Winn talks with Josh Cloer, the Director of Sales for Mujin. Mujin is a robotics technology company that is changing the way that robots are programmed. Instead of programming the waypoints and the workflows like a human programmer would, Mujin is creating autonomous robots. They’re modeling the environment, telling the system about what the environment looks like, what the surroundings are, what the type of application is, what type of gripper and what type of robot, and ultimately allowing it to make decisions about how to solve the problem within the application.
Winn Hardin: [00:00:07] Hello everybody. I’m here in Boston at the Autonomous Mobile Robotics and Logistics Conference, which is co-located with the Vision show this year. And I happen to be lucky enough to be with Josh Cloer for Mujin, director of sales.
Josh Cloer: [00:00:17] Thanks for having me.
Winn Hardin: [00:00:18] Thanks Josh. This is our first time to meet. My first time to see your company. So let’s start off with just a little bit about Mujin before we go into the really cool demo that you got over at the booth.
Josh Cloer: [00:00:27] Yeah, absolutely. So Mujin is a robotics technology company. We’re really flipping the way you program robots on its head. So instead of directly programming the waypoints and the workflows with a human programmer for each application . . .
Winn Hardin: [00:00:42] Or pendant?
Josh Cloer: [00:00:42] Yeah, you know, a human controlling the pendant and telling the robot what to do. Basically, we are creating robots that are autonomous. So we’re modeling the environment, we’re telling the system about what the environment looks like, what the surroundings are, what the type of application is, what type of gripper and what type of robot, and ultimately allowing it to make decisions about how to solve the problem within the application. You know, maybe picking a case off of a pallet and putting it onto a conveyor depalletizing is what we’re showing here at the show. So doing that really autonomously is kind of what’s changing.
Winn Hardin: [00:01:16] Are you doing those as part of dropdown menus in your software department? Sorry that I haven’t been able to work in your environment, but are these software dropdowns, are we using CAD?
Josh Cloer: [00:01:24] Yes. So it’s like a model-based approach. So as the user, as the person that’s deploying the robot instead of, you know, creating waypoints and telling the robot what to do, you’re creating the environment. So you’re drawing the walls, you’re pulling a dropdown menu to choose what robot manufacturer, what model, what type of gripper are you using, and where is all that in space, in a CAD-like environment.
Winn Hardin: [00:01:47] So do you guys have an extensive API library right now?
Josh Cloer: [00:01:51] We do, absolutely. Yep. It’s to really connect to the system for things like analytics and operator-level control and interacting with other systems. You can communicate to us with an API or we can communicate across some of the standard Ethernet protocols that are out there.
Winn Hardin: [00:02:08] Awesome. So are you getting to the point when you’ve even got all the major EOTs like a Robotiq or a Soft Robotics or any of these others?
Josh Cloer: [00:02:14] So another thing that we do kind of as a necessary extension for some of the applications that we do, we make our own grippers typically. So we do use a lot of standard grippers out there in the market. We use a lot of standard components when we’re even making our own gripper. But one thing that we do that’s really interesting is we put force torque sensing into most of our grippers so that not only can we see it with some sort of vision system, but also we can kind of have tactile feedback and really confirm that what we saw, it feels like what we saw. So we have a lot of redundancy built in with that.
Winn Hardin: [00:02:48] Are we doing the same thing, either measuring, air compressor — because I see you’ve got the vacuum gripper, and I want to hear a lot more about that depalletizer unit you’ve got on there, so feel free to put both those together. Tell us about that awesome demo you’ve got.
Josh Cloer: [00:03:00] Yeah, sure. So we usually mostly use vacuum suction for a lot of our grippers. And yes, we are monitoring the vacuum pressure actively as we’re picking things, so we can kind of have a lot of tactile sensing there within the gripper. Actually, we’re here launching a product at the show called QuickBot. So QuickBot, the idea is that you can quickly deploy an entire work cell that’s focused on depalletizing. It comes on a skid, a big metal base plate. There’s a short piece of conveyor already within the cell. There’s some arms that fold out. So basically you can deploy this thing in a couple of hours without having to drill into the ground or connect it really to anything else. And because it’s depalletizing, there’s really no system-level integration either. You don’t need to send anything from the WCS or anything like that. Basically, you just bring a pallet and put it in front of it, and we figure out the rest.
Winn Hardin: [00:03:53] Is there any intent to maybe marry that with an AGV down the road so that we can go from truck to truck to truck to bay to bay?
