Speaker 0 00:00:00 We make software for robots that gives them more spatial intelligence.
Speaker 1 00:00:06 Hello everyone. And welcome to the eighty-three robot industry podcast. I'd like to welcome our listeners all over the world from BJ WADA, Aurora, Ontario, and Sussex England. My guest today is Jared Glover. Jared is the CEO and co-founder of Capson robotics, a company that makes software to give robots more spacial intelligence. Jared received his PhD in computer science from MIT in 2014, where he developed and applied new theoretical tools for processing 3d orientation information to applications in computer vision and robot manipulation. Prior to that, he completed his BS in computer science from Carnegie Mellon university, where he led a team developing robotic walkers. He has 20 years of research experience in robotics and computer vision and over 500 paper citations. He's also a board member of catalyst connection at private non-profit that provides consulting and training services to small manufacturers in Southwestern, Pennsylvania. Jared, welcome to the podcast. Thanks. Thanks for having me. Hey, I have to tell you that I'm very interested in bin picking and really looking forward to this, uh, discussion. So I have to ask you, Jared, how did you get into robotics and automation and why did you want to focus on Ben picking, which is often called the Holy grail of automation?
Speaker 0 00:01:30 Uh, well let me start with the Ben picking. So, um, you know, basically we started the company six and a half years ago after my PhD where I'd been working in robotics for a long time. I can tell you how I got into robotics in a minute. Uh, but basically we started off with some software that I developed in my PhD, uh, for 3d vision, helping robots to, uh, find and recognize objects and so on. And we just started talking to people in the industry. We've talked to, um, you know, factory owners. We talked to systems integrators and there's kind of this resounding theme that everybody wanted to, uh, have robots that were better at picking objects out of unstructured containers and Inns. Um, and that just kept coming up again and again. And so, uh, one thing led to another and we ended up developing a product specifically for that.
Speaker 1 00:02:25 And what's been your traditional market, and I assume it's been heavy duty parts with big bins, maybe the automotive sector
Speaker 0 00:02:32 It's, it's, uh, it's kind of evolved. We're not really specific to any single market because our technology is very general purpose, right? We're giving the robot the ability to locate and pick up and manipulate objects, uh, from these cluttered spaces. And so that can be applied in a number of different industries. Uh, so we started out with a lot of traction, um, in manufacturing of small metal parts because those were, that couldn't really be done before we came along. Things like small metal hooks, they get entangled and other things where you have to do very precise placement, precise assembly steps, and so on after you pick them up. Um, but you know, now we're, we're in medical packaging, we're in logistics where nuclear. Um, so we're, we're, we're spreading out, uh, pretty widely already.
Speaker 1 00:03:22 I forgot to ask you about, about your robotics. How did you get so interested in robotics?
Speaker 0 00:03:27 Yeah, it was, it was kind of an accident. I was an undergrad in Carnegie Mellon 20 years ago and, um, I just happened to wander into a robotics lab run by, uh, Sebastian Thrun when he was still here in Pittsburgh and, uh, um, you know, kinda just got lucky and got, uh, got a summer job, uh, helping them out and, uh, kind, never, never looked back since I loved it. So you were
Speaker 1 00:03:49 In the right place.
Speaker 0 00:03:52 That's right. Yeah, definitely.
Speaker 1 00:03:54 What are some of the challenging parts that you have to deal with? Because you're what you say you're in so many industries and so many part geometries.
Speaker 0 00:04:03 Yeah. I mean, it's, you know, it's not like any specific part is, um, is necessarily that much harder than another, but the challenging thing is providing software that will work for such a wide range of objects, right from the tiny screws and hooks that would deal with washers, uh, those kinds of things, all the way up to big boxes and warehouses, um, you know, large metal parts in the nuclear industry. Um, you know, we, we have to deal with a wide range of, of objects. And so that's why we've designed our software to be, um, able to be used with the variety of different 3d sensors and robots, uh, so that we can kind of select the right hardware for each one of these objects. Cause, cause cause mainly it comes down to the choice of hardware. It comes down to, you know, whether your 3d camera can see the parts in enough detail.
