Episode Transcript
Speaker 0 00:00:00 <inaudible> hello everyone. And welcome to the robot industry podcast. My name's Jim Beretta. And I'm your host. Our special guest today is Eric Nita is from plus one robotics. And I'd like to tell you a little bit about Eric during his summers. While in college, Eric worked in a factory that made cement mixers, and then finally a supervisor figured out that he might be able to fix the unused robot in the corner. And it happened to be an industrial robot. Eric then went to work for the robot manufacturer for 25 years in the training department and then managing field support staff and eventually working as the technology group director for the North American region fast forward a few years. And Eric is now CEO of a new company called plus one robotics. Welcome to the webinar, Eric. I wonder, did I get all that
Speaker 1 00:00:50 Correct? Uh, yeah, so that was, uh, thanks for the reminder of my days holding a 10 pound electric grinder over my head, you know, grinding down welds on cement mixers until somebody turned me loose on an industrial robot and kicked my career off. That was a good day. Now, granted, that was 1988, a long time ago, but robotics has been good, a good career for sure.
Speaker 0 00:01:14 And that's a good motivator, I think for you too, we're in a lot about robots, right? Is it makes your job easier?
Speaker 1 00:01:20 Yeah, I mean, once I figured out how to program that robot, you know, then I would just load a bunch of parts in some fixture and hit a green button. And then I got to sit there for 15 or 20 minutes. Well, it welded up whatever the widget was and I got a raise out of the deal. I'm like, this is the best college gig ever. Um, little did. I know it would be the start of, you know, effectively my life's work. So very grateful.
Speaker 0 00:01:44 Can you give me in the audience an overview of plus one robotics and kind of what areas that you operate?
Speaker 1 00:01:51 You bet. So I've been doing robots, you know, my entire adult life, right? So 30, some odd years now, but I'm really a people person and it's not that, you know, I'm a friendly chap. Although maybe aside from that, it's that I really believe people are exceptional. I am a human exceptionalist and plus one is just sort of the outgrowth of that. It's a recognition that robots are terrific, but what they're really valued for is their repeatability. And that's great in the factory, you know, where I've got to do the same thing, you know, ad nauseum for the next five, 10 years, but it doesn't work in the warehouse. You know, when I look at warehouse operations, everything is variable. You're not doing the same thing over and over again. I don't know what packages are going to come down the line next. I don't know what order Jim is going to put in and what, you know, skews, uh, the robot's going to have to handle.
Speaker 1 00:02:56 And this is where traditional robotics has broken because it is predicated on repeatability. But supply chain, warehouse automation is predicated on variability. And that's what really intrigued me about this market is, Hey, how can I teach that old dog new tricks? How can I take industrial robots that are so successful at building cars and make them equally successful at processing everybody's orders that we so desperately need now? And the answer is perception. You have to let the robot see the world. And that's what plus one did is we built the 3d vision stack for the warehouse and that's, what's become, you know, plus one is all about, is bringing industrial automation to the warehouse by taking industrial robots and adding really good 3d vision to it.
Speaker 2 00:03:55 And so what's the customer pain right now.
Speaker 1 00:03:58 The customer pain is us as the consumers, right? We're continuing to buy more and more of our, you know, goods, uh, over the internet. And so where we used to be our own shopper, we would go to the grocery store and pick the stuff off the shelf ourselves, and then check out now whether it's goods and services off your favorite e-comm site, or even grocery, we do more and more of this online. And so there's just more and more requirements of labor in warehouses, the growth of e-comm and the growth of the parcels associated with e-com have driven such a tight labor market in that space that they're really looking for automation to help them leverage the people they got because there's only so many people to go around. So the pain point is fold. One of them is just straight up labor scarcity. I don't have enough people to keep up with the growth of my business as a warehouse operator. One, the other problem is more acute and that is, Hey, in the COVID era, we have things working at cross-purposes. We have even more demand from the online experience, more people staying at home and shopping at home and the specter of fewer people being allowed to go to work. So how then is a warehouse supposed to provide us with the goods we desperately need in an environment where they may have even fewer people available to do it. So those are the pain points that a warehouses are dealing with in real time today.