Josh Cloer: [00:03:58] So actually we’re already doing that. We’re doing that today in the booth. One of our partners, RoBex, one of our integration partners, has brought along a mere 1350 and it’s bringing pallets to the robot cell to the QuickBot.
Winn Hardin: [00:04:09] Awesome. So the QuickBot, is that something that’s also going to be mobile in itself or retaskable?
Josh Cloer: [00:04:15] Yeah, it’s definitely redeployable. There are four pockets on the big metal base plate, so you can easily get it in to where you want to put it, and you could also move it to somewhere else. The biggest hurdle with installing it will be just dropping power and air and Ethernet to the cell.
Winn Hardin: [00:04:31] Right on. But if you’ve got quick connects right nearby, then those problems are not — so maybe the warehouse of the future. Absolutely. Let’s talk about a 50,000 foot — let me ask before we get into the 50,000 foot question, here in Boston, what’s the coolest thing you’ve seen on the show floor here? I know you’ve been kind of married into the work stuff, but I know you. Did you do a presentation?
Josh Cloer: [00:04:49] I did do a presentation, yep. So we spoke about removing barriers to automation in the warehouse space. And we’re kind of combining the integration company RoBex with Mujin. And Formic is another company that was on stage with us. So really turning this idea of the QuickBot that’s easy to deploy and can integrate with things like mobile robots into a RaaS model through Formic, who is a company that purchases automation equipment and then sells it as a robots-as-a-service model. So that’s probably, you know, selfishly, I think what we have out here is the coolest thing that’s on the show floor. I haven’t got to walk around too much to be fair. But you know we’re doing an actual live demonstration of a real-world application in this mixed-case depalletizing.
Winn Hardin: [00:05:38] Absolutely. I know you’ve been keeping the rest of the guys busy kind of reforming the pallets. And I think you added a chute to kind of . . .
Josh Cloer: [00:05:44] Yep. We shipped a gravity conveyor at the last second, so we only need one guy to be there now.
Winn Hardin: [00:05:49] I don’t even want to know what that will cost us. Let’s not even have that discussion. That would be a rush order here in Boston. Nothing cheap in Boston. So tell me a little bit more about the key markets that Mujin is targeting. I know we talked a little bit about logistics, general, but you see a lot of different subgroups within that.
Josh Cloer: [00:06:05] Yeah, absolutely. So the folks that are really seeing a lot of mixed pallet loads, and that’s primarily what we’re doing on the case-handling side. We also do piece picking, which is doing some order filling or putting stuff away into an ASRS. You know, the primary packaged goods side of side of the robotics world there. What we’re doing on the case side, what we see is it’s retailers that are receiving a bunch of different pallets and mixed pallets a lot of times from their suppliers. It’s those suppliers to retail locations that need to build mixed orders to go out to the retail distribution centers. It’s e-commerce companies that are shipping stuff to a UPS or a FedEx on a pallet to then go to the end location. And really the high-volume manufacturers — the consumer packaged goods companies, shampoo and razor blades and toilet paper, stuff that they make a lot of. And it comes in a box, and you need to move a lot of them. Well, that labor, it’s really backbreaking work to have to just move cases all day, so robotics is a great way to kind of resolve that.
Winn Hardin: [00:07:08] So are you leveraging machine vision as part of that?
Josh Cloer: [00:07:11] Absolutely. Perception is a huge piece to what we do. We have a vision system in almost every one of our applications that that we’re working on here. And what we do there is really have specific solutions within each application. So using things like 3D vision and 2D vision, using some of the best suppliers of hardware that are out there in that space. Three-D vision’s come a really long way in the last five years, and we’re leveraging all of that progression.
Winn Hardin: [00:07:38] Is that what’s really helping us go from cartonizers, where everything is the same? Because the demo that I’m seeing on your show floor, none of the boxes are like any of the other boxes, but you’re not missing a pick, and your throughput seems to be pretty solid. So is vision the key component there, the enabler that’s there? Or the programming itself also seemed to be a secret sauce.
Josh Cloer: [00:07:59] Yeah, I would say there’s two things. Yes, of course, vision is a huge part of it, right? Being able to actually get the 3D point cloud and do something with it, being able to get photorealistic and with the 3D point cloud together, combining that dataset to get the best perception of the environment of what you’re trying to pick is a huge part of that. I’d say the other piece that’s kind of a secret sauce for us is the motion-planning side of it. So being able to have the robot take the most efficient path every single time. We take over the servos of the robot maker, so we’re kind of bypassing all of their normal programming languages and really driving the kinematics of the robot ourself.
