Developments in Robotic Warehouse Automation - Scaling Physical AI with Yaskawa and Mujin

Welcome to today's Yuskawa Moto Man webinar. We're glad you joined us. My name is Mike and I'm part of the marketing team here at Yuskawa Motoman. Today's webinar is development in robotic warehouse automation scaling physical AI with Yuskawa and Mujin. And now today's presenters.

Mario D Cruz is vice president of marketing and strategy at Mujin where he leads marketing strategy positioning and portfolio eval evolution for advanced robotics and automation solutions. In his role, Mario works closely with corporate leadership to sharpen the company's strategic vision and market positioning of Mujin solutions to meet the rapidly changing demands of modern supply chains. He brings more than 20 years of experience in strategic consulting, having advised global companies including Swiss Log, Honeywell, IBM, Dell, Hulip Packard, and uh, how do you say it? Hawaii industries.

>> Huawei. It's a Chinese company.

>> Sorry, I should have checked that beforehand. Huawei technologies. Throughout his career, uh Mario has helped technology organizations define markets, accelerate growth, and align solutions with evolving supply chain demands.

Next, uh, Jacob Dillingham is a senior solutions engineer with Mujin with more than eight years of experience in material handling and warehouse automation. A mechanical engineering graduate of Arkansas State University, Jake has designed and led large-scale automation projects valued at up to 150 million uh, delivering end-to-end material handling systems from inbound through outbound operations. At Mujin, he focuses on integrating robotics and advanced automation technologies into cohesive customer solutions, leveraging Mujin OS as a universal software platform to connect and optimize diverse automation systems.

And opening today's discussion is our very own Chris Caldwell. Chris is Yuskawa's senior product manager for material handling. He has over 20 years in manufacturing operations. He has worked in every corner of organizations to provide exceptional products and services while optimizing costs, maintaining reliability, and ensuring quality.

Remember, this is a live Q&A session. As you have questions during the presentation, please drop them in the chat. We'll address them during the Q&A at the end. And now, developments and robotic warehouse automation. Chris.

>> All right. Well, first I just wanted to say that I'm excited to have this discussion. Uh it's great to have a strong partnership like we have with Mushian and two companies that align on such a strategic opportunity as the logistics market. Uh just a little bit as what we're going to cover today. We're going to spend just a little bit of time as I'll advance this uh talking about what the overall logistics industry is. I'll go through a little bit of a definition of that and we'll also get into the e-market metrics. Uh after we talk about those, we'll just talk just broadly about where we think the entire industry is going. Uh and then uh as I wrap up, I'll hand it over to Mario. We've got a new product that Musha and Yascow are collaborating on and he'll introduce that and I'll talk a little bit more in depth on that. Uh then once we get there, we'll have Jacob talk just a bit about their platform overall and we'll go through a case study to really see how our two companies have been able to align to meet customer needs. Uh so he has a a pretty detailed case study there.

Uh getting started here, logistics, uh one of the important things is just to define and understand what we're talking about. Predominantly automation has been through history focused a lot on the manufacturing side of things and so making goods uh getting those goods through production and ready to market. Uh and it's really been in the last 15 or so years that focus has been on getting those products out of warehouses and to the end users whether that be to warehouses to distribution centers uh to case packers or through any type of other thing. That's what we're really focusing on when we're talking about logistics. Uh it's the post-manufacturing delivery to end users. Uh one of the things that comes up with that is there is a significantly different set of needs and demands for the market. So while traditional manufacturing has a large amount of high volume low mix uh when you get into the logistics everything is very high mix. Uh you're moving from single skew into mixed skew and have significant changes in terms of the flexibility that the automation systems require. Uh looking at where we fit into all of this. Uh I have kind of a chart up there and we can see that as you switch from manufacturing logistics a lot of that is the single case palletizing certainly has a home there but uh where this conversation is going to be predicated is as you get into the transport and distribution side of things where you're taking those full pallets and having facilities that specifically strip those products off combine them into mixed case pallets and or uh combine those mixed cases to recreate those products.

The market size overall, there's a lot of growth coming in this market. So, it's estimated to be between 9 to 12 billion in 2025. Uh the growth rate has been in those low teens for the last five years. Looking forward to the next five years, it's projected to be into the higher mid-teens. Uh just under 20% total. So this is a large market, a large demand for automation and it's only been growing. Uh some of the reasons for that growth have specifically been the significant increase in e-commerce. Uh in the post-pandemic world, the labor constraints, a lot of these jobs had previously been very manual. Uh moving those into automated allows companies to scale better, grow better, and just solve some of those labor challenges. And then also customers as they've become more adept at just adding to cart and expecting delivery their expectation is to have a faster shorter delivery cycle. Uh with that the overall US economy is spending just under $2 trillion on transportation logistics and at about a trillion and a half on e-commerce. Those are the fuel that are justifying all of this automation and a lot of the investment and AI based solutions such as the one we're going to be talking about today. Uh putting the kind of bow on the overall market, we have just over 15 and a half billion square feet of warehouses in the Americas. A lot of that footprint uh it's getting more and more difficult to find the employees to fill that. And so being able to come up with solutions that meet our customers needs and keep our products in motion is something that both of our companies align on.