Speaker 0 00:04:56 Uh, it has a, you know, wide enough field of view and it comes down to, you know, whether the robot is big enough and has the payload and more importantly, the end defector have you design the end, effector are using suction, what kind of suction cups, uh, are using fingers. So we sometimes get involved with the finger design. Um, but you know, unfortunately we're still not at the point in the industry where you can have a general purpose robot end effector that just allows you to pick up anything. You know what we're getting towards that on the software side, that's what we've been working on so hard, but there's still customization to be done on the hardware side. And so, uh, we've put a lot of effort into just supporting lots of different hardware, um, and getting good at it, helping to design the hardware ourselves on the end effector side.
Speaker 0 00:05:43 Like I said, um, because there's really no, no, um, no shortcuts yet in terms of, you know, completely general purpose hardware solution. Um, but you know, getting back to your original question, did the hardest, the hardest objects traditionally and been picking have been ones that are, um, very shiny or transparent, um, or objects that are small and get entangled and, uh, you know, those, these are all things that we've been able to make progress on. Um, but they're so object types that are kind of beyond our reach, even our reach, um, you know, things that are completely flexible, like, like just jumbles of wires, if they're not structured in a nice way, um, or things that are completely reflective, like mirrors are still beyond us. Um, but pretty much everything else we've been able to make some progress on.
Speaker 1 00:06:33 And I assume that you're, when you're doing end of arm tool work, you're either recommending them or buying them on behalf of the client, but you must have a very wide variety of say of a tribal knowledge to say, well, this part we really need to pick up with this end effector.
Speaker 0 00:06:47 Yeah. It really is, is an art more than the science, uh, unfortunately. Um, and so you, we, we definitely have some experience on our own of, of designing and selecting end effectors. Uh, but we also rely heavily on our partnerships with system integrators because they've been doing this, you know, maybe not specifically for bin picking, but for wide variety of picking place tasks for decades. Um, and so we really, we really do rely on our partners a lot, um, for that expertise.
Speaker 1 00:07:14 And I was going to ask you this question, like, do you deal with end users or do you deal with automation integrators or both?
Speaker 0 00:07:22 Both. So we, um, you know, we have customers that are the large end users, the fortune five hundreds, uh, and we have customers that are the integrators, but, uh, we're heavily involved in the integration right now, either way. Uh, you know, we're right there side by side with the integrator, even when the integrators, our customer to help them configure and train our system for the new objects and tasks. Um, we are working on, um, finishing up our user interface for integrators, which is going to be ready in a few months. And, um, that will allow the integrator to do, uh, more of the training and configuration that we do now. Uh, but, but the point is that, um, you know, we still will be there providing very thorough support, um, and help with the integration, even if, you know, we're just kind of watching the integrator and guiding them.
Speaker 0 00:08:13 But, but the point is that one thing that I feel is held back in picking an industry, and maybe we'll be talking about this more later, but, um, despite this act that, uh, you know, it's been proclaimed as kind of being a solved problem for a few years now, I think that's pretty far from the case because, um, there's so many details to really make it work in a live installation. And, uh, you know, oftentimes, you know, the integrators don't really have the tools or the support from the Ben taking vendors, the 3d software, the 3d vision benders to actually get the job done and meet the specs. Um, so that's something that we've put a lot of emphasis in is, you know, getting good at integration and integration, um, so that we can make sure that each project is successful.
Speaker 1 00:09:00 And of course, lots of things happen along the way, right? Like lighting changes or maybe there's a new batch of parts and it's not the spec and all of a sudden something reflect something's more reflective than it used to be.
Speaker 0 00:09:12 Yeah, absolutely. So, you know, one of the benefits of 3d vision is that it's less sensitive to lighting than a more traditional 2d vision methods, but you still do have some lighting issues. I mean, we had an issue recently where there was some led lighting that was very shiny and, and, um, and focused, it was up in the ceiling, 60 feet up above the robots, but it was still shining through, into the robot cells and reflecting off of, um, one of the objects that was extremely shiny, extremely shiny, a covering on the object. And, uh, and that was causing some issues for the vision systems. So there are still some issues like that where you kind of need to take a look around the whole environment and make sure nothing is going to be interfering. And that's just based on experience.