Speaker 2 00:05:55 Tell me a little bit about supervised autonomy. What is it and why is it important and, and who is the crew
Speaker 1 00:06:01 Chief? Okay, so let's talk about plus one and our two sort of main products or offerings to the market. One of them is the vision system itself, and that is pick one. So pick one is a set of 3d sensors and an industrial computer with a high end GPU and its whole life is to look at whatever the input stream is, whether it's a cluttered pile of parcels from FedEx or it's totes full of, you know, items for a particular fulfillment center, whatever they might be pick one's job is to look at a clutter pile and go, huh? I think the next one that I should pick up is that one and it picks one out and it'll process it downstream. So pick one is this self-contained autonomous vision system that you can attach to an industrial robot. But what you're asking about is supervised autonomy, which is our second product, which is yonder.
Speaker 1 00:07:04 So yonder is this human in the loop piece. I said, I was a people person. And that what I mean by that is in the warehouse. People are so much more capable than any vision system or AI or model will be. We just have really good sensors attached to our head. So the vision system of pick one is capable and it's going to do the work and it's picking and it's placing, and it may do that 98, 98 and a half 99% of the time. But then once in a, while it encounters a scene, it simply doesn't understand. Maybe the scene is too cluttered. Maybe it's a new item. The robot has never experienced before. In either of those cases, the robot vision system does not have confidence in what to do next. And when that happens, that's when the other product comes in, which is a yonder.
Speaker 1 00:08:07 So yonder is this human in the loop piece where when the robot says I got nothing, it phones a friend. And what happens is we people, a crew chief is what we call them, gets a ding on their computer screen and up pops a robot somewhere and they go, Oh wow. That robot in Tallahassee is confused. Boy, I can sure. See why I would probably pick this one up right there if I were there. And you've now got put human eyes into the control loop and the robot says, thanks a lot. I'm going to go with that. And the robot goes back to work and we may not hear from that robot again for hours. But Jim, I got to tell you, we will hear from that robot again because variability is the rule in the warehouse. So supervised autonomy is just the technical term for this idea of knitting a human set of capabilities in when the autonomy fails.
Speaker 1 00:09:15 And that's what that's yonder is. And that's why we call it yonder because we do this remotely, when a robot needs help, where does it get it from over yonder, meaning at a distance, we do this. So I should tell you that, how does this pragmatically work? Right? So the robot is picking and placing like we talked about, and then it says, Hey, I need help. From the time it says, I need help. Till the time we get the signal, see what hap needs to happen, command the robot and have that robot back running again is under eight seconds. Wow. No matter how far away we are. So that's what a supervised autonomy means. And that's what is so helpful to these users is they know the robot will always keep
Speaker 0 00:10:07 And the crew chief is at your location. So how many crew chiefs per se dozen robots do you need?
Speaker 1 00:10:14 Yeah. So what you want is obviously crew chief leverage, right? These crew chiefs are very adept at their work so they can manage scores of robots. So it's, you know, easy to, uh, have a crew chief manage 20 plus robots at a time. And we have, you know, crew chief support 24 seven because our customers run 24 seven. You know, FedEx has been very public about their partnership with plus one. And you know that when is it that FedEx does most of their work? Well, it's obviously overnight. So they have, uh, robots that are running in their facilities now and crew chiefs that are supporting them.
Speaker 0 00:10:58 That's a good segue because I was going to ask you to tell our listeners about a project that you did recently with FedEx. What kind of ROI did they achieve and, and why did they do it?
Speaker 1 00:11:08 Yeah. So FedEx is overarching fundamental structural issue is there's not enough labor in the Memphis metropolitan area to keep up with their volume. FedEx is a thousand people short every single night in Memphis, a thousand people short that just gives you a sense of the scale of their operation. So, you know, ROI, of course you have to check the box to make sure that this makes sense from a dollars and cents perspective. But the truth of the matter is they need automation because they don't have the people available to do the volume of work that's before them on a night to night basis. So they've deployed four robots. They're very public that we're doing lots more after the first of the year. And we anticipate that that relationship will, will grow into the future. Uh, FedEx is, uh, uh, obviously a very good partner and their needs for vision guided manipulation, right?
Speaker 1 00:12:19 If you were to put kind of a circle around what plus one does for our customers is we give them vision guided manipulation robots that can see and react. Well, that's going to be more than just robots picking parcels out of a pile and putting them onto a conveyor, which is what we do with FedEx right now, because you can imagine 50 other ways that they need vision guided manipulation. And that's true, not only in the parcel space, but certainly an order fulfillment de palletizing order pack out. All of these are implementations of hand-eye coordination in the warehouse. And it's just that with plus one it's artificial eyes and mechanical hands.