Uh looking at it, we view it. Both of our companies see the logistics industry as a very strategic growth opportunity. Uh one of the reasons for that as a robot company, there are a lot of different types of automation available. Uh there's a lot of specific robotic applications when we talk about parcel induction kitting uh packaging and then mixed case palletizing depalitizing uh but then there's also a lot of integration and solutions that require an additional software layer uh that's when you start adding in your AMRs AGVs and really relying on that physical AI to be able to turn the uh concept and material flow into a turnkey solution. Uh so with that we we have a nice full line of traditional industrial and collaborative robots but those only go so far and then after we get past the point where you need flexibility and adaptive programming uh technology partners such as Musen provide that vision and that robotic guidance to help us meet the needs of a changing economy. Uh so with that I'm going to turn it over to Mario. He's going to discuss kind of the foundation of our partnership, the way that we've opened up our robot controller, and this latest collaboration that the two of our companies have.

>> Sounds good. Thanks, Chris, and thanks to everyone for joining us today. Uh we really appreciate you taking taking the time to spend part of your day with us on this webinar. Um in today's industrial automation world the landscape the the the autom the technology is evolving extremely rapidly but what drives successful deployments is strong collaboration between technology partners. It's actually very rare to see partnerships in this industry that last for many years and continue to improve over time. Now we are really proud to be working with Yuskava Motorman a company with more than 115 years of industrial automation leadership. Yesava has been a pioneer in automation for decades and is known to be known for building some of the most reliable industrial robots in the world. In fact, they were the they were responsible for developing Japan's first fully electric industrial robot and that has helped shape the the modern uh the modern robotics industry.

Now, the other half of this partnership is Mujin, an industrial automation software company that is focused on what we call physical AI for robotics. Now, what exactly does that mean? Physical AI is kind of a buzzword that everybody uses today, right? But when you when we look at physical AI, we talk about a platform that allows robots to understand and react to real world variability. And let me repeat that, understand and react to real world variability so that they can handle complex logistic tasks that traditional automation struggles with. Now what do I mean when I say understand and react to real world variability? You know, unlike traditional automation, the Mujin platform does not rely on fixed programming or skew teaching. So every time you have to make a change, you don't need a robotic engineer to come in and kind of help you with that. Whether it's vision, whether it's uh whether it's motion planning or even even programming new SKUs. It uses real-time perception and dynamic motion planning to handle variability as it happens in real time.

Now what happens when you combine a world-class uh robotic hardware from Mujin and uh uh sorry from Vascava and with Mujun's intelligent automation software you get a system that can handle the real complexity of modern warehouse operations and on this platform you can build a number of different applications you know ranging from mix palletizing to depalitizing which is extremely complex where every case, perhaps every pallet and even every order can be different. And this relationship is not new for us. Over the past 15 years, our teams have worked together in 10 different countries and we have more than 30 successful deployments. Now this experience has allowed us to refine how we design, deploy and operate these systems so that our customers can get up and running faster, quicker with a and with a high level of precision. And this collaboration has become a trusted approach for companies that are looking to modernize their logistics operations with industrial grade uh with industrial grade automation.

So we I've talked a lot about this, but now let's kind of look at an actual let's stop talking about technology and let's let's show you what it looks like in practice. When it comes to mixed case palletizing is perhaps one of the most challenging areas when it comes to warehouse automation. Actually in many facilities this process is still highly manual primary because handling products of different shapes, different sizes, different weights in constantly changing order profiles and this is becomes really difficult to uh automate using traditional systems and the reliance on on manual labor actually uh creates several operational issues. Chris alluded to the labor shortage. You know, it's very difficult to find staff to um to fill these positions in a warehouse. As well as the cost of labor seems to be increasing in many states. You know, they've increased the cost of labor that tends to squeeze your margins. When you look at repetitive lifting and awkward movements increase the the risk of workplace injuries as well. And then manual stacking. When you manually stack these pallets, it leads to inefficient pallet utilization and that increases your transportation costs as well. And as auto volumes continue to grow, as e-commerce continues to grow, um you know, rainbow pallets out to uh to different stores, manual palletizing just does not scale efficiently.