Speaker 1 00:10:01 What are some of the trends you're seeing from your customer base? Like you must be getting a lot of questions from both the OEMs or so, but also from systems integrators and I assume applications. Yeah.
Speaker 0 00:10:12 Uh, well, you know, the big trend for many years, um, really hasn't slowed down in the last year is just that everybody has labor shortages. I'm sure that's probably a recurring theme on your podcast, but, um, you know, all the end users that we talked to, all the factory owners and warehouse operators, they all say that they can't find enough people to do the work and they can't keep the people they have and that's for a variety of reasons and even more reasons now after the pandemic. But, um, I think that's a long-term trend. That's gonna continue to be the case. And that really is the number one reason why people are looking to automate more now, um, in terms of other trends in the industry, um, you know, in terms of what integrators are talking about, you know, they're all looking to stay ahead, um, and, and be, be one of the ones at the front of the pack in terms of using technology like Ben picking.
Speaker 0 00:11:09 Um, but at the same time, you know, they don't want to get too far ahead of the curve and, um, you know, take on too much technical risks. So there's, there's kind of a balance there. Um, so it's interesting. I mean, we've talked to a lot of companies that, uh, you know, really love what we do and really want to get involved in Ben picky, but they're all a little bit, you know, scared, they're a little bit hesitant, um, to take that leap sometimes. And, uh, so one thing that we try to do is just, you know, do very careful, um, analysis and testing ahead of time and, uh, you know, make sure that the customer's really comfortable that this is a real thing, and it's going to meet their cycle times. We've worked really hard on, on cycle time in particular our software, um, by accelerating all of our basic math with, uh, with GPU programming, um, and also by allowing the robot to move while we're doing the image processing and motion planning. Um, so the robot never has to stop moving. And so that allows you to get the cycle times that, you know, haven't really been possible with most been taking systems until now. Um, so anyways, we, we try really hard to, um, to alleviate some of those concerns.
Speaker 1 00:12:20 And I assume that, you know, if there is some risk that you might bear some of that by doing a, um, engineering study, whether it's paid or whether it's free.
Speaker 0 00:12:31 Oh yeah. That's been happening for sure. A lot of companies, especially the larger ones, want to do a proof of concept, a pilot first, where they want us to set up a demo in our office first and then they'll deploy it in stages or something. And we're set up to do all of that. Um, and you know, I, I, I understand, uh, you know, it, it definitely, if you haven't installed something like this before you do want to get that comfort level before you go and revamp your whole factory floors, especially if this is part of like a larger assembly line, which we've done, um, you know, where if this doesn't meet the cycle time or it's, uh, you know, has more faults than the rest of the system, you're, you're slowing down your whole line. So, uh, it, it definitely makes sense to do some kind of, um, uh, you know, some kind of a test, um, you know, field acceptance, acceptance test, uh, in our facility or the integrators facility before you install something
Speaker 1 00:13:28 With the pandemic, you've got a lot of interest from medical device and diagnostic industry, um, except for the obvious why is that?
Speaker 0 00:13:36 Well, I think it's, it's all the same trends that are in the rest of the industry. Uh, you know, I think labor shortages are, are huge and are going to continue to be huge. And, um, you know, I think social distancing, obviously from, from the pandemic has, um, accelerated those trends in some ways. And, uh, you know, the, the, the, um, the medical companies, especially the ones doing packaging and so on, they, they, they have, um, you know, unique requirements compared to manufacturing say, or, or, um, logistics and warehouses, they're dealing with boxes, for example, because medical supplies come in all shapes and sizes, you know, so there is non rigid objects to being some clear objects, flat objects, um, and, you know, so it it's a challenge. Um, you know, it has been a challenge for them to find bin picking vendors like us that can really get the job done and, and at the cycle times that they need. Um, but we are, we are doing that now in the medical industry. And, um, we're gonna be highlighting a lot of those capabilities over the next few months on our website with videos. So if you check out capsule, robotics.com, uh, we'll be updating that, uh, with our capabilities over the next two months. Great. Can you give
Speaker 1 00:14:52 Our audience a bit of an idea of what your products are? You've got cat send, pick, and cap send scanner, and so give us the overview of those.
