[00:00:00] Speaker A: The core technology that we have is two factors within this robot cell, which is one, the automated robot programming software. The other side is the onboard vision system that we integrate with the robot.
[00:00:19] Speaker B: Hello everyone and welcome to the Robot Industry podcast. Thank you for subscribing and we're glad you're here. My guest today is Matt Kostaragi from Tetragen Robotics. Hey Matt, welcome to the podcast.
[00:00:31] Speaker A: Hi Jim, it's, it's a pleasure joining you today. I've been a long time listener of your podcast, so it's an honor to join you.
[00:00:39] Speaker B: Oh, I'm glad that we finally were able to put it together. And you're from Western Canada, correct?
[00:00:45] Speaker A: Yes. So we are based in Winnipeg, Manitoba. Right now it's almost like minus 30. So it is not very unusual for Winnipeg at this time of the year.
[00:00:55] Speaker B: Yes, that's when you have to plug your car in. Right. For a lot of our listeners, if you don't plug your car in, in, in Central Canada or Western Canada, sometimes you won't be going anywhere.
[00:01:05] Speaker A: Especially when it's minus 30 and minus 40. That's when you actually need to plug your car.
[00:01:12] Speaker B: Absolutely. And for our, our friends in America, that is really cold. Like minus 40, like you minus 50. It's pretty cold.
[00:01:20] Speaker A: Absolutely, absolutely.
[00:01:22] Speaker B: Matt, tell me a little bit about your background. Like how did you get into the, into the industry?
[00:01:27] Speaker A: Yeah, so I've been in the industry, I would say for the past 15 years.
So initially I started as a researcher working with manufacturing companies and of course over the past few years in Tetragen or at Tetragen as a provider of solutions. Maybe a little more specific. I basically completed my PhD in the area of robotic in the area of manufacturing and automation.
And I completed my PhD in 2017 at UBC in Vancouver. After that I joined the University of Manitoba. I started a research lab focusing on developing AI enabled manufacturing robots and really intelligent robotic systems for manufacturing applications.
I worked there for maybe five years or so and that actually became a very successful research lab. We work with many manufacturing companies in defense and aerospace companies or in aerospace applications and also heavy vehicle manufacturing companies.
And honestly we heard the same kind of pattern of problem in all of these companies, which is they want to adopt robotic automation, but just those traditional fixed program, rigid type of robotic automation systems just don't meet their needs. They need more adaptable, more intelligent, more flexible robotic automation systems.
[00:02:47] Speaker B: And of all the applications you could have picked, you picked welding, right?
[00:02:51] Speaker A: Yeah. And we can get into that as well. Why we actually chose welding, but that's basically when I left that position that I had at the university and started Tetragen and the rest is history. We've been working on it full time for over three years and we have deployed several robotic systems already in production that are making parts. But yeah, that's how I got into the industry.
[00:03:14] Speaker B: Oh, that's exciting. And what does a system look like for? Those people are walking their dogs and they're like, hey, I get they're doing this, but what does it. So there's obviously going to be a robot.
[00:03:26] Speaker A: Yes. So primarily we focus on industrial robots. So first of all, you can't imagine a robotic arm, typical 6 axis robotic arms that you may have seen and images and videos. And then at the end of that there's a torch that basically, basically does the welding. At the moment we focus on what we call mig welding or arc mig welding. And yeah, these are typical robot cells that are mounted kind of fixed in one location at the manufacturing company and operators bring welded components or basically tacked sub assemblies and put them in front of the robot and the robot does the welding. But primarily industrial robotic welding.
[00:04:08] Speaker B: Right. And so who kind of are your bullseye customers? Are there anybody with a welding problem or anybody who's got a volume welding problem?
[00:04:16] Speaker A: Yeah, so I would say our target customers or kind of ideal customers. Right now they're manufacturing companies in heavy industries and that's really companies in agricultural equipment manufacturing heavy equipment and heavy vehicle manufacturing, manufacturers of trailers and attachments for transportation.