Now, on top of that, the warehouse environment is extremely dynamic. Product flows are constantly changing. Uh auto priorities shift throughout the day and traditional palletizing systems simply lack the adaptability to respond to these changes in real time. And what Chris is showing right now in in the video, this represents 15 years of collaboration between Yuskava and Mugen and what it actually produces in practice. Now over time our teams have refined how robotic hardware, vision systems, intelligent automation software work together to solve some of the most complex problems when it comes to palletizing and warehouse operations. And the outcome of that is the of that collaboration is packmaster powered by Mujin OS. It's a mixed case palletizing solution designed to operate efficiently in a high-mix warehouse uh where skew variability is the norm. And as you can see in the video, the the Muja the pallet is built using Mujun software and the platform packmaster platform moves down and you're able to shrink wrap that so it's ready to u to to be sent to transportation or to get out of your warehouse. Um, Mujin combines, you know, the advanced vision, integrated motion planning, and the software to enable consistent high-speed palletizing even when product dimensions, weights, and perhaps even packaging formats are constantly changing. And the result of this is a dense is densely packed you know very stable plat uh pallets with a high throughput allowing warehouse operations to maintain the speed without sacrificing pallet quality or even the stability of the p pallets which is something that is really important that operations looks at.

So I'll hand it over to Chris to talk a little bit more about packmaster and u um and Chris over to you now.

>> Thank you. So one of the things the the first thing is what is packmaster and packmaster is the result as Mario said of multiple years of collaboration uh as well as a global attack on Yuskawa's front to really address the needs of the logistics market. Uh so at its core packmaster is a multi-level mixed case palletizing solution. Uh it combines robots uh the mezzanine pallet lift uh some product based conveyors, special grippers and the pallet wrapper as well as all programming controls and software uh to be able to accept the inputs from a facility that'll be some type of material conveyance solution. Uh combine all those sequence products into a densely uh stacked, tightly wound pallet and deliver those pallets out to pallet conveyors to be loaded onto another uh either truck or into a storage location. Uh we have the two different versions that we're offering to the market here. The single robot uh with 4 to 600 cases per hour and the dual robot with 800 to,200 cases per hour. Uh the entire package here gets brought together to meet the needs of an industry with significantly increasing demands. So in the past where it was uh fairly simple to have single case palletizing or you could have homogeneous layers uh with different products by layer as demands have gone on up and up. We have to really be able to adapt and overcome whatever is thrown at us. And this type of solution and collaboration between our two companies really results in a turnkey solution to one of warehousing's toughest applications. Uh the way that we lay there is really a payoff and trust of just over a decade of coordination between our two companies. Uh pictured here in the center is uh Ross Dancuff, the founder and CEO of Mujin. He visited and traveled to Japan. Just to his right there in kind of the the light shirt is Yuskawa's uh representative director and president Masaherro GAwa. Currently leading all of Yuskawa at the time he was just in charge of the robot group and those two got together about 15 years ago and had a meeting to really discuss what the future of robotics and what AI robotics would look like. The result of that meeting was Yuskow being the first company to really open up the robot controller and provide what at the time was very unprecedented access and direct control of our robot to Musen's platform. I saw in the comments as things were rolling through just talking about the the fluid movement of the robot and that is something that as you work in the robotic industry you really understand the kinematic and the motion planning that each company brings. Each robot OEM has their own special version of that. But at Yuscow, we give motion full kinematic control over the robot. They're controlling things at a joint angle and torque-based uh level so that they're able to take complete control of that robot and eliminate any kind of discrepancy that they would have on their motion planning versus the actual robot's execution. That type of forwardinking look was something that the industry did not have at the time. Pretty much every other robot OEM was a very closed environment uh giving very rigid tools to partners and not really embracing what AI could offer to the robotic industry. With that we've been able to hand over and get full optimized motor level control to Muian. And this relationship has been very deep. It's resulted in the development of very application and industry specific arms that we have launched specifically for projects to where we've coordinated with Mujian globally. So what does that look like today? Right now we have what is the most battle tested proven in the industry implementation of physical AI powered industrial robots. Going back to 2016, so 10 years ago, Mujun made the decision to initially standardize on Yuskawa's platform and we had a significant amount of success in all regions of the globe taking over large opportunities in the logistics and warehousing industry. A lot of that was very specifically due to that shared vision and Yuskawa embracing the ideas and concept of physical AI transforming industrial robots. What does it look like moving forward as we're looking into the evolution of Mushion's platform and different places that it can be deployed? We're looking for more and more areas of success and we already can look back and lean on that battle tested physical AI implementation. To talk a little bit more about that, I'm going to hand it over to Jacob. He's going to talk about the core motion OS functionality and capability and then he has another case study to wrap up and talk about the real world results that our two companies have collaborated on.