Speaker 0 00:15:01 Sure. Yeah. So Capson pic is our software product, uh, and it, it comes on a computer. So it's not just a software. We ship on our computer that you plug into the devices in the cell, you plug it into the robot controller, um, you know, a PLC, if you like, uh, if you want to have a PLC control, the whole thing we shipped with the 3d cameras typically as well, and we control them through our software. Uh, and so Capson pick provides a software that, um, that both recognizes where the objects are such as the vision 3d vision for the cell, uh, as well as the motion planning to plan the motions for the robot to pick up the objects and manipulated them in different ways, uh, and also the control. So that's, that's one area where we're a little bit unique from our competitors right now is that, um, our, our computer doesn't just act as a resource and information resource for you to query about where objects are and how you might grasp them.
Speaker 0 00:16:06 Um, but we actually have a fairly advanced scripting language, which we call action scripts, which allows you to set up the whole tasks, the whole program for the robot picking cell in our software. Uh, and that's really what allows our software to take the most advantage of parallelization to do the things that I talked about before, about having the robot move at the same time as we're processing, uh, the images and motion planning and all of that. Um, and the other reason that we set it up that way is that, um, you know, one of the other things that we do better than anybody, and the only we're the only company that does a lot of this is being able to do things with the objects after you pick them up out of the bin. So, um, the hook picking example I've talked about before, we're picking up small metal hooks, uh, from a bin and they can tangle.
Speaker 0 00:16:58 So after you pick them up, you have to figure out how to disentangle them. Uh, and then they have to get inserted into a press that flattens one end it's part of a larger assembly task. And, um, you know, that, that process involves putting the hook down again and picking it up to make sure you have it by the right side to go into the press accurately. And so doing all those things, um, is really necessary and necessary to use our software, or I won't say necessary cause it could be done, you know, maybe with a separate piece of code by the integrator or something, but it it's ideal to do it all in our software because our software knows how the object is being held by the robot. And it can kind of go back and forth between the vision system and the motion planning system and the control system to accomplish these more, uh, complex, you know, manipulation tasks after you pick them up.
Speaker 0 00:17:52 And it's the same story with, uh, with packaging as well, like the med medical packaging applications that we've done, where, you know, after you pick it up, you have to now place it into a box, a kit, uh, in a very specific way. And, uh, so you want to, you know, verify how you're holding it and make sure that everything's aligned properly to get it in there very precisely. Um, and so having everything set up in our, in our software to control that whole process, uh, makes us able to achieve those more complicated manipulation tasks. So that's caption pick, um, that's our flagship product then caption scanner is just sort of a companion product right now that we, we will set up sell that separately. Um, but, but basically that's a 3d scanner scan table with the software that allows you to acquire new 3d models of objects, um, too important to caps and pick primarily, uh, so that if you don't have existing CAD models of your objects, you can acquire them easily.
Speaker 1 00:18:55 Thank you for that. So I'm kind of assuming that you are a fairly robot agnostic and are you seeing any trends in robots? Are they getting bigger or smaller or,
Speaker 0 00:19:05 Yeah, I mean, I think one of the trends is definitely smaller. We're seeing several new, small ones, um, one from academic that, uh, we like, uh, in particular. Um, but, but a lot of vendors have had smaller robots now, which is interesting because, um, you know, a lot of objects that, that, um, are kind of at the cutting edge of what's possible with Ben taking are also small and a small fasteners and washers and so on. Um, and so it is a good fit in that respect. And then you save the, um, you saved the footprint, you saved the space, uh, and you save some, save some money, right? So the price is coming down when you use these smaller robots. And then the other trend that we're seeing is just a lot of new, uh, new companies entering the robot arm markets. So you will have to wait and see how it all plays out. It might be some consolidation, um, you know, some winners and losers obviously, but, but there is a lot more competition than there used to be. And so that, you know, I think has already started to drive down prices a little bit and we'll see how far that, that, uh, can really go.
Speaker 1 00:20:10 So if I'm a potential client out there and I've, I've got, I'm thinking, wow, this would be a really good fit for Ben picking. What do I do? Do I call you? Do I send you my parts? Like how do I do that?
