 Hi, I'm Drew Festini. I'm one of the board members of the Beagle-Border-Dork Foundation and we're here at embedded world showing off the new Beagle-Bone AI. So if you're familiar with the Beagle-Bone, it's a single-board Linux computer. That's open-source hardware This is our latest board has a newer processor from from Texas Instruments, the AM5729 It has a dual-core ARM A15 and along with that has dual-core DSP and Also these embedded vision engines and that allows us to accelerate and offload Running a neural network from the ARM onto these accelerators from TI so we can do something like load a TensorFlow light model Onto it and then it'll accelerate that using the DSP and the embedded vision engines and we can run our application logic in Linux on the ARM We also if you I don't know if your viewers are familiar in addition to the ARM We have these real-time microcontrollers called PRU's for programmable real-time units And these are separate 200 megahertz microcontrollers that are on the same die And that allows us to do things that we need very low latency or timing critical task force the things like Controlling motor drivers or over here. We're we're controlling an LED panel so we're doing that by using the PRU to handle the timing sensitive parts like Controlling the LEDs in the matrix and then we're handling in higher-level things like running a web interface on the ARM in Linux So like this is a great example here with the pocket beagle we're doing a Panel here, so we're driving this Matrix of RGB LEDs We're controlling that from the PRU the PRU is handling refreshing the matrix And then meanwhile the ARM processor and the same SOC is handling running Linux and if we were to connect to it through Wi-Fi or USB we could Load up the web interface and change what's shown here So it's a good example how we can break up higher-level tasks like running a web server in Linux and lower level Latency sensitive tasks like driving a LED matrix Another good example is one of our Google summer of code students a few years ago made something called the ego logic So this is a 14 channel 100 mega sample logic analyzer And we're able to do that because the PRU has access to the pins at very low latency five nanosecond It can read a state of a pin and then on the Linux side We then grab that information out of memory and we can display it display the waveforms in a web browser So it's a good way that we can take Low latency tasks and put those on the PRU and we can take the more complex higher-level tasks and run them in Linux and When you talk about the embedded and embedded vision Parts right on the SOC. Yeah on the chip. How does that compare with some of the recent AI? Specific parts of the SOC's that people are doing is it similar like this? Yeah, it's similar So there's something called the TI Texas Instruments deep learning SDK Now allows us to take something like a tensor flow light model or an NVIDIA cafe model and load it onto the Beagle bone and then it will break that up and Accelerate the inference of the neural network across both the DSP and these embedded vision engines and the bed of vision engines essentially co-processors that allow us to accelerate vector operations So you maybe achieve similar performance to the I think the Jetson nano is one of the board So you can get similar performance in terms of accelerating the inference and offloading it from the arm And TI has been doing that for a long time doing all kinds of different parts on the SOC and trying to optimize like DSP for Very long time. Yeah doing it and there's all these different ways of doing it And when you talk about the Cortex M4, right? Yeah, is it also in the SOC or is it on the yeah So this the TI SOT the satara AM 5 7 2 9 It has a bunch of different cores in it. So in addition to the dual core arm a 15 We have two Cortex M series M4s that you can use for doing lower level Like you can see sensitive code and we also have four of these programmable real-time unit microcontrollers Those are each 200 megahertz. We also have the dual core DSP and the bit of it embedded vision engines And we can manage all this from the Linux running on the dual a 15's One of the other things that I think kind of differentiates the Beagle bone versus other single-board computers Is it's based on a Industrial SOC called the satara from TI So we have all the peripherals that you'd expect in a microcontroller platform eight channels of PWM six channels of ADC And the other thing I think is important is we're open source hardware Which means the the schematic and the board design are all open source including the bill materials All the parts are available in quantity one from distribution So if you're going to embed the Beagle bone in a product you don't have to worry about It being discontinued because even if it were you'd still be able to buy the parts and you still be able to manufacture the circuit board So you're gonna have confidence that this is a platform that you can build the system around you need support for Ten years or more. So you have what do you call those people doing clones? Yeah, we do Both ones that are branded as Beagle board org and other ones that People anyone that wants to it's it's a open source so they can take the designs and make their own If they want to call it a Beagle bone, we have a program where we have a couple different programs One is called Beagle board or compatible. So seed studio does that seed studio has the Beagle bone green Which they decided to take the HDMI off to save costs. So it's a lower cost version And they call that the seed studio Beagle bone green And then other times people will just take the designs and make their own derivatives like Autodesk took the design of the Beagle bone black and used it to base the controller for one of their 3d printers on So there's a lot of flexibility there. There's also these two headers so these two expansion headers expose a lot of the IO and We can then have expansion boards, which we call capes So we have several different capes that people can buy to add functionality in addition You can design your own capes as well to add additional functionality to the board Maybe external ADC or external DAX stepper drivers that you want to have integrated into the system So is this also integrated using the Octavo systems SIP? This one is not so you can't see underneath it But underneath the heatsink. It's just a TI package for the AM 5729 However over here in the pocket Beagle we're using the Octavo system system and package So this has all of the chips that were in the Beagle bone black and Octavo Which is a company in Texas took the individual die of the of the sitara SOC and the Power management IC and the DRAM and also a bunch of the passive so on the back of the Beagle bone black so here for example Beagle bone black here We have a bunch of separate chips. So all these chips are