Generally companies that do sell a lot of welding. However, they don't make millions of the same exact machine or make millions of the same exact product line each year. They actually have relatively large SKUs of products. And also within each product they have lots of, lots of different welded sub components which really, even though over their whole operations they do have lots of welding, they don't have enough welding per part or per subcomponent to allocate the robot to make only or to weld only that one subcomponent every day and every night. What we provide are actually robot cells that have the flexibility that first of all you can very quickly reprogram, repurpose to start welding a new type of part. And then secondly is that the vision system that we have on board allows these robots to adjust on the fly. And that means that you don't really need a lot of very repeatable fixturing, very sophisticated fixturing. As long as you can hold that part somehow in front of the robot. We will be able to find every single joint located very accurately and make sure that you get a very good one. So really providing flexibility to manufacturers to use that robot for a much broader range of parts and also not need to rely on the repeatability of everything, whether it's part fixturing and all those limitations that are common in traditional robot cells.
[00:06:11] Speaker B: So we kind of start. I started asking this question about why did you pick robot welding? And obviously it was a specialty of yours.
[00:06:20] Speaker A: Yes. So one thing that maybe I should mention is that at the very beginning of our company, we actually focused on some other applications like sanding, trimming as well. And in fact, we deployed some robot cells that are still in production, making production parts that are powered by our vision system and AI software. However, once we started working on robotic welding, we realized that that's actually much better fit and much better application. Application. And there are several reasons. One is that of course, welding is a very, very big industry. In fact, over 50% of all manufactured products have welding as part of their process somewhere in that flow. Second is the need. Honestly, welding, if you're familiar with it, it's an art. It's not just something that you can learn overnight. It takes years and years of experience to be a good welder.
And right now, the average age of welders in the industry is 55 years old. And that means that as the previous generation of welders that are leaving the workforce, just a new generation is not as interested to go into these hard manufacturing jobs. So honestly, we have talked to so many manufacturing companies that have said that just their bottleneck right now is finding skilled welders. And of course, as that labor shortage becomes more and more of a challenge, the need for adaptable and flexible robotic welding systems is becoming more and more of a necessity. And the third aspect is, I would say we, and I mean our whole team, we are uniquely qualified to solve this problem in that robotic welding is an application that requires understanding of lots of different aspects. Understanding the manufacturing side, the robotic side, the robot error and calibration side of things, the machine vision and AI software aspect, it's not just about one component. You have to understand this whole system. And we come from a background of precision manufacturing, metrology, and at the same time, we have very deep expertise in vision and AI. So I would say we have really the set of skills required to solve this complex but much needed problem and develop a solution for it.
[00:08:34] Speaker B: So I'm interested to ask you about the vision system that you've set up. I'll ask you about AI in a sec, but tell us a little bit about your vision solution.
[00:08:44] Speaker A: Basically, our robotic cell. So first of all, we provide a full turnkey robot cell. But the core technology that we have is two factors within this robot cell, which is one, the automated robot programming software. The other side is the onboard vision system that we integrate with the robot. So in very simple terms, we say that we build the brain and eyes for, for industrial robots. And of course right now focusing on robotic welding. So the eye portion, which is a 3D vision system that we have on the robot, what we use it for is first of all, before doing any welding, we take some captures of the part. We do basically a check beforehand to make sure that there are no jigs and clamps in the way of the robot. Honestly, it's very common in the industry that people forget their clamp or something there.
And of course those traditional robots, if you crash into that clamp, all of your calibration will go sideways and then you will not even be able to reuse those programs. But with our system, first of all, we do all of those checks at the beginning to make sure that it's safe for us to do that process.
Second, as you know, the CAD models in industry never match the real part. Your parts, even part to part, is not consistent in welding especially because you sometimes just tack part which is really you join the pieces together with a kind of a dot weld.
Each time the part may sit a little differently. During welding, your part will undergo warpage and kind of thermal deformation.
So if you're relying and for welding, as I've mentioned, you have to be always within sub millimeter accuracy of your seam to get a good weld. Especially for mig welding.
Really the vision system, what it does is that of course it does all of the pre weld checks, but also during the process it makes sure that no matter if your fixturing is not consistent, if your part has warpage, if your CAD model doesn't match the real part, at the end of the day, the robot program will do a good job in welding that seam and really adjusting to all of those variations that we expect to see. And that's the reality of the manufacturing industry, especially the type of industries that we work with, heavy industries, you cannot expect them to have very, very tight tolerances. So that's really what the vision system does. We are using the vision system for some other aspects as well. So for example, if you are missing some sub components in your tag assembly, we can flag those. If you're way out of tolerance, we can actually flag that as well, to say that, hey, we can weld it, but you sure you want to proceed? Something seems not right and at the moment we are not doing any post vault inspection. However, that's something that down the road we definitely want to be able to offer our customers.
