that. I would love to look at the different solutions that are out there. The system we use has, as mentioned, we create ROS drivers and then use the ROS operating, the robot operating system to create images that then are processed in further in our -- in the pipeline in ROS. I definitely like to invest that more and see more options than are -- what's out there and what is available for that. If I can relate that back to our product because at the end of the day, we're selling image sensors in that scenario. So I'd like to see if I can enhance our image sensors with specific features that I incorporate into the hardware and initially see that they are used everywhere in the software, for instance.

Wonderful. All right. And then we do have a quick question about being sent to the presentation. It is being recorded, and it will be on onsemi.com, and you'll receive an e-mail whenever that's up. Another question though is, can you share more about the types of edge AI model supported by your solution?

Yes. So in the slide where we use this RSL10 node, we are -- we partnered up with a company called Sentinel that has 2 kids to do a time domain AI protocols. So this is over temperature or acceleration over time, anything over time, so not special, not video and looking at objects in that because this is a smaller system. It's an ARM3 CPU and it creates then the learned behavior or what you want to detect with it and it creates the code in that system that you can download on our RSL10 in this case, we're looking at to announce that to our new RSL15 and other products as well.

Another question, what is the total weight of the robot, if you're able to get that?

Yes. So maybe one quick remark to the previous question. So in industry, the dominant applications we see for HAI is, for instance, if you have a big HVAC or fan system that has the filters in front of it, we would have an AI that looks at the vibrations of the fan. And there's a relationship that you can then back and say, okay, I got still 10% filter life or my bearings are running out. I do some preemptive maintenance solutions on that. So that's the prevailing current HAI solutions, I would think that I have seen. So the weight of the robot? I actually don't know the weight of the robot. I have it somewhere on paper because we shipped it around into different trade shows but I don't know at the top of my head. I do know structural integrity a little bit because when we created it, the thing was specification we said that it should be able to drive me around. So I should be able to sit on it and be driven around the office. That was the team's requirement for its structural integrity. So I'll look up to weight share that.

Wonderful. All right. And you did mention shipping. We do have one question sort of related, but where can I see the ON Semi AMR?

So the latest version, which has at least the updates that I talked about, we'll be at CES on January 9. So you can definitely see it there. And then during the year next year, there will be more opportunities to see it. It will also go to some of our distribution partners that I know of so far, but I don't know the details the dates as such. So maybe something we can share as well lookup and share.

I think that information should be on our website, we do have some an event section.

Okay. Good.

All right. And then how can I get a charger demo board for the AMR charger?

Yes. So there is I think overall, 3 ways that we have available our demonstrations larger demonstrations. For instance, like we have a total pulp PFC, AC/DC solution for onboard charging, which is 3-kilowatt and even an 11-kilowatt solution. Those are larger systems that are positioned in each region, and you can make arrangements with your sales force to have that demonstrated somewhere. This is the same with the AC/DC solutions because they don't -- they tend to have open exposed parts so they're less safe to just have available online that you can order them, which we have other platforms like that. So in this case, we have these available and you can connect your salesperson or a field application engineer and arrange for that.

All right. Questions. Can you repeat the name of the partner you worked with to deploy the Edge AI with the RSL10?

Yes, there was a SenseML so that Sense and ML, machine learning, SenseML.

All right. We've got a few more questions. Do you need dedicated firmware for using 10b T1S? Or can it run on a Linux system?

It can run on a Linux system. You don't need dedicated software for that. We have implementations with microcontroller and M3 arm core and 3, so microcontroller that runs a free OS and then in IP stack. We also have solutions where we show it connected to a Linux machine. So you can do various options there.

All right. How can I access the NVIDIA virtual simulation environment?

Okay. That's through NVIDIA, it's called the Omniverse and you can download it onto your machine if you have an NVIDIA graphics card, I believe that's the restriction. And then it has all kind of trainings and interesting information. They have 2 environments they promote one is ISEC, which is the industrial simulator, which has building blocks for warehouses and synthetic data. We talked about boxes, all that is available. And then they have an automotive drive simulator that does the same for automotive self-driving applications, where it has a road system and buildings and all kind of things. And they're really looking almost real if you work in those simulation environments. So yes, through the NVIDIA website, Omniverse, definitely something to do. It's free. So it's the best time. We live in all the schools are freely available.

Wonderful. All right. Is functional safety considered in your AVR?

No. This is a demonstration unit. We do have, as I mentioned, e-views. So some of our products that are in the system, but we don't have a lot of thought through this. We are more interested in functional behavior and how this interact, how different systems in the different systems interact and then into the microcontroller are available through drivers and other products, yes.

All right. And I think this is our last question, but is ON Semi providing the mechanical parts and 3D printing files for the AMR they created?

Yes. I mentioned that. So everything is -- they are available, if anybody is interested. We just don't have them readily. I can give you a link here, is where it is. So we might see how we can distribute that or make that available.

Beautiful. for you. All right. Well, it looks like those are all the questions that we have for the day. Thank you very much, Theo. And on behalf of ON Semi, I would like to thank everyone for attending, and I wish you a nice rest of your day.

Thank you. Thanks, everybody. Bye-bye.