Episode Transcript
[00:00:00] Speaker A: We saw that there was a core that could be made into a real self standing product.
[00:00:05] Speaker B: The AI model we're dealing with are mostly now zero training or zero shot, which means that from an operator perspective, the ability to deploy or redeploy is very facilitated as its leveraging model has been trained on thousands, thousands of shapes in SKU already.
[00:00:28] Speaker C: Hello everyone and welcome to the Robot Industry podcast. I'm Jim Beretta, I'm your host and for today's edition of the podcast I am joined with my friends from Vention out of Montreal. Although they're not in Montreal today, they're on west coast. So I've got Etienne Lacroix and Francois Jiguer joining me live. And so welcome to the podcast, gentlemen.
[00:00:46] Speaker B: Thank you Jim for having us today.
[00:00:48] Speaker A: Hi Jim, thank you for having us as well.
[00:00:50] Speaker C: There's probably not very many people in the robot industry, Francois, that knows Vention. Can you tell me a little bit about who Vention is?
[00:00:58] Speaker A: So we are basically a LEGO meets IKEA meets SolidWorks turbocharged environment to build industrial machines. So we build a 3D CAD, which was our first product called Machine Builder.
You can access it on your, in your web browser and in there you, you find basically a bunch of compatible components to build industrial machines. It's our components, our LEGO blocks. You can put them together and you know pretty much infinite ways to create the machine you want. So it can be material handling, pick and place with robots, it can be multi axis gantries, all kinds of different machines. People build them on our website. And the beautiful part with Vention is all the components are compatible. You can also simulate and program on our platform.
Once everything is to your liking, in our 3D environment, you can actually get all the parts shipped to you. Next day you assemble them just like when you were a kid with your LEGO blocks, and then you get a fully functional machine.
[00:01:55] Speaker C: Francois and Etienne, one of the things I really like about Vention is the fact that your graphical user interface is exceptional. Like it's the best in the industry. So if I'm an applications engineer, and I was one a long time ago, put up the conveyor system, I can put in the collaborative robots, I can put in the press, I can put in all these, some of the things you may not actually apply, but I can put all these things in and I'll know that they're all there and I can look at the bill of materials and I can order it. Is that a fair assumption, Etienne?
[00:02:22] Speaker B: That's true. We like to say if it's available it's compatible. So we're trying to make industrial automation accessible to a much wider audience of manufacturer versus the traditional approach. And to do that we need to a lot of simplicity. So we're working in this environment where all the LEGO bricks are compatible with one another. So now somebody who's maybe has strong process knowledge, who's a manufacturing engineer, but not necessarily a control robotic expert can still with this simplistic user interface, design those machine end to end and take investment decision afterward.
[00:02:58] Speaker C: Now give me a little bit about Etienne. Still give me a little bit of information about Vention from it. You've got thousands of customers, you've got hundreds of employees, you've got locations in Canada, you've got locations in Europe. Can you just give me a little bit of a 411 for those people that aren't keeping good track?
[00:03:13] Speaker B: Vention has a great trajectory over the last few years and it's really fueled by the amount of companies seeking to automate. They're facing labor shortages, they're facing cost pressure, and they want to find alternative ways to automate at a much lower cost profile than traditional technology. Today, invention is present roughly in 4,000 factories, different site. Right. It's around every month it's around 6,000 users that come on the platform to design those equipment. We serve around 25 discrete manufacturing industries, from consumer electronics to automotive to defense to sports good to medical devices. A lot of those companies are using Vention on their manufacturing automation side to solve their own manufacturing problems. Vention is around 320 employees and we serve our North American customer from Montreal, Canada and our European customer from Berlin in Germany.
[00:04:06] Speaker C: That's great. And you were actually, I was doing a little bit of research just before I came on the call. And you're supplying a lot of collaborative robots, like a lot of cobots. Right.
[00:04:15] Speaker B: I think we became the third biggest top three player in the in the United States. Now cobots are by nature easier than industrial robots, but still in a robot cell, there's a lot of element that gravitate around the robot. There's an infeed, there's an outfeed. There may be sensors, there's controllers, seven tack systematic system. And putting all those things around the robot is quite complex.
You know, I like to say, you might have heard that from me, Jim, but I like to say there's nothing more manual than industrial automation. And you know, so the robot is one thing, it's often a very important thing. There's so many other subsystem that needs to gravitate around the robot and obviously Vention makes that very easy. And as a result the, you know, the, the, you know, Vention has become a big source of robot in Europe and in, in the United States across our various partners that we support, including Universal Robot, Fenic and abb.
