Blog

Discover the ins & outs of 3D vision for robots in our podcast!

Learn about the 3D technology Pickit has introduced to the market place - how it helps with singulation, what the ROI looks like, and what applications Pickit's cameras are best suited for. Listen to the interview with our in-house bin picking expert for the Focus on Automation podcast series of our US partner the Knotts Company.

Prefer reading? Here's the full transcript:

The Knotts Company welcomes you back to focus on automation where we explore and celebrate advances that empower manufacturing.

Todd Youngblood  00:17

Welcome back to focus on automation. I'm your host Todd Youngblood and today we're going to dive into something that frankly sounds a little bit like science fiction robot vision, but to help us understand it and the reality of all of it. I'm joined by Mr. Sam Biermans. He's Channel Manager with Pickit. Sam, thanks for joining us.

Sam Biermans  00:35

Hi Todd. I'm happy to be here and as you said, some things do sound like science fiction, but it's not always what it is.

Todd Youngblood  00:43

I know that and I can't wait to dive into it before we get into the, the actual technology. Sam, tell us a little bit about yourself and a little bit about Pickit.

Sam Biermans  00:51

Okay, Pickit is a Belgium-based company and we were established under the name of Intermodalics and we still do a lot of projects on the software side, so purely software, high-end vision and robotics. And it goes from indoor GPS positioning to as broad as you can see, self-driving cars, drones, household biotics during the 2010, 11, 12, we got a lot of requests from different companies about, hey, can you write me software for my camera and my robots? Uh, because we got so many requests, we knew there was a market question about a 3D vision solution that will help people pick up objects from a bin basically. And that's where, that's from where we started and then we launched a product, what's Pickit 3D. And so I've been working the company now two years and I've been mainly focusing on the U.S. manufacturing market. 

Todd Youngblood  

Got It. What'd you do before Pickit, Sam? 

Sam Biermans  

So before I worked in the data center construction, uh, and I was a commercial technical responsibility for the whole project from getting an order into completion on two to three year projects of data center, manufacturing, construction. So the building side, the technologies, and inside the data center.

Todd Youngblood  02:13

Interesting stuff. That's pretty cool. Um, right off the top here as we dive into this thing, I want to spend a little time talking about the differences between 2D and 3D. On the one hand that's perfectly obvious. Uh, but what are some of the implications of that and moving into this world with all this 3D technology?

Sam Biermans  02:30

Yes. So, what we're basically doing as a complete that is, moving into the 3D market is also we're building our own markets. So a lot of people,  especially in the industrial automation world, they know 2D. They know 2D cameras, they know their capabilities, they know their strengths, but they also know their weaknesses because people know 2D, there's also a lot of workarounds to get a 3D situation into 2D scene where a 2D camera can handle it. For example, vibration tables, mechanical singulation projects. So 2D cameras are great for a lot of solutions that are out there, but a lot of people are also working with 2D where they would benefit more from 3D. What is the biggest difference that a 2D system is perfect for doing quality control, Barcode rereading, color scanning, ah, surface controls, those kinds of things.

Where a 2D system struggles is parts singulation. So once a part is singulated, a 2D system is brilliant. That's also not what we're aiming for with our company because we know that the 2D systems who are out there [inaudible] and all the other guys, they're perfect in doing a quality scan on the singulated objects. Where we're moving in is before we get the object singulated, that's where the 3D comes in handy, because then the product line, different, um, orientations that are on top of each other. They like touching, uh, they lie in a dirty environment, all those kinds of things. And that's where 2D system has more, I wouldn't say limitations, but it has not the capabilities to handle this. 

Todd Youngblood  

In terms of the markets that you're aiming at, Sam, is it, it sounds like it can be very broad in terms of any bin picking sort of applications. Is that true? Are you looking under more narrow set of application?

Sam Biermans  

Yeah, it's good. Just that question because we see a lot in the world that people on Linkedin and wherever the marketing strategy say, Hey, been picking solved, so we as a company, we're not. Yeah, you probably have heard of seeing those kinds of sentences. 

Todd Youngblood  

Yeah. Actually, I thought bin picking was solved about 20 years ago.