[00:11:34] Speaker B: Oh, thank you for that.
Now also too, when did you add AI or if you've always had AI in your system and what does it really do for the customer?
[00:11:43] Speaker A: Yeah, so one thing that I should mention is that we have taken more of a high hybrid approach to AI. We didn't want to integrate AI just for the sake of integrating AI. We integrated where it makes sense and when it makes sense.
So really AI is good at adding that more of intelligence and flexibility and adaptability.
So some companies, for example, they have decided to kind of jump to the last step, which is like they're trying to build a welding robot. They can learn from scratch how to weld, how to choose the weld parameters, how to control the robot motion, all those kind of things. What we have done is kind of more of a layered and step by step approach.
We basically started with existing robot cells that are working already in manufacturing environments.
The challenge is that they have, they take a lot of time and effort to program, but also they don't have the flexibility to adjust on the fly. So then we started kind of in the first step, we just added the vision system, said that, okay, you still do the robot programming. However, we use the vision system to make sure that it works on your actual parts, even if you have some warpage or inconsistency. Then in the second stage, we actually added the automated robot programming. So now through our software, which we call our key guide weld selector software, you just bring your CAD model, you say where you want to weld, and then we generate the full robot program.
[00:13:10] Speaker B: So you do the motion control, Everything?
[00:13:11] Speaker A: Everything, yes. Well, we still rely on the motion control system of the robot. We just tell the robot, we make all of the decisions as to where the robot should go, like the approach, the retract all of the angles. So from the operator end user, they just simply bring their CAD model, they select where they want to weld. At the moment, we rely a little bit on them choosing some of the vault parameters. So, for example, how much heat we really want to put into this joint.
But they do not have to do any robot programming. In fact, some of the robot cells that we have deployed, people who by trade are welders, but they've never done anything with robotics. They've never even touched a 3D CAD model. Basically, in their life they are in charge of setting up that robot cell by themselves because really the only thing that they have to do is to bring the CAD model, tell the system where they want to weld, subject to what wealth parameters. And we really do all of the robot programming based on the nominal CAD model. But of course the vision system will ensure that that nominal robot program will work on the real part. And that's really where we are bringing and leveraging and bringing in and leveraging. AI is adding that autonomy and intelligence more and more to our system.
[00:14:30] Speaker B: It must be kind of challenging, like you say with, when you come in with the fancy new automated weld system for a new user, are you finding that you have to do a lot of hand holding at the, at the very front end of your commissioning?
[00:14:41] Speaker A: So for our first deployments, yes. But as our system actually becomes more and more, more polished, it's actually, it's much easier to hand it off because one thing that we, from the very beginning we have made a top priority for us, it's ease of use. You can make a system that is very, very capable, but if it is so complicated that no one other than you can use, that just defeats the purpose.
And our vision from the very beginning has been that we want to deploy a system that we can hand off to the customer and non expert users will be able to set up the system, operate the system by themselves.
So whatever we do in terms of the capabilities most often is in the backend. But the front end is really still as simple as, okay, just bring your CAD model, select where you want to weld, and then say generate robot program and then run it on the robot.
And again, that's the part that actually without much training, definitely no training on the robot programming side, it's just kind of helping them understand the logic behind our front end, our software interface. That's really the part that honestly, a few hours we can train an operator, and then after that, of course, maybe the first couple of parts that they're setting up by themselves, they may sometimes have questions, but after that they can definitely do it independently.
[00:16:04] Speaker B: Your customers must be thrilled to have you come in and create a solution for them.
But they're losing people and they're not seeing a new influx. So automation's really their only choice, right?
[00:16:17] Speaker A: Yeah. And honestly, some people think that automation is taking jobs away. And generally speaking, it's the opposite. Most companies that we work with actually the bottleneck is finding the right welders or skilled welders so once they release that bottleneck and fill that gap, actually it allows them to increase their whole production and throughput, which means that then now they have to hire a lot more people for all other aspects of that manufacturing plant that they have.
So almost always any company that we talk to, once they deploy robots, they actually increase their workforce rather than decrease. It's really about filling the gap rather than replacing people.
[00:17:01] Speaker B: No. That's good to hear. And tell me, do you have any partners that you're always on the phone with?