>> Yeah, thank you Chris. Um yeah, just to kind of start out here, um Mugen has been around for about 14 years now. And just this slide, while it does explain kind of what all we we like to do, it also kind of explains our journey as well. Where we started out to where we're currently going. Um starting out, you know, very humble, basic, simple, trying to just solve the problem of robotics and intelligent robotics. And so for that we partnered with companies like Yuskawa to you know open up their platform to us, open up their robotic software to us to allow us to control their robot arms at a very like servo based level. Um and through that you know we we gained immense knowledge on how production systems work and you know furthered our technology and our capabilities. But you know, it ultimately not enough right? Just understanding in how to control a robot is not enough to provide a full turnkey production system and so for that we started having to integrate out to the the different components that might exist within a a robot cell whether that's vision sensors to detect uh changes within the environment or the different grippers and different gripper technologies that allow us to you know pick different products in different ways or you know safety sensor and light curtains that allow us to understand the the safety aspect and the interaction between human and the robot cell or barcode readers that allow us to kind of gather information from the products uh that that are being presented or picked by the robot cell and you know that really comes to the full robot cell. We understand and know how to utilize everything within the robot cell as a whole but even then that that's that's not enough right? We need to understand how to efficiently and effectively provide product and take away product from that robot cell. And that's where we expanded kind of beyond the robot cell to working with different AGV companies to control the AGVs that deliver pallets or cases to and from the robot cell. Or maybe it's a you know CNC machine for us to do machine tending or pallet ASRS's for kind of high density deep storage automated storage and retrieval system or just simple or complex uh conveyor systems that allow us to present new product or take away product from the ro robot. So it's all as an automation solution as an automated system all of those things need to interact and integrate very finely. and efficiently. And it's important to be able to do that and control that interaction on one platform. So it's not you having to learn how to work with, you know, a Yuskawa robot and then you have to know how to, you know, do a PLC uh ladder program to control the the the gripper and then learning how to, you know, communicate with a six scan tunnel. It's like you don't have to do that. You just know how to use our program, our platform, and we give you access through a easy-to-use UI that allows you to utilize all of those automation platforms. So, we can go on to the next one.