Sam Biermans  

…so that's the funny thing because that's not what we are claiming. What we're trying to do is we're trying to help manufacturing companies who struggle with finding labor forces to do repetitive tasks. The first, there was a big wave in automation when robots started. People started learning, hey, we can do repetitive tasks, we can do certain things. Now a lot has been automated last years and then with collaborative robots there was a new wave coming in. What suddenly was possible that you will work as alongside the robot without having fences, gauges or without having people that are highly-skilled trained for it, because it's now accessible to more people and that's what we're trying to do.

 

We're not trying to solve every been picking application. We're trying to help people solve applications by doing it themselves. You can go to an integrator and pay 200-300 grand to get him fixed a 3D bin picking solution for one product in your factory. That has been solved, that has been done. What we have is off-the-shelf product that you can integrate with your robot and it helps you to solve certain applications. How do we qualify these applications? We always test the products so we have a comfort zone, what we call the typical applications. What are these? These are most of the time picking steel billets and I'm putting them in heat furnace, picking driveshaft couplings, picking objects that are fist-sized that are metal or plastic and they're in a bin and now you need to singulate them because you're going to grind them, you're going to heat treat them, you've got to whatever you want to do with them and that's something we, we know we do what and we do it fairly straightforward. 

 

Todd Youngblood  06:16

Okay. It sounds like the payload that you can handle is all over the place from steel to small plastic parts. I wanted to make sure I heard that correctly.

Sam Biermans  06:25

Yeah, but so the payload is not something we do. So we are, you can couple us to any robots. 

Todd Youngblood

So I got it 

Sam Biermans  

…so basically what we are, we're basically a sensor where your robots can ask a question to what's the question? Hey Pick-It. I'm looking for this object. Can you find it for me and pick it. We'll tell the robot, yes, I find it for you and it's lying here. X, Y, Z , that orientation but also xyz rotation. So that means that the robots can go into a bin and grasp an object on a certain spot of hole. For example, if you have a dry shift that's just been milled, you don't want to grab it on the middle surface, you probably want to grab it on a thicker edge or something where you don't damage the middle surface. 

Todd Youngblood  

Right. 

Sam Biermans  

The thing about 3D that we don't really understand as a human how it works because we use it every day and our eyes are the best 3D camera in the world. Our eyes together with our brain can handle a long, dark light, very bright light surfaces that are complete flat because we have a brain and that's the artificial intelligence behind it. We know what we're seeing. A 2D camera only sees in one plane while a 3D camera sees in two, three dimensions. So not only a plane, but they can also see the Z or the depth information.

Todd Youngblood  07:39

I'm the guy on the factory floor and I'm saying I'm, I'm looking to automate more and more of the, of the whole manufacturing process. Walk me through how do I, what's, what's my decision process? How to first do I identify what I've got that lends itself to a 3D application and then how do I begin to step through that doing the design part of it, figuring out the, the financial part of it. Just walk me through how I would do that.

Sam Biermans  08:04

So what we always suggest people is if it's possible, start easy. So that's also how the first robots came into factories, how the first collaborative robots, that's also how you should work with vision. So what you should try to do is you should try to run into your manufacturing floor from where raw goods or materials come in. So that can be either purchased plastic parts, purchased, uh, forge parts. They come in and at some point they need to be singulated like everything because you're gonna handle those parts or you're going to assemble those parts into something bigger. 

Todd Youngblood  

Got It. 

Sam Biermans  

A lot of the automation has been done between once a part, is singulated until the finished product and there's basically two zones where you could use 3D vision. One is the parts manufactured into something else and now it comes out of a conveyor, but the robot doesn't know always how the part is lying, right? Is it orientated around? Is it rotated? But most of those things has been done mechanically because in the past there wasn't 3D systems, but on the other spectrum, in the beginning of the manufacturing, a lot of people have trays full of objects where people are singulating, just taking a product the whole day and putting it somewhere or they're using big vibration tables who make a ton of noise. Take up big floor space to get those parts into a one by one position. If you have a product that you think, hey, I have an idea how to grasp it because we come, we will come later to the grasping of objects. Finding an object is most of the time not that challenging for our vision systems. The way you want to take an object is most of the time more challenging and there's a lot of integrators in the US who have knowledge about building grippers but these guys are very good at keeping their knowledge.