[00:05:07] Speaker C: So I was in Informa trade show, the MDM show, the ATX whatever, whatever particular brand we're calling it. I was chatting with their staff and I happened to look at some of the AI robot motion controllers. And who wants to support this question? Is that Francois or is that Etienne on what's happening in AI?
[00:05:25] Speaker A: Well, there's, there's a couple different things we do with AI and you know, we understand it's becoming a super popular word so we're actually very cautious because we also get worked up when people use it the wrong way.
So there's two main things we're doing right now with the AI controller that we call Machine Motion AI. The first one is to do accelerated motion planning. What we do in our stack is to use external motion planners. So basically the motion planner is not inside the robot. So we don't send commands to the robot. We actually stream a set of points because that's the most universal interface you can get with a robot. So that's what's the most portable for us and the most agnostic. Now that being said, if you do all the motion planning in your actual controller, it can take quite a bit of time for long moves for applications like sanding or path following. And there are new ways to accelerate motion planning using GPUs and using AI algorithms to make them faster. So that's the first thing we do. And there's technologies from Nvidia that do that as well as others. So that's the first section of the AI component. The second is pose estimation and also vision applications where we will use accelerated hardware, but also AI workflows to do object detection, pose estimation and vision systems. And this is what we'll showcase at gtc. And why did we brand it Machine Motion AI? I mean, we're using an Nvidia Jetson system on Chip inside and Orin NX16. So we thought that, you know, it was the natural progression of the system to be a little bit more AI centric. And we also see with the emergence of skills based programming where you focus more on skills and a little bit less on discrete operations, that this trend is going to continue to help us make it easier to program the system. So that's, that's kind of the thinking.
[00:07:23] Speaker B: Around AI maybe it's worth even mentioning the why do we have Enchantment's own motion controller to start with? Because there's a lot of great product out there. So why another controller was needed in the market.
And for us that story started with the very founding story of engine where we wanted not control expert to be able to deploy by themselves. And today we when we're looking at automated equipment or an automated robot cell we have those gray automation enclosure and inside you'll find a PLC's, you know, power supplies, drive safety system and all that is quite complex. To just piecemeal those enclosure you will add probably three weeks of project time and tens of thousands of dollars of costs to a project. And we've asked ourselves can this be Lego? Ify to go back to Francois analogy, can we have a box that fit them all? And that gave you know, the emergence to our first motion controller which back then was very simplistic but still deliver the main element of value which is plug and play with all the devices inside the robot cell. So it's only a connector based wire. You don't need the concept of pigtail wire and terminal block is forbidden adventure. Right. So everything is connected.
You know this was present you can program with traditional tool like from a, from a software developer perspective which means not ladder logic, which means Python. Right. So all those things were in the first version. Then we currently we're still, you know heavily relying on the second generation motion controller that we've introduced in 2021 and now we're slowly migrating to this third version which bring full fledged AI capabilities to run inference pipeline and to run accelerated motion planning.
[00:09:05] Speaker C: Etan. You're on your way and thank you for that. You're on your way to or you're actually on west coast. You're talking to Nvidia today and talking about some of the case stories you're doing. One of the ones you mentioned was a UK firm.
[00:09:18] Speaker B: Yeah. You know McAlpine UK is a century old companies that manufacture a lot of household plumbing products that we're used to in the United States. The equivalent will might be Kohler for those that know that that plumbing company. Well and like a lot of those companies they have a lot of tasks that are fixed operator position. So think about an opera human operator in front of a machine doing a highly repetitive task that involves some level of unstructured grasping and placement.
And those jobs are extremely hard to fill and extremely hard to retain. So to some extent even the salary cost of the operator is you know A fraction of what the whole cost is because there's so much replacement needed in those, in those jobs. So we've been working with them since the summer to come up with an autonomous robotic solution to realize that task for 2 of their product. And we're very happy to be at Nvidia GTC this week to showcase the technology that emerged from that collaboration.
So you know, as a company we always prefer developing great technology rooted in real client problem.
And that was the reason McAlpine joined us.
[00:10:33] Speaker C: No, that's exciting for you. And Francois, you've just recently as well released Cabinet a released product for the cabinet making industry. And can you tell our audience a little bit about that?
[00:10:47] Speaker A: We like to build products that are a market pull right when the market pulls on us. Had a lot of customers coming to us with sending needs a little bit the same story.
Operators that will be sending cabinets day long will make mistakes, get tired and don't necessarily have the throughput necessary. And also it's hard to find these people that stick around and don't rotate.
