Impact Your Inbound

Hello everyone, and welcome to today's webinar. Impact Your Inbound-- Automate Unloading with Stretch. My name is Mike Fair. I'm the lead product manager for our warehouse robotics team here at Boston Dynamics. I've been here just over a year now, and before that I've worked in the industrial automation space, as well as in the public security robots sector. I am joined by my colleague, Spenser. Spenser, would you like to say hello and introduce yourself? Hey, I'm Spenser Brouwer. I'm the Director of Field Application Engineering here at Boston Dynamics. I've been in the robotics and IoT industry for about 10 years now, most of that in the supply chain and logistics space. I've been with Boston Dynamics specifically for about a year. And I see a lot of familiar names on the webinar today that I've worked with in the past, so thank you all for being here. Yes, so today we're going to cover a wide variety of topics. We're going to talk about some of the key challenges in the warehouse industry. We'll give you a little bit of the background on Boston Dynamics history, the path we took to get here, and how we're solving some of our customers' challenges in the warehouse. We will talk a little bit about what the key specifications of the robot are, as well as the installation and deployment experience. We'll give a sense of what it's like day to day to run a Stretch in operations. Then we'll talk about our road map and our future with Stretch, providing an overview of the future applications. And then, finally, we'll end on some Q&A. So today, one of the biggest challenges in the warehouse industry is labor. We hear that everywhere we go. There's a real scarcity in having reliable workers and lots of rapid turnover for these types of jobs, and that's not going to go away anytime soon. Our challenge is really to find solutions to mitigate these different types of challenges so that companies can continue to remain productive, profitable, competitive, and also relevant. There has been tremendous bottlenecks and stresses in the supply chain that are causing lots of challenges in order fulfillment. The pandemic definitely helped to cause a surge in the demand for goods and e-commerce. Coupling that with the shortage in labor really has identified some weaknesses in our supply chain. And getting products closer to customers to maintain a high level of service that they expect, like overnight shipping, really helps to contribute to these key challenges. Warehouses generally are pretty early in the adoption curve of automation, and that's where a need for flexibility in solutions for automation because these warehouses weren't really designed around automation. It's causing big pain point for some of our customers. And last, some of these tasks in the warehouse are the most demanding in terms of the types of heavy lift that's required to do some of these applications, and it puts people in unpleasant environments such as high heat or cold temperatures. So safety concerns and ergonomics also continue to be a big challenge. At Boston Dynamics, we develop robotics technology that have extraordinary mobility and manipulation, and that can be applied to solve real world challenges. [CLEARS THROAT] Excuse me. Boston Dynamics is about a 30-year-old company, and we got our start in the research space with DARPA, government-funded projects for cutting-edge technology exploration. The company was acquired by Google in 2013 and then by SoftBank in 2017, and during that time, we began our journey towards commercializing our robots. In 2021, the company was acquired by Hyundai, which will be a great partner for us as we continue to push the boundaries of robotics forward while also building the scale that we need in the manufacturing capability to deliver commercial robots to the market. You might have seen some of our robots and their crazy antics dancing on The Tonight Show with Jimmy Fallon or partying it up in the Sam Adams Super Bowl commercial. And you might be thinking, what purpose do these robots serve? Well, at Boston Dynamics, we're in it for the long term. We aim to help robots become a part of our everyday lives, to improve the way we work and how we live. And our goal, really, and our mission, is to imagine and create exceptional robots that improve or enhance people's lives, that add value to making our daily lives and tasks safer, easier, and more productive. We have three different product areas. Spot is our quadruped robot. It's used for industrial inspection applications and security applications. We won't spend too much time today discussing Spot, but it is available for sale in those applications. Stretch, which is the focus of this webinar, is our flexible warehouse robot, and we'll go into to depth about what this robot is and how it's helping to add value at our customer sites. And of course there's Atlas, our biped humanoid robot, which is our advanced research platform. It's not a commercial product, but it helps us push the boundaries of robotics technology, and we bring those innovations forward to benefit all of our product areas to solve some of today's challenges. For warehouses, our journey began with Atlas. The team posted a video of Atlas doing a very simple maneuver of picking up a box, rotating, and placing that box. It was posted on YouTube, and all of a sudden we had a bunch of inbound inquiry of people wanting to get Atlas to do that task. And as I just said, it's not a commercial product. So that really gave us our aha moment about there's something in the warehouse and logistics automation space that needs to be explored. So the team developed, very rapidly, the Handle robot, which is the second robot over from the left. That was initially incepted in 2016, and that's where our journey to autonomous unloading started, by proof of concept testing Handle at customer sites, getting real world experience in learning, and conducting discovery on what it would take to solve some of these challenges in new ways. That led to the Stretch prototype robot, which was rapidly prototyped, designed, and field tested, again gaining more experience and learnings in the industry, that ultimately resulted in the Stretch robot that we now have available in the market. Stretch has a small footprint. The base is about the size of a pallet, so it can go anywhere in the warehouse that a pallet can go. It has an omnidirectional base so it can move in any direction. There is a independently actuated perception mask with 2D and 3D sensors so it can see its environment. Lidars in the base allow the robot to detect obstacles in the environment and navigate. It has a very long