Um, so now I'll kind of talk more about the the technology behind the scenes, right? Um, so at at the core of it is the digital twin, right? That this is a virtual representation of the real world. It's very important to understand with the Mugen platform that the robot does not react unless the the digital twin is updated to match the real world. If the box exists in the real world, then it needs to exist in the digital twin. And that's where we start leveraging, you know, vision systems or simple sensors um or perception systems over there that that allow us to update uh a dynamically changing environment, right? Whether it's creating new cases within the robot cell, then we could start doing the motion planning and orchestration that allow us to identify, okay, I know what the ultimate goal is. I know the pallet that I need to build. What is the best way for me to build that pallet? Well, I have this case in front of me. I need to figure out how to grab that case. Right? Our system on the fly in real time figures out how to do all of these things without any kind of you know prior programming that that needs or understanding of robotic kinematics inverse kinematics or anything like that that needs to be known by the user. And so through all that there's a bunch of different layers within our system that that need to be controlled in sequence. And that's where we get to kind of our orchestration layers that kind of take that ownership of controlling those different uh parts of the cell whether it's the gripper or the robot specifically or the conveyor system. All of that needs to be orchestrated in a harmonious system um to operate efficiently. And so that's where we have our orchestration layer and then you know the control layer that allows us to manage communication with those devices and make sure that we're efficiently and effectively communicating and that if any wires are getting damaged or if you know the the Wi-Fi goes out that allows that allows us to communicate with the AGBs having that understanding within the platform allows us to monitor that in real time and provide errors at the point of the error occurring. Um, and then all of that is more like the integration layer. And then there's different plugins that allow you to visualize a lot of this. Um, so whether it's fleet manager, you can see all of the AGVs within the fleet, all the movements that they're making, how you know the one AGV is crossing paths with another and how it has to wait for that AGV and just uh optimizing that motion planning or parts manager where you can induct inject new SKs into the system um within under a minute, right? You can program a new skew to be handled by this system. Um, and you know, it's ready to go within a minute and all you have to do is, you know, the go through a simple UI workflow that allows you to enable that. Or, you know, if you wanna I if the the system is having a difficult time with, you know, certain types of product, certain package typing, maybe you want to go to the the vision manager and look and actually see what the what the vision system is seeing, see the point cloud that it's gathering, see the detections that are made within that system. Um, you can get to that granular level data if you so please. So, all of this is possible and controlled through the Mugen platform. And so that brings us to kind of the user interface, right? What what are you going to see as a user of the software platform, right? And that's where we get to our web-based user interface, uh commonly called web UI. Um, so this is what you'll see if you pick up the pendant at the robot cell or if you're on the network, you can access it through your local laptop. and it gives you an easy-to-use platform for setting up, configuring, running this system as well as you know just generally operating this system. Um, so down in the bottom left you kind of see the the scene what as we call it. This is the environment around the robot that gives it the understanding of where obstacles are within uh the robot's reach that it needs to motion plan and uh avoid right whenever it does it its trajectory planning. It's very important that you design this scene to as high of an accuracy as you can, right? That so that the robot knows where every corner is, where every nook and cranny of the system is. Up in the top right is again going back to that vision manager. You'll be able to see all of what the robot sees. Every 2D color image, every black and white image, every point cloud for every pick that it captures, we'll keep a historical record of that. that allows you, okay, that that, you know, three hours ago a problem occurred, right? That that we couldn't pick something. Why why did that happen? Well, you can dig in and see, okay, well, the the labels on this package were messed up that that, you know, didn't allow it to be detected. So, that that's a benefit of our system is we track all of that information over time. So you have that historical analysis and ultimately what that comes into is you know probably the most important to any kind of manager of a of a facility operations is understanding how the robot is performing and how it's performed over time. And since we collect all of that data from not just our system but all the external systems that pass information to us, we can pile that compile that into uh reports or graphs or charts that show the performance of the system. How how many cases uh it is putting out through per day, how many pallets it completed in the past hour, how many errors the downtime for those errors, how many you know multi-picks were performed. All of that can be monitored within our system as well as you know having some root cause analysis there where you can dig in a little bit deeper. If you identify a problem and want to know more, you can get as granular as you want down to, you know, individual device IO signals and timestamps. So, um, a really comprehensive platform here that that really allows you to either be as high level as you want or be as granular as you as you, um, are comfortable going. So, that that takes us to the next one. Um, what how does this all culminate into you know the the creme de la creme of a of a solution right and what I would say is robotic case pick is that it is the culmination of all of that uh 14 years of development and partnerships with Yuskawa and HV manufacturers and conveyor providers to provide a comprehensive s system for building these mixed skew pallets for outbound storage replenishment and so what What we do is we provide a robot cell that integrated with an HV system that and a mass storage system that delivers single skew pallets to the robot and we'll design the pack in a stable and you know uh structured manner that we're going to build and we'll pick from those single skew pallets to build that mix skew outbound replenishment pallet um you know automatically you know it's not an operator coming in and building this while um so and it's been very beneficial in a couple key uh markets, specifically third-party logistics and food and beverage, but we've also seen extreme benefits within, you know, consumer goods and retail and e-commerce. Uh having this kind of flexible, adaptable system that can grow with you, you can expand the fleet, you can add robots to the system um over time or you can introduce new SKUs, right? You know, the these large conglomerates, they like to buy and sell companies uh over time, which means new SKs come into the building, old SKs go out of the building, some stay forever, right? So, having the ability to easily uh inject new SKs into the system and have it just work is very important. and and um that that's kind of what you get with our platform, an easy-to-use uh platform that you know we work with you to train you on on you know how to use it effectively and how to manage that operation. Um, and so next we'll go into kind of a case study uh around uh Tresco Nakayyama. Um, so they are a leading uh industrial tools and materials distributor. Um, they, you know, they have a a a distribution center uh uh Planet Saitama that is kind of like their showcase, their center of excellence of all of the the automation that they want to bring to uh their their distribution network. And so the the problem there uh that that we solved for Trusco was random mix skew palletizing, right? That's a very difficult problem and you know it if you say it to any one customer everybody's going to give you different problems that need to be solved by random mix skew palletizing and what what we worked with them to do is to okay infinitely random SKs infinitely palletized at a high rate you know there there's not a company in the world that can solve that for you but what we can do is give you a platform and and work with you to identify what can be controlled within your operations if we work with you to understand your operate your current operations right what what are okay okay maybe it's not infinitely random SKs but maybe there's a couple very well-defined SKs that helps us kind of limit that that infinite possibility down to a set range of possibilities and so within that we work to identify you know three skew sizes so one of them uh three skew unique skew dimensions so they have length, width, and height. They have variable height, but they have three unique uh length, width um combinations. And so for that, hey, that that really starts to limit the scope of possibilities that we have to plan for at the point of picking. Um, and so we identified those three and we started working with them on an algorithm that allows us to do this random on-the-fly palletizing at the point that it's presented to the robot cell um which you know uh with the the logic that we created we basically make like a tiered uh stack um we build in a tiered way so that we can um always have a presentation available for the robot to place to um and regardless of what height skew comes next right there there's always a logical comparative process to go through to identify the ideal place to put the next case right and that that that's part of the benefit of of the Mugen platform is while yes it's easy to use it is also highly customizable if needed and those random mixed skew palletizing applications tend to need that that high customizability right they there there's no one shoe fits all for for these type of applications and so you see in the video down there in the left you'll kind of see uh in in the examples what I'm talking about with that tiered structure of the pallets um you know it it really allows us to simplify that packing algorithm and that motion planning that we need to do in a fairly quick cycle time. You know, you're looking at, you know, 7 to 8 second cycle times. And within that, we need to, you know, move new SKs into the position into the pick position. We need to figure out what that skew is. And we need to figure out what to do with it. Uh, all without seeing any kind of slowdown or stoppage of the robot. Um, and part of that, as you see, we have AGVs delivering the pallets. Well, we need to know where the pallet is. They're on casters. We can't say that within plus or minus 10 millimeters they're going to be in this position always. So, we have to incorporate a vision system onto the gripper itself and allow us to take a a snapshot of the of the pallet being delivered so that we know exactly where it is in reference to the robot so that we build a stable pallet that is centered on that pallet. So um yeah and and so that that's Tresco Nakayyama. Um, you know, some some overall outcomes of that was you know they they we were able to double their order volume with with the same workforce. You know, the the the Japan network uh experiences different um you know troubles in in the workforce than what maybe the US looks for. But you know uh overall having double the efficiency of your workforce is a net benefit no matter where you are. You know, but what allowed us to do that for them was having a a being able to integrate multi-pick on the fly opportunistic as the the opportunity presented itself in in this random operation. You know, we were able to identify those opportunities and perform multi-picks um as needed. It's not guaranteed. It's not every time, but you know, when it's there, when the opportunity is there and presents itself, you want a system that can do it, right? That just gives a little added boost to to the throughput. And then um you know, having a a uh some kind of way to intelligently and quickly identify the the stack that we need to be building that, you know, doesn't take too long to compute, but also builds a stable pack is kind of the constant battle that we have to fight here. And so, you know, we were able to work with Tresco and uh and identify what really matters to them and kind of, you know, simplify that that uh that computation that needed to take place. And, you know, down at the the bottom you see kind of the overall output of that the the performance 500 cases an hour. You know, deployed the system within six weeks, which is, you know, a crazy kind of turnaround time for a system of that complexity with AGVs with, you know, the the cases, random mixed skew cases coming into the cell um and then ultimately achieving an error rate within uh less than uh 0.05% within two weeks of production. So um you know o overall a very successful um solution uh for both parties there um and has continued to be a the kind of pinnacle or you know what what what can be achieved through Mugen OS to to solve the problem of mix skew palletizing.