[00:17:09] Speaker A: Yeah, I would say a robot cell is a complicated system, and as much as possible, you don't want to reinvent the wheel and try to go alone. You want to leverage other partners who have really done decades sometimes of R and D and development.
So for our robot cells, we rely on robot OEMs. Of course, all of the robot OEMs that you're familiar with particularly, we have some good partnership with certain robot OEMs. We are a Fanuc authorized system integrator. We're a Fronius authorized system integrator.
And we really rely on these partnerships and OEMs and suppliers because whenever our customers call us, if it's something that we cannot answer, then we have to rely on our suppliers and OEMs.
So definitely that partnership is something that we've been working on for several years now.
[00:18:04] Speaker B: Matt, did you have any surprises when you first got started in starting up tetrogen?
[00:18:09] Speaker A: Um, surprises. Of course, starting a company is always, every single day is a surprise.
But maybe one thing it's not a surprise, but something that really still stands out to me is how much this industry relies on personal relationships and trust.
No matter how good your tech is, at the end of the day, it comes down to, can that manufacturer trust you as a person, trust you as a company to deliver what you promise, and can they trust you to support them when they, they need your help? And, and maybe that goes back to what we just discussed, is that the same way that we have to rely on our suppliers and vendors to provide support to us, our customers, really, at the end of the day, the way they think about it is that, do we want to go in partnership with this supplier, this solution provider? Because the robot cell has a lifespan of at least 10 years, sometimes 10 to 15 years. So it's a really long term part partnership in a way between that customer, end user and the solution provider.
So I've seen again, times and times again that when that customer sends the po, it's not just because this is the best Product is you build that relationship. And honestly that's why the sales cycles in this industry are long. It's not because building a proposal takes time or negotiating a contract takes time. It's building trust and relationships. That's the part that takes time.
Thank you. Yeah, that's again not a surprise, but definitely something that we see it a lot.
[00:19:45] Speaker B: I totally agree with you. Thank you for that answer.
I wanted to sneak in a question about data and to find out if is this something that is important to your new customers or to your existing customers?
[00:19:56] Speaker A: Yeah, that's a very good question. Of course, anytime that we work with a company, everything that we do is subject to NDAs and we are very cautious about making sure that that we respect all of the IP and confidential information of that company.
In fact, we have in the past actually we have worked with some event defense companies that those companies, when you work with them, no matter how insignificant or insensitive something may seem to you, that does not come out of their company. Right. However, with most of the other companies that we have worked with, of course all of their designs, CAD models, all those kind of things are their own ip. We never try to train stuff on those parts unless it's for them. However, some other aspects. So for example, what are the best weld parameters for joining a plate to plate, a half inch plate to another half inch plate? Those are things that are less sensitive and honestly, like most of the manufacturing companies that we've talked to, they're not that sensitive about that type of information for us to add it to our library and the more information we can add to our library and our training data, the more value it adds across all of our customers. Because then maybe the next customer, if they want to weld that same type of joint, then they will be able to leverage that information.
But again, it's always based on that consultation with that customer just to make sure that they're okay with us to use some of that data that the robot is actually using to then weld some next set of parts.
[00:21:32] Speaker B: Matt, how does a company get started with Tetrogen?
[00:21:35] Speaker A: As I mentioned, again, this industry is all based on building trust and relationships. So the first step is that we always try to start that relationship by having a meeting oftentimes with a decision maker of that customer that we're targeting.
What we are providing the system can cost hundreds of thousands of dollars, sometimes even more depending on the system. So of course you have to be able to talk to the decision maker who will be able to assess if that's something that they can allocate the budget for.
So once we have that first meeting, if possible, we try to go visit them.
And that's where we go through basically the walkthrough of the plant and we see their operations, how they do things. Right now we have seen many, many manufacturing companies and each time we see things being done differently.
So we see their operations, we see their challenges, we see their parts and that's really where we can start talking their language, refer to their own parts and kind of start again building that relationship. But once we are past that stage, we oftentimes try to show them some demos as to how we can solve that problem that they have.
Sometimes we even ask them to send us some sample parts, some physical parts. We have actually a physical demo robot cell in our facility that we can show some full scale tests for that customer.
And then at that point, hopefully if they're convinced enough, we go into kind of the contract negotiation, sign the contract, we start building the system and then we deliver the system, commission it and support it afterward.