reach with a vertical reach of 10 and 1/2 feet and a horizontal reach of 6.4 feet. And we custom built a smart gripper, which has adaptive sensing and control, to pick up a wide variety of SKUs and payloads of up to 50 pounds, or 23 kilograms. The battery can run two full shifts of operation before it needs a charge. Stretch is purpose-built for addressing a wide variety of applications in the warehouse. There are many applications from the inbound side of the operation to the outbound operation and everything in between. Those are typically broken down into the inbound operations, or unloading. There's de-palletizing or palletizing applications, case-picking or order building, and then truck loading on the outbound side. Our first application with Stretch is unloading. This video is showing our demonstration from the ProMat Show a couple of weeks ago in Chicago. This is Stretch's first application, and it's unloading floor loaded cases from both containers and trailers. Stretch is picking a mixed variety of boxes, both different sizes and weights, including highly graphic branded boxes as well. The built-in machine learning and perception system can detect all of these boxes and autonomously pick them. There's no pre-programming of the boxes required. In this demo, the robot is unloading between 500 and 600 cases per hour, and we have some optimizations and exciting new features on our roadmap that will improve the case rate up to about 800 cases per hour for a single case pick. And it's only getting better. So with a piece of automation like Stretch, this can definitely help improve worker safety by tackling the most tedious and the most demanding tasks in the warehouse. It can run continuously. Most of our customers are running 24/7 or multiple shifts of operation, gaining additional productivity in their workflow. It is a mobile robot, so it can go anywhere that the work needs to be done. And it's flexible. It can drop into existing facilities or existing infrastructure and get to work within just a matter of days. And that's where I'd like to bring Spenser into the conversation to talk about what it's like to deploy a Stretch robot. So Spenser, you've been to a lot of warehouses recently, and your team is responsible for deploying Stretch at our customer sites. Can you talk about the different types of conveyance in warehouses and what their differences are? Yeah, so there's two really common solutions we see in warehouses. And I'll go into detail on how each one works here in a couple of slides. But you either have a telescoping conveyor, so a fixed rigid or a transversing telescoping conveyor that fixes itself to upstream automation. And about 60% of our customers are using those sort of telescoping unloaders. For the remaining about 40%, we see them using an actual flexible conveyor. They might be doing less automated operations or just being introduced to automation. And so for those customers, we're seeing them use these flexible powered conveyors that you see on the left side. Nice. Yeah, so can you talk a little bit more about the accordion flexible type conveyors? Yeah, I'll talk about that. But first I'll kind of say, on the last slide-- so installation itself? Super easy. So with either of those two options, we're going live on production freight in as little as one day, and most of our sites are averaging about two or three days right now. So for the QDU specifically, which is the fencing solution here, we call it the quick deployment unit. It's a traditional fencing option. So the side fences collapse into the wall to get out of your way when you need to do manual operations. And then the actual safety interlock system there in the center transverses between doors. So one or two people can push it between the doors. It makes your change over times really fast. And all the safety interlock connections are actually built into that solution. So you can go from one set of fencing wings to the next set of fencing wings on the next door and drop some alignment pins in, and then it'll hook up to that and be ready to go. And there's also options when you run the standard fencing solution like this to actually tether the robot to power so you can basically get continuous operations without the need for opportunity charging. For the telescoping solution-- so we call this the TVG, or the virtual guarding solution. So it integrates directly with that fixed telescoping conveyor. It's lidar-based for your safety interlock, so a small piece of fencing to cover blind spots between dock doors and then we actually use a lidar to identify areas where Stretch needs to go slower or stop altogether. These are static per dock unless you have a transverse conveyor, in which it gets fixed to the conveyor. And then Stretch is actually seamlessly controlling that conveyor. So just as an associate would, if we need the conveyor to come out or go in, Stretch is actually wirelessly communicating with the safety system on that conveyor to make that happen. So I'll talk a little bit about operating interfaces. So sort of two operating interfaces-- and we'll spend a little bit of time talking about how easy is it to use Stretch. I think that's a common question we get, Mike. So we can do basic operations using our pendant system, and most associates are trained within minutes. It's a gamification, if you will, just to get Stretch into the area it needs to perform its job. Everything we do is built for low level operators. We do not need a master's degree associate to run our robots, and that's a major focus for us. So when we go on to the actual control interface, you end up getting something that looks like this. So this is the control console. It's a tablet-based system to start, stop, drive, see job metrics, get notifications when Stretch needs help, perform health checks. And it also has a more traditional safety interlock built in there. So you mix a little bit of the new, more advanced technology with the tried and true industrial automation. One big thing here is sort of the same low level operator concept. So you have that UI and the interlock, and we can get a trainer program up and running where you actually have a qualified trainer in your facility within about a week. And then operators from their point can be trained within hours to cover about 80% of Streth's operation and actually be up and running as you continue to teach them the edge cases there. Awesome. Thanks for that overview. So just to sort of wrap it up on the deployment installation and user experience. From the time a Stretch arrives at our customer site, how quickly are people up and running? Yeah. So we're seeing about