So we get for example, a lot of requests from companies, hey, I have a very special part, let's say an engine head because everybody can imagine how the engine head looks and I wanted to take it into the two holes where I go in with my spark plugs for example. 

Todd Youngblood  

Okay, 

Sam Biermans  

…it's all fine, but if you don't see if we can find your objects [inaudible] and the object is still that, we can still tell you where those holes are. But how are you going to go in with a robot? Because the robot is not a hand that has unlimited movement of joints. You only have six or seven axises. Correct. So from every side you look at it, you need a different pick frame and a pick point. So it could be that if you look from one side, you can go perfectly in the holes and open up and grab it from the other side. The holes are maybe not visible or not reachable. So you maybe need to finger grip at a 10 clamp and close, pick it up, drop it somewhere else, and then you can go back into the holes. And then from another side, you may need a different kind of gripper. So the whole design of a gripper compared to the complexity of a product is important. When you start with automation, with 3D vision, you don't have to worry that much about finding what's in there. Because if it's not translucent and it's not a mirror-like stainless steel, most of the objects can be found by 3D system. 

Todd Youngblood  11:11

Do I need typically, what I need an integrator to help me with that gripper or is that something you, you talked before a little earlier about being able to do it yourself in terms of programming. Is that that apply also to the gripper? Do I typically need an integrator to help out with that yet? 

Sam Biermans  

So that's why I say start easy. If you have cylindrical kind of parts or plastic parts or parts that you want to grab out of a bin and you know that you can approach them from above, that means either here was [inaudible] gripper or a magnetic gripper to pick them up. That will give you a ton of freedom. Why? Every flat surface that you have, you can pick something up straight from above. You don't need fingers to go in holes. You don't only fingers to open, you don't need things to close and it gives you more freedom to play around because most of the time, every visible surface from every side that is flat will you to pick it up and that's something you can do yourself. The moment you go into a high end custom grip first you probably need an integrator unless you have some engineering in house, but the more simplified applications where you have a magnetic gripper or a [inaudible] gripper, it's easy to buy. It's easy to build, especially now, I mean the collaborative world you have a few guys who have even vacuum generators that you can mount on the robot head, magnetic grippers supplied by the typical gripper companies. So there is a lot to do there and it's moving very quickly. 

For example, we have Vanamatic, KYB, 21 Century Plastics. They're three companies who had time and resources and technical knowledge in house and they just bought a Pickit box. They bought a UR, Universal Robots. They designed a very basic gripper and then they figured out, hey, with this technology that we have, we can do this. They found in their plant one project that they could work on and they learned by doing it what are our applications they can do and now they're all running with multiple vision systems, 3D system.

Todd Youngblood  12:56

I got it. So that, yeah. The old adage of walk before you run really, really makes a lot of sense. Very applicable right here. Sam, let's switch a little bit and talk about the economics of it and you give it, give me an idea of what kind of break even points you can get or return on investment. What's the, what's the money side of it?

Sam Biermans  13:11

So if we talk about the money side and return on investments, so a lot depends on what kind of robot brand you're using, but you have to count for vision systems that they can vary between 20K up to 40 50 60K. Now Pick-It, we have three different models, so we're more between the 20K and in the 37 to 38K depending on what kind of model you want. So we're in that price range. That also means if you add a robot, you add some grippers and you add fixtures and everything. You are around 100K to have a small cell up and running. 

Todd Youngblood  

Okay. 

Sam Biermans  

If you look at labor costs in the US and you count the shift hours and everything, then for a lot of projects where you have a product that you can run in three or four shifts, you should be able to reach five to six months on return on investment. 

Todd Youngblood  

That's pretty quick.

Sam Biermans  

Not by eliminating those people, but by helping those people quit repetitive tasks because everyone that works in a plant as a labor force should be able to add more value than doing a repetitive monotone task.

Todd Youngblood  14:19

Yeah, no, there's no, no doubt about that.

Sam Biermans  14:21

And that's our true belief in our company. So we're not trying to get people out of the jobs by putting robots there. We're trying to move people from doing repetitive monotone jobs who are basically not good for motivation, into a state where they add the human knowledge and the human interference into a manufacturing process.