>> All right. Well thanks for that. Uh just before we get to the questions and one more call for questions as we're getting ready to go through to that I did want to toss out there that both Minion and Yuskawa will be at Modex next month. So just under a month actually will be in Atlanta, Georgia. Uh I've got here on the screen that Mush will be in booth 9519. We'll be in booth uh 13550 566. Uh you can certainly stop by our booth, learn a little bit more about Packmaster. We'll have a a video feature there to showcase a lot of its capabilities and we'll have our staff on site to be able to talk through that. And also if you stop by Muian's booth, you can learn more about their uh operating system, their physical AI, their vision system, and how they coordinate robots and meet customers needs. With that, I'm going to go ahead and switch over to questions. Looks like the first question that came in was uh talking about packmaster asking on those mixed case uh pallets that it was making. Is that uh palletized to ship to stores or to store in the warehouse? Uh I can go ahead and field that one. Typically, Packmaster is going to be in a facility uh pulling product out of the pallets that are stored in the warehouse uh to prepare them to ship out to uh either a retail store or going from the manufacturer into their distribution centers. Uh so, a lot of times places that can use that are going to be much more of the retail front where they need the uh multiple skew uh mix there and they build those out. A lot of grocery stores will do things like that or be food and beverage companies. Uh additionally, some of the distribution companies like Granger, Fastenol, large houses that have a lot of different products when they're uh filling from their master distribution centers into their local stores, they'll build very similar pallets to that.

The next question which he he had noted with that it was answered but we can probably talk about it just a little bit more that there was a lot of AMRs on the Muian slides. Can you give examples and kind of discuss through kind of how Mujian OS coordinates those AMRs and keeps track of the goods through the warehouse and the both goods to robot and flowing that through.

>> Yeah. Yeah. Uh I think I can speak to that a little bit. So, you know, we have our own fleet manager application that allows us to control the movements and and plan the movements of the AGVs themselves. But, you know, that that that's only one part of the problem that it's the the effective orchestration of those movements. Making sure that the movements are in service of a greater goal, right, of completing, you know, an overall task within a within a facility. So depending on uh you know what those HEVs are servicing whether it's a robot and whether you know there's a hundred tasks that need to be completed so the robot can perform one task right or if it's you know onetoone right that that orchestration can be kind of dynamic and and very complex and that's where we would you know work with you to identify the operation that we're trying to solve and you know work uh if you have a preferred vendor we can work with those preferred vendors for AGVs and ultimately um provide kind of that overhead orchestration layer um that that you know manages all of the the different priorities uh of what tasks need to be completed in what sequence and you know um the the time estimates to complete those. Um, typically, you know, you might call it something like a west or uh uh yeah, a west type system, but you know, for an AR AMR only um solution, you know, that that might be, you know, much less of a scary word than you're thinking like a WMS or a WES. It it's much more toned down and specific to you know that piece of automation or the automation underneath what we are providing.

Uh the next question we had here was asking on the Musion OS does it require CAD part data or can it do all 3D scanning for registration or no does it require CAT data or a full 3D scan prior to operation or can it just handle new parts on the fly?