actually inside this Octavo system and package So this is the SOC so we have the bear die of the SOC power management IC also is included and Also all of these passives so we have lots and lots of passives on the back So like 140 passives are also inside this package. So the great thing about that is this pocket Beagle Which is what this board is here on on this shio on this cape So it's just a four-layer circuit board So it's great for people who want to customize and make their own boards based on this anyone using ego or a key Cad we have the design available and both can take this design and modify it and add on the other boards that they other Is it easy for them to to do the four layer? Yeah, so one of the nice things about Now that all the high-speed things like the DDR memory in the SOC are inside this one package The rest of the board is relatively simple So you can take this design and that's so you want to have stepper drivers or something like that You can make up for the integrated that all together Or you can use the headers and have it plug into a expansion board, which we call capes So actually one person the person that did that beagle beagle logic I told you about project. He ended up making a variation of the pocket beagle Called the beagle logic which added on probes for the logic analyzer All right, what were you doing mostly this during this embedded world? I've been talking to a lot of people about the projects they're making one of the one of the engineers It's not by today was interesting to hear about their product, which they do a welding monitoring system using the beagle bones So using different sensors They're able to monitor the weld to make sure that it's a good welding joints being formed And they're using the beagle bone to do that also talk to students that we're building robots at university So it's so we're always excited to see what people are using the beagle bone for in their projects So it's a bunch of years now already the beagle board. So we started off with the beagle board in 2008 So this was using an omap processor from TI So that the project started with two TI 4410. I believe so. Yeah Same there wasn't an archive since death. Yes. So this is actually the beagle board XM Which I think was the second version. This one was the DM 3730 I believe So we started off with this in 2008 and then the beagle bone white came out of 2011 This board actually came out in 2013 the beagle bone black It's still really popular because many people have designed products based around it So we still see a lot of beagle bone blacks being sold despite the fact that it's been out for Over seven years now, I guess How would you define the difference between what you do and what some raspberry pi people do or some other? I think the key is that the beagle bone is a really great platform that people for people that are interested in industrial applications and Traditional embedded applications. There's a lot of single board computers out there now due to the proliferation of Socs for things like tablets and phones and media boxes. One of the things I think is special is that the TI Soc is actually built for industrial uses. I think the other key thing is that all the parts and let me come this way It's fun to see. I think that's another important thing Because there's a lot of really fancy arm out so sees out there now like octo core Massive ones, but they aren't usually available with industrial life cycles So all the parts that we use in the beagle bone are available with ten-year part-life Availability so while you might be able to system around one of these SoCs that are used in a tablet or a phone or in a set top box You might not have the same confidence level that you do with the components that we use in here That you'll be able to source them in the future So I think the beagle bone is great for people that are prototyping something that might end up becoming a product because you can have confidence they'll be able to source and build those boards in the future and TI is doing a lot of enterprise Embedded stuff. No Yeah, I mean so the parts that we're using kind of come out of TI's portfolio so that beagle bone AI Automotive. Yeah, so the the beagle bone AI which we had earlier So this chip is was designed for the automotive market. So things like saw self-driving cars In car entertainment systems, so that's why it has things like the embedded vision engine the acceleration for running neural networks So we're kind of following their roadmap of you know for us It's great because people are building rovers that can self-driving rovers. There's this community called DIY Robo cars Where people are building RC scale autonomous vehicles, and this is a good platform for building one of those DIY Robo cars. Cool. We have the rover and the other video was one around here. Cool. All right So checking out what's next is there like something announced for next what's happening? One of the things that we're going right now is the idea of a microcontroller based board that would act as a remote sensor board So this idea we've been working on for a few months now the idea of having a Board that's running in microcontroller with a Bluetooth interface that could then send sensor data back to a Linux board like a beagle bone In one of the ways that we're thinking about doing that is by leveraging some of the subsystems that already exist in Linux So we can we can essentially see the sensors that are plugged into the remote microcontroller board As part of the Linux system. So it's something that we're working on right now on the microcontroller board We're using Zephyr So we're implementing some things in Zephyr that will allow us to communicate with the Linux kernel That's running on the beagle bone That's on that board. It's not on this board It's another board that we're working on that would be a small inexpensive maybe ten dollar sensor board They would talk over a wireless link back to you a beagle bone Hey, hi, I'm Kathy. I'm with beagleboard.org. Welcome. We're excited to be here at embedded world 2020 And today I'm here to talk to you about beagleboard.org. We're a completely open source hardware open source software Foundation we're a non-profit that supports the education of embedded Linux. We started with beagleboard You know, it's a 1 gigahertz Linux PC. It's based upon the AM Sitara from Texas Instruments and you know runs low power and we have you know Millions boards out there including the beagle bones. We've developed to get better access to the IO This is the beagle bone black as you can see the beagle bone black has Plenty of access to the IO. It's small. This one has ethernet You can do a lot of projects with communication moving control and then last year we introduced The pocket beagle pocket beagle is the same performance as beagle bone black but comes in this small mini tin mini mint tin size and