Todd Youngblood  14:42

Talk a little bit about the quality aspect of it. I, I imagine there's gotta be an increase in the quality of the manufacturing products. Just cause I got the orientation of everything is correct. Is that another area where I can get some ROI?

Sam Biermans  14:52

Yeah, so why I'm saying five to six months, that's the ROI that we calculated that if you take into account everything. So you're not only taking into account the labor force, you're all set, you're also eliminating human errors. It's not bad because you learn from human errors, and I think human errors are essential for manufacturing from moving forward, but in some processes that are fine tuned, we should eliminate them. For example, we had a customer who is basically just order picking. So he needs 10 bolts in a bag, 20 washers in a bag, and 40 support struts in the bag. Why is it? Because that bag goes with a lot of metal frames and he sells it as um, warehouse reqs. So, uh, projects for example, of 200,000 US dollar, which warehouse reqs gets bought by him, he assembles the metal parts. But then he also needs to bag the small components that are there. But imagine you buying 2 or 300,000 project and one of the important bolts is missing. The whole customer experience is bad, because the whole product is not good. And so he said, my return on investment is a zero on saving cost, but on my customer experience it moves gigantically because the robots counts and it doesn't make mistakes.

Todd Youngblood  

You only need to prevent one of those with a major customer. Sam, talk a little bit about the support structure. How are you, how are you going to market particularly in the US to make sure that the, the knowledge in terms of identifying applications, getting the right design done, implementing, et cetera. How, how are you supporting your customers? 

Sam Biermans  

As I mentioned, we're a Belgium-based company. Our headquarters is in Belgium. We have  certified channel partners in the US. We have about 15 of them. They all have a demo unit. So what means, they can all test products. So what we're trying to do is especially not sell our technology where it is not perfect for that application. So we're trying to test every product that we get in as a question, after some evaluation on the sales cycle and all the sales evaluation. Okay. Yes, we will make a proof of concept for you. We will show you what we can do with our technology that allows the customer to get more confidence in a new technology as 3D. By doing that, on the other hand, we are training our partners in how to set it up, how to run, so our partner network does the first line support consult systems in the US. On the other hand, because we are a global company, we also [inaudible]. We will establish normally by the end of 2020 in office in the US so local presence. Why do we believe that we need a local presence in the US due to the time shift, but especially because it's one of our focus markets, we know that there are about 5 million job openings will be created in manufacturing in the next 10 to 15 years and probably 40% of those will not be possible to fill. 

Todd Youngblood  

Right. That's going to be, there's a heck of a problem. 

Sam Biermans  

Yeah, so that's heck of a problem. What's really important for us on the support side is that we, because we'd still designed software and we designed our own software in house, we build in a lot of capabilities. One of the things we have is that you can connect your system online. I hear now already on the other side of people listening here, everybody works in IT is now moving back into his seat and saying nobody goes online.

Yeah, yeah, we're ready for IOT and we are ready for big data and we can transfer the data. On the other hand, what do we have? We have a tool that's called a snap shelter to try to simplify what a snapshot is. It's basically taking a screenshot on your computer, but it's a snapshot and it's a document that saves the entire 3D scene. So what the camera's seeing in 2D, what the camera's seeing in 3D plus all the settings and it crops it into a very small file that we can open on any Pick-It system. You can download it from your system. So even if the Pick-It system in production is not online, you can download that file, put it on a flash drive, email it to us from your normal computer and our support engineers anywhere on the world can work on it on a virtual system.

We can tweak the settings, we can tell you, hey, but you put a value on 100 that should not be 100 that should be 10 and send back with the right settings so we can do onsite support without being onsite. 

Todd Youngblood  

That's pretty exciting stuff.  Sam, before I let you go, I wanna I want to ask you to peer into your crystal ball a little bit. What do you see coming next over the next three to five years? How are we going to build on this technology and what sorts of things do you see happening? 

Sam Biermans  

So I think what's the most interesting for us is that more people are moving into the 3D, 3D vision market, because we need everybody. We need everybody to work together to get people who are biased on it's not working and it's not ready, uh to convince them that it is ready and that there's an era coming now where you can do a lot more with robots because we have 3D capabilities out there. So that's the main thing. And the ball is rolling slowly, but at some point it will get at a snowball effect, if you know what I mean. 