So yeah, we had a lot of people coming to us with sending needs, so we built the product for them. Initially we created sales solely using the Vention platform. So basically nothing special, our standard components put solutions together. But we saw that there was a core that could be modularized, that could be made into a real self standing product. So we partner with 3M, which is a great customer of ours and also have an abrasives division. So we partnered with them to fine tune all the abrasive parameters and also to make sure that the product would suit our customers need. And then we released the sanding product that you've seen recently.
[00:11:51] Speaker B: In both of those cases, the project we've done with McAlpine, that is the technology we're showcasing here at GTC or Descending Cell. What's very interesting is it's a clear transition to autonomous robotics. And by that I mean there's very little setup time from SKU to Q to Skew. The model that we're dealing with, the AI model we're dealing with are mostly now zero training or zero shot, which means that from an operator perspective the ability to deploy or redeploy is very facilitated as its leveraging model has been trained on thousands, thousands of shapes and skew already.
And with the compute stack that we're now having with machine motion AI you get to inference cycle time of just a few seconds of compute. And you know, if you rewind back, Jim and I know you to vision in the early 2000, you know, there was 10 to 15 second worth of compute time every time you wanted to do a move. And it's just not compatible with the pace of manufacturing. Right. If you want to be able to have a product that can replace those hard to fill fixed operator position, you need to be in the 2, 3, 4 second words of compute time. So you can really move at the pace of human. And because those cells have been designed roughly for that velocity. So we're entering that realm of, of, of, of reliability, of speed, of functionality and I think what we'll see here at GTC and, but also with the robotics, such as the beginning of autonomous robotic cell and that, that's very exciting because it means more simplicity for more.
[00:13:24] Speaker C: Manufacturers and quicker time to market. Right. And I love this zero training idea because it's such a big in impediment to automation.
[00:13:35] Speaker B: It is, it is.
[00:13:36] Speaker C: Do you see yourself, Francois, doing more and more work with Nvidia?
[00:13:42] Speaker A: Well, the, the release technology at a very fast pace and I mean like they have the capabilities, they have the means, they have the engineers.
So yeah, for now we, we're mostly working on pose estimation and accelerated motion planning. But I was yesterday with the, the group here at the Nvidia headquarter, we had a partner day for the robotics and edge computing team and it was exciting. There's a lot of good stuff coming up every year. They're not slowing down. So I see, I see a lot more we can do.
[00:14:16] Speaker C: Yeah, they're operating at the, at the, at the speed of vention.
[00:14:20] Speaker A: Yeah, the Jensen speed.
[00:14:23] Speaker C: So, so we're really here to talk about palletizing. Right. And I know it's this, is this podcast won't come out until a few weeks. So can you tell us a little bit about what's happening with palletizing invention?
[00:14:36] Speaker B: Palletizing is an interesting use cases because it's probably the one that saw the most competition, the most quickly in the realm of productization. Just rewind back five years ago, every palletizer was an orphan machine with orphan code and adding a sku. You had to call back your system integrators because you had to navigate pretty complex interfaces both on the PLC side and on the robot side. And what has happened with the productization? The platforming of automation and innovation is probably the poster child of platforming and productization. But we've came up with fixed product. Right. And when you start to deliver products to client as opposed to project, you change the dynamic a little bit. First of all, you know what we saw last year is enterprise client that we serve and we're very privileged. We have the chance to work with some pretty large name. They understood now that they no longer have to buy project, they can buy product. That means as an enterprise buyer, you have scale, right? You can buy a certain amount of unit that will be the same. The training for the operator will be the same, the maintenance plan will be the same, the reliability, performance will be the same. It changed everything for them, providing them significantly better economic value than the project based approach to automation. And that has fueled a lot of growth specifically for Vention in that, in that segment of productized application. And what that means we continue to invest more and more in those product. I think our Palletizer product line is now at version 4.5. So that must be 40 release that have been done since 2021.
So it's a very stable, very mature product. As we embed more and more AI technology, what we see is those machine get more and more connected, more and more connected to run more complex inference pipeline, connected to provide better analytics, better alert, better video recording. So we see all those things making their way into product, a base product. I think the main winner here are the client themselves and they get more feature, more functionality, lower price. And the manufacturers of those get the benefit of having a reputable products that has more consistency than a and more scalable than a project by project approach to automation.
[00:16:52] Speaker C: Thank you for that. And Francois, what about data? One of the things that I love about Vention and what's so interesting is you've got so many clients in so many different industries, you're actually able to see a lot for not on both just the ordering side, but now data side of the systems that you put in. So I'm wondering a little bit about is data a product?