Speaker 0 00:13:11 Now I forgot to ask you, should I ask you about the check one product that you have?
Speaker 1 00:13:16 Yeah. So check one is interesting. Check one was actually something that one of our customers asked us to build for them. In our typical applications, we do two sets of cameras. Well, two cameras, one camera is above the call. It the picking area and its job is to tell the robot which parcel or which item to go pick next. And then it picks it and then it places it on the outbound and the outbound might be another conveyor. The outbound might be outbound, maybe a sorter. The outbound might be another robot to mobile, whatever it is once we place it on the outbound, we then have a second camera above that outbound area to ensure we did it correctly. It's not just a matter of rate in the warehouse. It's not just how fast can you pick and place you better place them the right way.
Speaker 1 00:14:07 So that's what we call place verification. It's that second camera? Well, one of our users saw how effective that was, uh, with the robot and said, Hey, can I use that piece? Just the place verification piece elsewhere without a robot. And we were curious as to why they would want to do that, but they said, look, we still have other operations in our facility where we're doing manual handling and we still need to ensure that we're doing it the right way. So check one became a quality control system for manual operations, uh, in their facility.
Speaker 0 00:14:48 Well, that's very cool. So let's talk a bit about AI
Speaker 1 00:14:52 And vision. Yeah. This is really important. Our pick one software uses 2d, 3d and AI. It has multiple pipelines to determine which is the next best available pick. And like I said, if the universe of items that that robot needs to manage is somehow constrained, right? If it's a known universe, call it and doesn't have to be a small universe. It could be 2000 different items, but if that's all the robot needs to deal with, then the 2d, the 3d and the AI will together be successful in running that autonomously. We have customers today that use pick one by itself, on a robot and never do use a human in the loop. Why? Because their input stream is sufficiently constrained. But we talk about when it's not a constraint problem, you know, FedEx, doesn't get to pick what you ship. There's all kinds of stuff they have to deal with.
Speaker 1 00:16:03 So then the variability requires a human in the loop. What happened? AI broke down. And this is a fundamental argument that the market is having right now, which is what do you believe about the sufficiency of AI in the physical world? In the mess that warehouse is, can be, and there is a school of thought that, you know what the warehouse is just another data problem, feed the camera, enough images and the robot behavior that we want, the picking will be emergent, but robots will teach themselves. We'll all get to go home, turn out the lights and stuff will ship. But I told you, I am a human exceptionalist. I am convicted to my marrow as is plus one that people are better than AI. And your AI is going to not be able to keep up with the rate of change that the warehouse is going to throw at that robot and that's their world. So yes, AI is, is an important tool, but no, it is not sufficient. You need to have a way to bootstrap that AI through a human in the loop. And I'll say this, obviously every instance of a human crew, chief clicking on a scene because the robot was confused in some way, obviously that feeds our reinforcement learning and the AI that we will ultimately deploy on the next update. Right? So the system gets smarter through the crew chief plus AI interaction.
Speaker 0 00:17:51 I must admit that the crew chiefs are this huge resource for you, right? Because they can tell you what's happening on the ground through their own eyes.
Speaker 1 00:18:01 Yes. You know, the facility that we have as people to see and determine and decide happens in milliseconds, right? So for one crew chiefs made the system not only more available, meaning it never breaks down, right? I mean the only way a robot would stop when you have a crew, chief is if actually the gripper broke or there was some mechanical failure, or there was cleanup on aisle eight because you know, a package blew open. Then you get into the physics of the problem. But you know what, having a crew chief see that immediately, they say, yep, we need to alert somebody local because you know, a package burst open. And now we've got stuff all over the input stream. Physics happens. So crew chiefs are very useful this way, but obviously the main work that the crew chief does is really helping the robot, not when it's a physics problem, but when it's a logic problem, the, the robot vision system can't discriminate well in such and such a cluttered scene and the person does. And that feeds, you know, the AI model subsequent to that. So effectively it's the crew. Chief is your resource for uptime, availability, quality, and the future improvement of the system over time.
Speaker 0 00:19:31 That's very exciting. What do you see as the future for you and your part of the industry?