[00:23:19] Speaker C: Hey everyone, producer Jeff here we. We lost Jim's connection at this point, but we've kept the rest of the guest Matt Kostoraki's remarks. For the full story, please enjoy.
[00:23:33] Speaker A: Yeah, for sure. Absolutely. Again, welding is a, is a very, very big industry. It's not going to go anywhere. And just considering all of the labor challenges, I'm sure that the robotic welding industry as a whole has still a very bright future.
What I would say is that there are a few different ways that companies are targeting that or at least a solution provider. Some companies are going building something for probably sometimes even decades from now. For example, I'm not sure if you've seen Hyundai has started working on humanoid robots for welding and that's kind of for shipbuilding applications.
So that's one approach. And some other companies have tried to pursue similar ideas. But industrial welding generally build an industrial robot that can learn from scratch how to weld, how to select the weld parameters, do all of the control vision side of things.
What we have done, as I mentioned, we've tried to kind of take a more step by step approach, more incremental, and add more and more autonomy and adaptability.
Let's say all of these things probably will converge somewhere down the road. So some people kind of start working from now on that vision that they have for 20 years from now. And some companies like ours, we try to deploy something that works right now, but also make sure that it continuously adds more and more and more Value. But we definitely see a very, very big market for industrial welding and AI enabled welding applications.
Yeah, I would say maybe one of the question that we get a lot is at maybe at the beginning of the process that, okay, even if we sign a PO today or send a PO today, how long does it take until we get that robot cell? I'd say the answer, it definitely depends on the customer and application. If it's an application that it's really just our standard robot cell with everything as is, I would say even we can get that out the door within three months. Provided again, there's no crazy lead time on the robot and welder.
If that customer relies on us to also provide some of the jigs and fixturing, of course that's the part that will take more, more time and potentially some customization. Let's say anywhere from three to six months. That's a very good estimate.
Some other questions that we get is that especially companies that have used offline programming in the past, for example, if you're familiar with the simulation software of some of robot OEMs, some engineers especially, they might have used that software before and they ask us, okay, how is your system different?
And really the answer is that through our system you don't do any robot programming. If you want to use offline programming, what exists on the market, you still have to have an understanding of robot kinematics, robot configurations, singularities, approach, retrack, all those kind of things.
Whereas in our system, really you do not have to worry about any of those aspects. You just define what you want to weld and we take care of all of those other aspects.
Then maybe one other question that we get a lot is that okay, in robotic welding right now, there are some other technologies like touch sense and seam tracking that are sometimes used. How is your solution, how is your vision system different? And the answer is that some of those techniques touch sense, scene tracking. First of all, you will need to have some level of expertise in robot programming to set them up. Oftentimes, robot integrators, when they deploy a robot cell, it's their job to set up some of those aspects. Let's say the touch sense and scene tracking. Whereas our solution is agnostic to the geometry or shape. The same way that you set up disjoint, it's the same way that you will set up some other joints.
In terms of generality, it's much more generalized. It's not very party specific, but also it can be set up with someone who does not have any expertise or experience in robot programming, it's really you just bring your CAD model. You say this is where I want to weld. And if you want to add any Touch sense, you literally we have the option to say, okay, introduce a touch sense point here. And then we do again all of that robot programming for that Touch sense event.
I think we covered a lot. Maybe just one thing to mention is that we have been exhibiting in of course Manitoba, in Alberta and we are looking to exhibit this year in Ontario as well. So if you see us around, please do stop by and say hi, we'd love to talk to you.
I would say the best way is probably LinkedIn. I'm very responsive on LinkedIn when it makes sense.
So yes, if you just find my name or if you find Tetragen Robotics on LinkedIn, I'm sure if you go through the list of people you'll find my name easily.
[00:28:41] Speaker D: 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 their
[email protected] and Earhart's hard to spell. It's E, H R H A R D T. And thanks to our other sponsor, Mecademic Industrial Robotics. Mecademic builds the world's most compact and precise robots in industries such as photonics, med devices, optics and electronics. Mecademic continues to set new standards in precision, footprint and flexibility, accelerating small component automation for manufacturers, robot integrators and innovators alike. And you can find
[email protected] 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. You can Visit
[email protected] and if 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 sponsors Earhart Automation and Mechademic Robotics.