I'm showing it here with our tech lab learner Cape where you can go around and learn all the different parts of the chip by utilizing the IO and so you can learn the PRU which is our programmable real-time unit by using the 7-segment LED You can learn the PWM pins by using the RGB and then also we have this nice add-on in the back here from microelectronica You can use many of their click boards to add Different sensors and actuators to learn different projects once you've done some learning like this Maybe you develop your own Cape like this community member developed this scroller Cape He is really into Christmas decoration and so his Desire is to do lots and lots of LED Display so he designed this to meet our Cape or add-on board Application for the pocket beagle lots of fun and you can make your own as well because everything we do is open all of our Designs, it's very easy once you've developed with Beagle board to get going. We've got really great Debbie and distribution of Linux that we have you download off of our website. There's libraries and today we are working with the New Beaglebone AI just introduced last year and we have a fabulous group of developers already working with the Beaglebone AI it is a Mechanically compatible with all of our Beagle boards. So you've got the same header Capability and it matches with all of our capes and then again, we've got ethernet we've got the Wi-Fi and Bluetooth and We've got a fabulous chip on the inside of it that has the SOC from Texas Instruments the AM 5 7 That chip is this demo is running over here so this is our AI demo and it's using the The Beaglebone AI which utilizes the AM 5 7 2 9 as you can see it has a multiple of cores We use the embedded vision engines to do the compare on the TIDL machine learning library And also we are using the PR use to do the image capture So what you can see here in our demonstration is we're running a full Linux distro on the Beaglebone AI We are capturing the image the first image that you'll see is replacing What might be in a machine vision application is a good part That's the Beagle so the Beagle dog is the good part it places down in front of the camera and it will recognize the image just a simple image off the internet Runs through the millions of images on the TIDL library. It's a deep deep learning library from Texas Instruments and then it will come around The arm will pick up the device and put it into the good bin And then again when we put down the Fruit which we've chosen here a small berry you put down the fruit in front of it It again will determine that that is not a Beagle and so it will put it into the bad bin So we're also streaming the information here on the display. It's all running over ethernet and 100% supported by the Beaglebone AI. We don't need a connection to the cloud or any other Processes we are very excited to Support the open-source community. We've been longtime members 13 years with open-source Linux and Have a lot of really solid professional membership Our focus is on education embedded Linux education Professionals or students. I'm really proud to have education day here at embedded world 2020 And talk with the students about the Beagles. So when we take the Beagle The AI the new one which is right here. Yeah, the white one. Uh-huh. Yeah, where'd he go? He's over here He's over there. I'm a little repetitive. Yeah, so what's This this chip has AI capability Yes So this is the AM 5729 from Texas Instruments. It has the dual Cortex a 15 microprocessor It has the embedded vision engine also two DSPs for programmable real-time units a GPU It's got a graphics accelerator a lot of performance and capabilities. You do a full AI implementation and here you have some Some different promotions here. I mean you talking about the different solutions here at the abandoned world exactly exactly It's $25 it starts $25 you can get a pocket Beagle, which is a great learning platform Same performance 1 gigahertz Linux PC a lot of different projects on going with this everything from an IOT and application to perhaps even your IOT gateway or even a simple sensor sending things to the nodes The pocket Beagle is a great opportunity to not only learn embedded Linux Through the Linux foundation it is their learning platform, but also to implement any projects commercial or educational and In the beginning of the video we had this car right here behind So what is this project? Oh, this is a fabulous project This is actually a collaboration between the University in Texas Texas A&M University and NASA This will be a moon rover that is going to go and collect more rock samples from the moon what they've discovered is is that we need to get more rock samples and so they're utilizing our Beagle bone blue which is customized towards robotics applications and They've made their own unique Suspension system so that this will be able to Run over the moon rocks now They're also using it for education in the freshman year at Texas A&M University teams 3d print and build their robot and program their device and that way they're learning electrical and computer engineering right off the bat All right, and I guess this loss of other cool Projects out there plenty of cool projects. How many how many students working with the boards there are Hundreds of students thousands of students per year actually working with Beagle boards and projects University of California, San Diego works with them in their mechanical engineering department. We've got projects everywhere from India I was learning about a project today in Austria So globally we have people working with our projects in both industry and in education. So getting to Design robots for the moon. Maybe some robots for Mars all kinds of all kinds of fun. I I wake up Later this year. I wake up every day with a new project being Exciting for me later this year. We're going to have Portland State University launching a keepsat and they utilize the pocket vehicle for their project, but You need to be radiation proof when you go into space You need to have all these kind of like you think that will work in space So one of the good things is is that you know our devices robust industrial temperature and then most of the time They in encapsulate the things to help it go into space and they work with NASA for that opportunity So there's a lot going on with collaboration with NASA and education space beagles It's gonna be out there all right and lots of lots of lots of all the projects and people can find them on the social media Where you post all the quick things at bigelboard.org if you put in slash P you go straight to our project page We love to promote projects from our community. So please let us know when you're loading up a project on the website We have University students we have a student as young as nine years old who's done a project and of course professional developers as well