[00:17:15] Speaker A: Yeah. So you probably have seen already our machine analytics product, remote view and remote support product. So this is really the backbone for how we channel data from our customers machine to the Vention infrastructure. So we basically give access to different set of parameters and data that customers elect to see. And it gets displayed on our dashboard that we call machine analytics in the client portal online on Vention IO. There's more that we want to do there. We want to make this a little bit richer. We want to give different views that are going to be more flexible.
What's great with Vention is because everything is connected to a single controller and everything is one product makes it very easy to get the data right. That can give you the literally the temperature of A transistor and a drive all the way to the number of pallets. A palletizer has palletized in a day. So it gives you a lot of rich data and customers can elect to. Of course, transistor temperature is not super useful for customers, but it's just to give an idea of how much richness we can have in the data. And we, our customers, select some of the data they want to see, they bring up the machine analytics and then they can see what's important for them.
[00:18:29] Speaker B: I think what we saw in the fast space of productization of Palletizer will now happen in a suite of other applications that are also productizable in the robotics world. Right. So I think the, this impellatizer got productized in roughly three years. You can imagine how fast we'll go for the next series.
[00:18:49] Speaker A: We serve very big customers and we also serve pretty small customers. And it very often happens that Vention is the first automated machine in a particular operation. Manufacturing operations are very often a flow. So you'll have in the flow somewhere just one connected Vention machine and it magically becomes the eye of that company on the factory floor because of the nature of the first connected machine. So people will, you know, start to follow, you know, what's been produced in a day, the throughput of the line with these connectivity products.
[00:19:21] Speaker C: So have we forgotten to talk about anything today? Is there any other announcements coming out that we could know about?
[00:19:27] Speaker A: Well, we haven't talked very much about machine Motion AI and what it is, so I'm happy to give more details there.
[00:19:33] Speaker C: Yeah. So how do I buy this or how, how does it affect me or do I have to buy a special machine for, for, for this?
[00:19:40] Speaker A: Well, machine Motion AI is, you know, the core of a Vention automated system or robotic system. So you can't really just buy machine motion on its own, really. It needs to be connected to our two Vention products. So actuators, sensors, pneumatics and other types of devices.
And this is where we run our core software, our control software. So it's not a standalone product. But you know, as soon as you elect to choose Vention as a platform to automate you, naturally you will have to purchase a machine Motion AI as part of your bill of materials.
[00:20:12] Speaker C: Let's say I'm building an assembly system to make a medical device, the machine Motion AI that will help me in the logistics in mapping the robot motion, for example. Right, yeah.
[00:20:25] Speaker A: So what you'll do is you'll go inside our machine builder ecosystem. You'll start to put your machine together, maybe interact with some of our application engineers to get some advice. And then eventually you get to all right, I have the mechanics of my system. I have to start automating it. So you'll choose some motors for your conveyors for your parts inflow for your part, outflow. You'll choose maybe some pneumatics to index your parts and additionally maybe a camera to do detection to make sure your parts are located in the right place.
And then all of this, you know, kind of magically assumes that there's a machine motion controller in the background. And the great thing is you don't really necessarily have to choose. Is it an NPN sensor, a PNP sensor? Do I need a gigabit ethernet port? Do I need an ether cat drop? You don't have to take care of any of these details. Everything you have from a control standpoint in our tool will connect to machine motion AI and will be interfaced appropriately.
So it's really the backbone of everything that gives you motion and controls in our platform.
[00:21:26] Speaker C: Thank you for that, Francois. Last question for Etienne. You have this really interesting view on what's happening in manufacturing around the world, especially in North America. Do you see any trends that are kind of like a lot of small people, a lot of big people, or is there any things that you could maybe note to the audience?
[00:21:43] Speaker A: Everybody's using Vention.
[00:21:45] Speaker C: That's great.
[00:21:48] Speaker B: I think there's a couple of things that are continue to be interesting. Right. You know, Jim, the last two years have been more difficult for the industry in general and versus going back to 2022.
But despite that, the fundamental trend for automations are better than ever. So there's a slight timing disconnect in the market, but things will re bridge themselves with some of the policies being put in place in the United States. Labor shortage will be problematic. Reshoring will be key and automation will have to play a role.
It brings back, going back to the fundamental and the long term trend. Automation is poised to be growing rapidly. And we know, everybody knows we're still at the beginning of adoption. Right. You look at the density of robot in the United States versus perhaps Germany or Southeast Asia, we have a lot of room to grow that. So it's a very interesting market to get into. What's interesting is really AI coming in that's going to bring those, you know, robot cells to an autonomous state and it will take years. Like I'm not saying this will happen in one year or two. It will probably be a five to ten year journey. But the, the step have already in place.