Speaker 0 00:20:21 Uh, yeah, I mean, you, you can call us our numbers on our website, um, you know, reach out to us. Uh, also we, like I said, before, we work with system integrators, we know our technology. And so, um, you know, reach out to your local system integrator, tell them you want to work with Capson robotics, uh, to do some Ben taking. And, uh, they'll, they'll talk to us and we can work out the details. You know, one, one of the things that, um, you know, it take a long time sometimes is, is for the end users to figure out, um, the partners that they want to work with on something like this. And so, um, you know, we recommend you just keep going with the partners you already have and, and tell them, Hey, I want to do some Ben picking and why don't you talk to captain robotics? And, and that's just a much easier way we think to, um, to get everybody aligned early on and, and, uh, be comfortable with the process because if you're bringing in, you know, two or three new vendors, uh, and then having to figure out not just what you're doing, but how you work together and everything that can take longer.
Speaker 1 00:21:24 Yes. It is kind of a long dance and it's always something I recommend to people to always, uh, start early, for sure. What do you see the future for Ben picking automation? Like, do you see better cameras, better software? What are some worse, some of the innovations going to happen the next bunch of years?
Speaker 0 00:21:42 Well, I think, um, you know, there are a number of companies like us that are innovating, um, the core technology, right? So, so everything you mentioned, so the 3d cameras, the end effectors the software, we're seeing innovations in all those fronts, um, from variety of companies, um, you know, I would look mainly to see it be easier to use, right? So, so at the moment it's still, um, you know, a fairly complex integration process, but, uh, you know, I, I would look for it to get more general, you know, for, for there to be more general and defectors more general software, uh, which were already sort of leading the effort on, on the software side at least, but there's still work to be done and for prices to come down, you know, prices of the 3d cameras and everything to come down. Um, I think that will certainly happen, but, but, uh, but basically for things to be, um, you know, a little bit more off the shelf and a little bit, uh, just, just quicker to deploy, um, because it's still, it's still, um, you know, I don't want to say it's a complicated process now, but it's, it's still something that needs to be thought out and planned.
Speaker 1 00:22:56 I wanted to ask you about the talent you're in a very competitive industry and region where talents at a premium, how do you attract the talent that you need?
Speaker 0 00:23:03 Uh, well, we, uh, we're lucky because we're in a, an area in Pittsburgh with a couple of top universities, Carnegie, Mellon, and university of Pittsburgh. And, uh, we also attract talent from the top schools all over the place because, uh, we're developing cutting edge technology, which is really exciting to a lot of recent graduates.
Speaker 1 00:23:23 I think you're right. I think, uh, grads want to work at exciting, interesting companies that are doing good work. So thank you, Jared. And how do, um, if someone wants to get in touch with you, what's the best way to get in touch with you at Capson?
Speaker 0 00:23:37 Well, the, the best way is to either reach out to us at our website, it capsule, robotics.com, uh, or to email us at info at caps, robotics.com, uh, and you know, in particular, we're really looking for people to tell us about their challenges, tell us what they're looking for, uh, because we're really looking to help solve problems, uh, and not just, you know, things that you may see in our videos, but, um, there's a wide range of capabilities that our software has. So even if you don't think it's a perfect fit for us, there's a good chance that we'll be able to help you solve it, or connect you with somebody who can solve it.
Speaker 1 00:24:14 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
[email protected] and Earhart is spelled E H R H a R D T. I'd like to thank and acknowledge our partner. <inaudible> the association for advancing automation. Three is the umbrella association for the R I a B a I a M CMA N 83, Mexico. And these four associations combined represent almost 1300 automation manufacturers, component suppliers, systems integrators, and users, research groups, and consulting firms throughout the world that are driving automation forward. I'd also like to thank our partner painted robot needed robot builds and integrates digital solutions.
Speaker 1 00:25:19 They're a web development firm that offers S E O and digital social marketing, and can set up and connect CRM and other ERP tools to unify marketing sales and operations, and
[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. We'll see you next time. Thanks for listening. Be safe out there. Today's podcast was produced by customer attraction, industrial marketing, and I thank my nephew, Chris gray for the music, Chris Colvin for audio production, my partner, Janet, and our partners, a three painted robot and our sponsor Earhart automation systems.
Speaker 2 00:26:03 <inaudible>.