Winn Hardin: [00:08:39] How important is throughput to an application like that? Is that a spec that your customers are always asking: Where are we at in rate?
Josh Cloer: [00:08:45] Yeah. So it’s an important number within every proposal that goes out the door. I’ll tell you, outside of price, it’s probably the number one thing that gets negotiated in these things. And you’re looking to get somewhere around what a human can do. And humans are really good at doing this kind of stuff, but they’re not usually as good in the long run. So instantaneously a human can do some of the applications that we do a lot faster than the robot. But can that person do it for an eight-hour shift faster than the robot?
Winn Hardin: [00:09:14] The seventh hour, are you still you still moving as fast?
Josh Cloer: [00:09:16] Exactly. So we do some “race the robot” stuff around our office, and it’s always fun. And you can keep up and you can go faster than the robot. But then by the time you’re unloading even that second, third, fourth pallet, you’re starting to slow down a little bit.
Winn Hardin: [00:09:30] This is when we need the masseuse service robot.
Josh Cloer: [00:09:32] That’s right. That’s right.
Winn Hardin: [00:09:33] So the marriage of perception and machine vision, along with robotic handling and motion control planning, it seems like this is kind of the perfect show for Mujin. I mean, we’ve got Vision co-located with AMR. Are you guys just having scads of good conversations?
Josh Cloer: [00:09:49] Oh, absolutely. Absolutely. Yeah. There’s a lot a lot of great folks here that we are looking to partner with. We work with a lot of integration partners there to take our technology and turn it into a big solution. And then there’s also a lot of big end users here. You know, we had some of them talk upstairs. With Procter & Gamble and DHL there. So there’s a lot of good, good folks here that are trying to advance the cause with automation and really get the most throughput out of their warehouse. Right. That’s what they’re after.
Winn Hardin: [00:10:18] I believe you’re one of the few people that I’ve spoken to who’s really got a good solution for 3PL. I mean, when you hear about the retail side, we hear about e-commerce, there’s lots of barcodes, so many different solutions there. They’re high volume — I don’t want to say monolithic — but you know, again, where all the packages are about the same size within a given area. Is it that flexibility again that’s helping you get into that 3PL in additional segments and logistics?
Josh Cloer: [00:10:41] 3PL is definitely a target customer-partner for this kind of thing. I mean, they work on contracts typically with their customers, right? So when you have a solution like the QuickBot that we talked about, that you can pick it up and send it somewhere else, that really works out for them as they remake their warehouse to work with a new customer. Maybe they don’t have the contract anymore and they need to do something with that asset. They can easily pick it up and redeploy it. I think another thing that’s really important for them, because they work on contracts, is how they finance this stuff. And do they actually want to own the equipment or do they want to work with someone like Formic, that that we spoke with here at the show, to do more of a pay-for-use of the equipment? You know, you pay for an hourly rate for use, robots as a service. I think that’s another critical piece to remove barriers is: How do you buy it?
Winn Hardin: [00:11:35] You know, and that’s one of the most interesting offshoots of retasking robots. And we always talk about robots, future possible refurbishment, retasking to get even more ROI out of that asset expenditure. But if you can just take that cap-ex big expenditure right out of the loop, it’s definitely opening up the opportunities to so many more organizations who may not have giant capital expenditures but could benefit from an automated solution in material handling. That’s super cool..
Josh Cloer: [00:11:59] Absolutely. You think of it from the payback perspective. A lot of folks are looking in the two- to three-year payback range, hoping for that with cap-ex. But ultimately these systems, they can last 10, 15 years without too much maintenance. There’s really not a whole lot of maintenance outside of the friction that’s caused on the gripper. You need to maybe change some parts there over time. But ultimately this RaaS model is going to help you get the full payback in a very short amount of time and then reuse that asset for whatever you need it to use it for.
Winn Hardin: [00:12:31] Absolutely. And roll it out to many more facilities. If you’ve got multiple locations, you get a singular solution that you know it will be the same, wherever, whatever building, with the flexibility to adapt to existing conditions in that facility or that plant.
Josh Cloer: [00:12:42] Absolutely.
Winn Hardin: [00:12:42]. Josh can’t thank you enough for coming and talking with Manufacturing Matters today. I really appreciate your time. And if you’ve got any questions for Josh or Mujin, please just send us a quick email, give us a call over at Tech B2B. We’ll get back. Make sure to share and like this video so that we can continue to spread the word about why manufacturing matters here in North America and across the world. Josh, thanks again for your time.
Josh Cloer: [00:13:03] Thanks for having me.