And I think if you think about this as if you're a machine builder or you're from the industry, you look at this is you can ask yourself what will it change to my typical robot cell design? And there's a few things I think are worth mentioning.
In a world where robot arms are continue, price continues to decrease, to become more and more ubiquitous. You'll see simply more design where people decide to, let's overshoot a little bit, put a robot arm instead of an actuator system.
It's a little bit more pricey, but I save so much stuff, programming and so on that it makes sense. Right? So we'll see that. We'll see a robot cell where there's hundreds of sensors today being replaced maybe with a few camera. Today with some of the AI model coming up, you no longer need to buy a $10,000 camera. You can go around with a $600 camera, right? And, and instead of having hundreds of sensors, you'll have 2,3 camera to orchestrate the entire cell. I think we'll see also PLCs have been the centerpiece of a lot of those equipment in the past and they will continue to play a role. But I think you will see a shift towards, you know, PLC becoming GPUs because you're going to need to be able to rogue those AI workload on the edge and, and the role of PLC may slightly diminish inside, inside the robot cell.
And with all of that, I think what's interesting is just jigs and fixtures, just presentation in general will become less and less of a problem because robots are gaining grasping capabilities. So if you think about the robot cells of tomorrow versus the robot cell today, it's not the same design. It's going to be easier to design, it's going to be faster to design and you're going to rely more on software, less on mechanical devices and they're going to become more agile, more flexible. And as a result cost of automation will continue to go down. To me, one of the big unlock I'd like to see in the industry is a solution where client can consistently achieve below one year payback. Because if you have below one year payback, you're flying under the fiscal year's budget and that's where you'll see the unlock software does that today. You can have a lot of software solutions providing below one year payback. But automation and I think Vention is the platform that provide the most economic value for Our customers and on average across all the portfolio of project we do, we are at 1.3 year. Right. So but I rewind back, you know, just three years ago I was probably around 1.8, so it's going down every year. But I think AI will be the last push we needed to really cross over that below one year payback. And the shifting in architecture of those robot cells will help get there. They will be just less pricey than today's cell as a result.
[00:25:38] Speaker C: No, that's great and that's interesting to hear because I've come from the two year. Most of my work's always been, well, it's payback's 2.1 years or whatever. So that's great to hear. Is there anything. So if somebody's out here listening, they're like, I didn't even know about Vention.
What steps, Francois, would you tell them to take to maybe talk to an applications engineer or just get up and running?
[00:26:00] Speaker A: There's a lot of fantastic content on our website, obviously. So visiting Vention IO and looking at our, our reference designs, our application, our design library is a great place to start because we've decided to build our platform to be fully client side driven. Everything runs in the Browser for a 3D environment or simulation or programming tools. People can go on the website and open a session, Design something in 3D, simulate it and play with the tool absolutely free, no download. So it's very straightforward and like you mentioned, we're always happy to talk to customers and give them a hand to get started so they can absolutely reach out to an application engineer. We'll get support much faster than they think they will. So these would be my next steps.
[00:26:44] Speaker C: Well, that's great. Well, good luck at the conference, gentlemen. Thank you Etienne, and thank you Francois for joining me today. And I'll put some of the notes, I'll put some of the notes in the podcast edition notes. And we'll look forward to seeing you both again soon.
[00:26:58] Speaker B: Thanks Jim for having us again.
[00:27:00] Speaker A: We were not too hard on Jeremy, so he's going to have an easy one.
[00:27:04] Speaker C: That's great. Thanks very much.
[00:27:06] Speaker A: Thank you, Jim.
[00:27:07] Speaker C: Our sponsor for this episode is Earhart Automation Systems. Earhart builds and commissions turnkey solutions for their worldwide clients. With over 80 years of precision manufacturing, they understand the complex world of robotics, automated manufacturing and project management. Delivering world class custom automation on time and on budget. Contact one of their sales engineers to see what Earhart can build for you. And Earhart's hard to spell. It's E, H, R H A R D T and I'd like to acknowledge A three the association for Advancing Automation. They are the leading automation trade association for robotics, vision and imaging, motion control and motors, and the industrial artificial intelligence technologies. Visit automate.org to learn more. Where you'd like to get in touch with us at the Robot Industry Podcast, you can find me Jim Beretta on LinkedIn.
Today's podcast was produced by Customer Attraction Industrial Marketing, and I'd like to thank my team Chris Gray for the music, Jeffrey Bremner for audio production, my business partner Janet, and our sponsor, Earhart Automation Systems.