Speaker 1 00:19:37 Wow. There is no lack of need for vision guided manipulation in supply chain. You know, you go to any warehouse in this country and there is a sign outside that says help wanted the growth is such that there is a lot of headroom for the work that we do. And it's not a robots versus jobs question. There aren't enough people as it is. I'll tell you this, no robot ever deployed in a warehouse resulted in a pink slip. There's not enough labor it's that the labor is so valued that we better put the people to go do these tasks that our robot has no chance of doing right. I'm thinking about the outbound truck problem. You know, loading trucks is hugely labor intensive, and it's probably going to stay that way, you know, for the next five years, for sure, because you know, when you have to load a truck, you have to play Tetris in 3d in real time. And some of the items coming at you to load aren't even rigid, where am I going to put a bean bag chair right now? These are the sorts of decisions that it takes humans to make in real time to do that work. You're a long way from having a robot load you a truck right now. So what it means is for the future, as robots continue to take on more of the mundane tasks in the warehouse, you can release the people to go do more value added, work like that.
Speaker 0 00:21:20 And that's a very exciting, I'm wondering if we missed anything in our conversation today.
Speaker 1 00:21:27 Yes. I would argue that supply chain automation is a lot different than automation in manufacturing or automotive. And here's how it's different. You know, if you take and sell a project to GM and it turns out it was a lousy implementation of robotics and they strip it out six months later, I'll tell you what GM is still going to buy a whole bunch of robots next year, right? It is a sufficiently mature market that they know that you can have a bad job and it's not going to wreck the way they view automation. That's not true in supply chain right now. You better get it right? Because when you put a robot into a warehouse, almost in every instance, Jim, it is the first robot they've ever had. That was even true of FedEx. These are the first four robot arms FedEx ever installed. That's crazy.
Speaker 1 00:22:33 You know, so if you get it wrong, you're going to ruin a facility for years, right? You can't afford to get it wrong. They're putting out a lot of their enterprise risk to try this. So what does that mean? You had better partner with your customers. You cannot throw technology over the wall in supply chain right now, if you have a client that wants a robot in their warehouse, you're going to need to come alongside them. And you're going to have to work with them through, you know, that first installation until it's productive and meeting the metrics and being a real win for O E. And then, you know, that you've succeeded. This is not a PO business. You've got to help them be successful. Then you're sticky.
Speaker 0 00:23:33 And this business is now seconds matter. Right? So it's time is of the essence.
Speaker 1 00:23:39 Yes. You know, uh, our applications are hundreds of pigs per you know, I keep referencing FedEx only because they're so public about the work they do with us, you know, it's 1200, 1300 pigs per hour, right? So yes, seconds don't matter. Milliseconds matter when you're talking about rates like that. So yes, it's important for them to get the throughput. It's important for them to make sure that throughput is at quality. And it's important for them to know that the system will keep running in the event of exceptions or edge cases. And it's really important for them to know that they've got the support from a reliable partner. All of those things have to be true for supply chain to meet the promise of being the next major vertical to adopt industrial automation at scale. And that's what plus one is committed to. And that's
Speaker 3 00:24:41 Great. And I think that can be a wrap for today. How can people get ahold of you, Eric?
Speaker 1 00:24:46 Plus one robotics.com will be your main gateway to all things, a plus one, or you can follow us on LinkedIn. And you'll always know us because our hashtag is robots. Work people rule. If you see it on Twitter, you see it on LinkedIn. It's because that's plus one. And that's the way we view the world. Robots are great.
Speaker 3 00:25:09 People are better. Thank you, Eric. And I'd like to thank and acknowledge our partner. <inaudible> the association for advancing automation. 83 is the umbrella association for the RIA AIA, M C M a N 83, Mexico. These four associations combined represent almost 1300 automation manufacturers, components, 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 painted robot cells and integrates digital solutions. They're a web development firm that offers SEO and digital social marketing and can set up and connect CRM and other ERP tools to unify marketing sales and operations. And if you do it to get in touch with us@therobotindustrypodcastoremailistherobotindustrypodcastsatgmail.com or 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'd like to thank my nephew, Chris gray for the music, Chris Coleman for audio production, my partner, Janet, and our sponsors <inaudible> and painted robot.
Speaker 4 00:26:20 <inaudible>.