Todd Youngblood  

I know exactly what you mean. 

Sam Biermans  

…and that's why we are at the forefront, we’re the forefront fighters. That's why we are also investing so much money in R and D because we want to stay on the tip of that mountain and move as the high tech best vision solution that's most usable and scalable for bigger companies. We're talking with a lot of bigger manufacturing companies in Europe because it's our home base and they all like what they see and they all say, we have to move with you guys into this market. And we have to see, maybe to adapt some things to really scale it up in our company. So what I see in five to ten years is that where we are now, we see with 2D, what everybody knows, what everybody can use, what everybody understands…that will be five to six years 3D. 

Todd Youngblood  

Got It. So it's not the least bit science fiction. This 3D robot vision is, is for real and it's today, 

Sam Biermans  

…it's so not only is 3D being used for bin picking 3D’s everywhere right now, um, everything that's suited with automation. So there's some self-driving cars. So they use lidars, they use 3D formation. There's drones who do inspections on high altitudes buildings on electricity distribution networks. There's indoor GPS app positioning that doesn't work, so they use 3D information to know where the robots are, household robots. There's logistics scanning in warehouses. There's even Walmart is doing scanning on their inventories. So the 3D allows robots to do a lot more and I cannot stress it enough. It's not that robots will steal human jobs. It's to get people out of monotone, dangerous, non-ergonomic workspaces. Because I believe, and I'm a strong believer in this company with everyone here that people should be able to add more value.

And there are more valuable jobs and if people are interested, you should look online to some of our testimonials where we talk with people who actually before worked on a production line where they had to, for example, a singulate small parts into a fixture and now by doing that they have four robots that they are responsible for. And they have more responsibility and now they are more of a manager of four robots who do the same task of them. They still have to check them if they're running, if they have enough material to run, but they don't have to do the monotone repetitive task everyday anymore. 

Todd Youngblood  

Right. It's an upgrade of the job. 

Sam Biermans  

It's an upgrade of the job. Indeed. That's the right, yeah, that's the right term. 

Todd Youngblood  

Just to get us wrapped up here, Sam, how can folks learn more about Pick-It, the applications, the technology? You had to give us an idea of how to, how to get at this.

Sam Biermans  22:30 

Okay. Yeah, so New York, New Jersey region, we have one of our partners who is Knotts Company and so the Knotts Company is one of our certified partners. So they have a demo unit, they have partnership with Pick-It, they do New Jersey and Metro New York area so they can test all of our products. They are already 67 years I think, or 76 years in industrial automation. And they have two upcoming shows. So one is on October the 4th, that's NJMEP’s 2019 Manufacturing Day in Somerset, New Jersey. And then October 23 and 24 Design-2-Part in Oaks, Pennsylvania. 

Todd Youngblood  

And how can folks get a hold of you, Sam? 

Sam Biermans  

Okay, so how to get a hold of us. So we have a website, pickit3d.com and on that website there are all our contact details. If you want to find me on Linkedin - it’s Sam Biermans. So Belgium beer guy, basically, in flemish. But if you email info or sales@ pickit3d.com you end up with any of our team and we'll get you sorted out. Outstanding. 

Todd Youngblood  

Sam Biermans, Channel Manager with Pickit. Thanks for taking the time to share your insights and expertise. This is fascinating stuff.

Sam Biermans  23:46

Okay. Thank you, Todd, for this interview. It's always nice to talk about where we are and where are we going. 

Todd Youngblood  

Great. Thank you, Sam. That wraps today's episode. On behalf of my guest, Sam Biermans and our sponsor, The Knotts Company. I'm Todd Youngblood. Talk to you again soon.

 

Do you want to learn more about bin picking as well? Here's some exciting news: we are launching our online BIN PICKING ACADEMY very soon.  We can already reveal it will feature videos, a podcast, a webinar and the ultimate mini guide to bin picking...

Sign up for the bin picking academy

 

 

Looking for a proven solution?
Let us test your part for free.

We make real-life demonstration videos every day. Join the club!