>> Let me Does a system require CAT? So we can work either way. Um, so you can either have the the SKS pre-registered within the system. Um, so we we typically do this for our visionless applications. So like single skew palletizing where you're going to be running a lot of one skew over and over and over again and it's going to go to a very fixed predefined position at the end of a conveyor line, right? You don't need vision over top of that, an expensive vision system. So that's where we would, you know, rely on preconfiguration of that skew. And that's where I was getting back into the the parts manager. It takes less than a minute to to go through and configure a new skew. You can put length, width, height, any kind of barcodes on there, a weight to it. Um, and you know that now it's configured within the system. Um, or you know, we can apply a 3D vision system that allows us if you don't really know and don't have time to to uh or there's just 10,000 SKs. It doesn't make sense or just a new skew every day. Does it make sense to always be doing that operation? And that's where we might leverage a a 3D vision system to okay, just give me the the good parts of the information about the thing that I need to pick, the length, width, height, maybe a little bit of a color identification. Let me know. Um, I could take a 2D color image of it and and so I can figure out is it a is it a brown cardboard case or is it a shrink wrap tray of cans or is it a pack of dog food, right? Get that kind of high-level information that lets us better interpret how we need to handle it, right? Whether or not we need to, you know, move a little bit slower because it's typically a difficult package type to handle or maybe we can move at full speed or maybe if the system has a side pad, we can actuate the side pad be to to kind of get that better grip on the part itself. Um, so yeah, that that's the situation where we can work either way, having it predefined as a CAD part or, you know, kind of identifying it on the fly. Um, really just depends on your operation and and and the application that we need to solve.

>> Uh the next question we had in here was from Rob uh from the Netherlands there asking you know who to contact specifically for container loading. I know uh with both of our companies uh there's Men Europe that he can reach out to. You guys would have uh some potential solutions there and then we also have some contacts uh locally there from Yuskawa Europe. Uh we have your contact information and we can go ahead and coordinate that uh request.

Moving on, physical AI is a term that gets used a lot right now. It's in the news quite a bit. I know we all three of us panelists had discussed it right now. But the question is what does that actually mean on the production warehouse floor and how is what we're showing here different from what uh a variety of vendors have been promising for years? Uh I'll say at the very highest level before I let Mosian answer that one. Uh one of the things there was the key that we had there in the um case studies that we have proven this we have deployments and we're getting those real world results. Uh we are taking that actual AI bringing it into the physical world and uh hitting production targets daily. Uh but with that I will turn that one right over. So, I'll take that first and then Jake, you can jump in as well. But, you know, I think I alluded to that dur when I was talking about the fact that you really don't have to pre-program your robots or, you know, feed in a bunch of SKs, especially in today's world where, you know, SKs are constantly changing, auto profiles are constantly changing. With the vision system, with the dynamic uh motion planning, the system can automatically sense what do you need to do? And you know I think one of the questions was what happens if things change? It can actually look at that and kind of you know program it in real time and kind of react to those changes. So I would say a lot of lot of the softwares kind of have to be pre-programmed out here. It does it in near real or does it in real time way it can it can really kind of you know um it can sense those changes and react to them. So it really helps the robot kind of um understand what the changes are, react to those changes in real time as you know auto profiles change or even new skews come in. I think Jay talked about a round container versus something that is square you know and then tomorrow you introduce something else and it can just take it and and it kind of looks at that. um you just prompt it a little bit if you need to, but uh from there it can pretty much pick up and and it plans everything, you know, it plans even the motion planning. So I think somebody alluded to the fact that the motion was also very fluid versus being very clunky. So that's part of what is included into the platform.

where it actually works and it doesn't have to be even on, you know, something in a warehouse. You know, we've seen, um, like some of our competitors are just about getting into this, and the platform allows you to use it for any different application that you want to, um, that you that you want to kind of use it for. And again, I'm going to stretch this a little bit, but one of our competitors that has just entered the market, a big, big competitor that has nothing to do with warehouses, is using it for defense systems, you know, so all that is kind of included and you don't have to build that up from scratch.

Jake, do you want to add anything to that?

Uh, yeah, I see a couple questions coming in about AI. So, I think, you know, talk, talk a little bit about that. So, um, I, I understand, you know, AI is generally kind of a a very loosely used term. You, nobody really knows what it means. Is it a black box? Is it not a black box? Can, can I look into the black box? And while our system, the the MCX is technically a black box, right? Um, we, we've worked very hard to kind of give you the ability to see behind the curtain and understand what is actually going on within the black box. Um, and so for that, you know, kind of touching on what what Mario said, there's, um, you know, the the the idea with traditional robotics is, you know, it's a pre-programmed position, pre-programmed set of movements. If that changes in any way, you know, that program doesn't work anymore, or it doesn't work in the way that you thought it would. Um, our system is, you know, built around allowing that change to happen based upon the the requirements of the system, of the operation, right? The new SKUs coming in, right? And, you know, not really beholden to this pre-programmed situation, but being able to look at the information that's in front of us, right, and and figure out what the next best move is, what the next best operation is. In a very similar way that, okay, I send a a 24 pack of Cokes to you, right? Well, you know from your experience of handling 24 packs of Cokes that there's generally two handles on the side. That's the best way that you could probably pick that thing up and move with it, right? We can have that same kind of, you know, very basic terminology and and, you know, pre-understanding that you've given the system and allow it to know that, okay, I don't have to know that this is Cherry Coke or it's, you know, regular Pepsi or it's Coca-Cola or whatever it can be, whatever if it looks like this, then I'm going to make an educated decision on what and how I need to operate on that package, right? And so it's giving, you know, designing the system in a way that it's not fixed, you know, outcome every time. It's a, okay, with the information that I have available to me, what is the best choice? What's the best logical direction for the robot to go in, uh, in a sequence of actions, right? And that's where it kind of gets to, okay, it kind of figures it out on its own, right? It's not, you know, this predetermined thing. So hopefully that explains a little bit. I know it's still kind of gray. Um, but, you know, hopefully it it calms the the nerves a little bit that, you know, it's not just this black box that you can't see behind it. There is real thought and real process going on behind it and it's very logical as far as how we choose to based upon the information we're provided to for the robot to go to the next operation.

Coming real close to the time here. So, I'm going to try and just limit to the, uh, I think the the two best questions remaining in chat here. Sorry, we didn't have a chance to get to all of them, but I did want to honor everybody's time. Uh, so, just really briefly, we had the question on what kind of stability and volume utilization rate can your palletizing robot achieve for pallets? Uh, selfishly, I want to have this one just stick with the, uh, Packmaster-based solution here because, uh, that's the one that we're in the process of our two companies collaborating on. Uh, I know specifically on stability, some of that does matter on the product that you're going to load in there, the variety that you'll have, but when you get into that stability, the design of the system to have the pallet start at the, uh, mezzanine level and wrap as it lowers down significantly increases that stability. But for kind of that volume utilization rate, I will let, uh, Jake go ahead and take that one.

>> Yeah. So, we've worked really hard to to kind of optimize that and we're working on it every single day, getting it better, getting the algorithm a little bit better. Um, you know, I think a good realistic, um, uh, uh, packet density, uh, that that we like to shoot for is somewhere in the mid-70 to 80 range, 80% utilization of the volume, uh, that that is given. Um, but, you know, that that's subject to vary. It might, if there's less SKUs or less number of unique SKU sizes, that might increase. Um, if it's high variability, it might decrease, um, but, but I think 70 to 75%, maybe even 80% is a good, you know, base range to, okay, yeah, it, that, that's viable, right, as a solution.

>> And I think, uh, you know, even though I wanted to drive that towards Packmaster, I think we could see on the Trusco case study there, that was a much higher, uh, density there, probably near 100%, above the 90, 95%.

>> But then the last question there, uh, which I think is a good one, and it's from Ron Burgundy, so you got to stick with that one, uh, but with a lot of facilities already having some amount of automation in place or being a brownfield, what does the path to deploying a system with Mujin OS look like? Do you have to rip out and replace, or can you just add on to and work alongside the the existing infrastructure?

>> Uh, so we've done it either way. Um, you know, we, we've done it where we've come in and retrofitted existing solutions. If there's automation that needs to, you know, be updated, right, and get a little bit more dynamic and get a little bit more capable, right, to squeeze, uh, if it's a 20-year-old robot, squeeze every last bit of performance out of that 20-year-old robot, um, you know, uh, until it's time to upgrade the system as a whole. Um, or, you know, we can look at it as more of like a foundational building block. So that that's where Mujin OS kind of, you know, shines on its own where it, we kind of see it as the building block that allows you to not only just do automation today, but allows you to add automation tomorrow without all the hassles, right? There's not many companies in the world that that say like, okay, you can have a robot this year, and then next year add AGVs to it, and then six months after that add a pallet ASRS, and it's all still one platform. That's generally going to lead to you have to learn how to work with the robot, you got to learn how to work with the AGV, and you got to learn how to work with the pallet ASRS, and all of those are three different softwares, three different support packages. You have to manage three different companies you have to interact with for any kind of support calls or training, and it becomes a nightmare, right? And and so that's where, you know, we kind of look to be that that foundational platform for automation within your facility that you can always kind of add to it. It's not, it's never, um, you know, this is the group of automation that you can do with here. Don't, don't, you know, this is outsiders, right? You know, we want to be very inclusive to all automation that could benefit our, our customers.

>> All right, I think that's it. If you want to go ahead and wrap it up, Mike.

>> Great. Thanks, Mario, Jake, and Chris for the great presentation. I'd also like to thank the whole Mujin team for their help putting this together, and thanks to you. Thank you for joining us. We appreciate that you chose to spend part of your day with Yaskawa.