two to three days right now. We've had some sites go live in production in as little as one day. It really depends on the prep work for that facility, but once we get all the safety system up and running, we sort of go through a go-live milestone with our customers. And then we start training. Our goal is every acceptance test we run at a customer site is run by the customer operations team. So we get straight into training and having the customer use the robot. That's outstanding. Awesome. Thanks for that. So inevitably problems get encountered in these containers. Can you talk a little bit about what types of issues are encountered during the unloading process and how Stretch can deal with those? Yeah, and this video actually opens it up for sort of a pre-submitted question of how Stretch handles miss-picks and recoveries. I think whether it's a robot or an associate, box avalanches are really common amongst floor loaded trailers and containers. So if we go back about 12 months ago, we completely aligned our strategy with a zero intervention strategy. And this was due to feedback from building general managers and inbound operators saying that, hey, if I have to have a full-time person managing and wrangling this robot, dealing with interventions, not only is it going to hurt the financials of the ROI for the robot, but it's also going to hinder them from doing their actual job, which is managing five, six, even 10 inbound dock doors at a time. So with that backstory aside, we sort of took box avalanches and drop cases happen with Stretch just as often as they do with an associate. And as you can see in this video, Stretch absolutely does clean up its own mess, and that's something that the engineers have worked really hard on. This means most of our production sites are running an intervention rate of 0.1% to 0.4% on average. And we intend to maintain that high level of automated recovery exception handling as we increase our inbound rates because our customers say it's really important to them. That's amazing. Awesome. So today we've started installing and deploying Stretch since the beginning of the year. Those are our first off the line production units that are now working at our customer sites. We've signed on a number of customers as a part of our pilot program that are ramping up deployments this year. Spenser is going to be very busy this coming year. And just a quick snapshot of who those customers are-- we've got DHL, Maersk, Gap, Puma, H&M, PHL, and NFI. So a pretty good mix of both sales and retail type of customers. And going forward-- it was mentioned earlier that Stretch is a multipurpose robot that is designed to handle a wide variety of applications in the warehouse. This video is showing the vision of that future. It's showing all of the applications that Stretch is capable of, starting with the application that we're solving today, which is the container unloading. And we'll bring you through all of the applications in the warehouse as we zoom through this warehouse of the future. We anticipate that we'll be able to do palletizing operations, both to a fixed cart at Stretch's base, as well as interoperability with other EMRs of companies that we partner with to work alongside Stretch. We foresee interacting with autonomous forklifts and other autonomous systems in the aisles of a warehouse. Stretch can perform case picking in order to build applications, building pallets alongside an AMR. And this is a demonstration of Stretch picking through its own cart and then towing that along to be dropped off into a staging area where it can go into a wrapping station, then go to the outbound area. But for floor loaded outbound, we also anticipate that Stretch will be able to do container loading. And that would all be done with the same hardware platform, but through software updates enabling those future capabilities. All of these are on our horizon and on our roadmap. OK. Now we have an opportunity. Let's take some Q&A. If you haven't already, please submit some of your questions. We also had a number of questions that were submitted before the webinar that we will address. So we'll just pick through some of them here. What is the ROI? How many robots are needed per site? Spenser, you want to take that one? Yeah, I can do that. So at this time, especially for our early adopters, we think a minimum of two robots at the facility with the opportunity to expand is the sweet spot for most customers. And when we do that, most of those customers running two robots across two shifts a day are regularly seeing about a two year payback in their own ROI calculations. Great. We have a question here about how does Stretch take care of a container with different SKUs or cases, sizes and weights? Do you want to take that one too? Yeah, I'll do that one too. So today Stretch is handling production freight at customer sites where the inbound container or trailer regularly has eight to 24 SKUs on the PO, so this means that Stretch could come across eight to 24 times per container a different box size, different box weight. So currently Stretch has a novel way of using its weight and perception sensors to identify missing dimensions to sort of make a little box catalog for each container or trailer. And then it will actually also use the weight sensors to determine the box weight. So it kind of fuses that together, and it's making inferences on how it should place the box. So this is also a good time to address two very different markets that we look at when we look at trailer and container unloading. So there's the inbound supplier, transload, cross-dock, et cetera market. And then there's a second really large market for mixed parcels, so think Postal Service, RMA processing centers, large shipment providers. So we think these are really large markets, and we are focusing on that inbound palletizing, de-palletizing, the stuff you would see in the inbound supplier transload cross-dock market. So when we talk about the question on different sizing and how does it handle that, does it do poly bags-- we kind of segregate those markets. And we're targeting that inbound market. Great. Yeah, good question. There's a question here about how do nearby personnel in the direct area-- how are they safeguarded? So Spenser did provide an overview of the lightweight safety infrastructure that is provided outside of the container to separate humans from the robot productively working inside the container. We do have lidars on the base of the robot. And in this case, we are expecting boxes to fall in that avalanche scenario where the boxes will fall towards the base of the robot, and we'll use the lidars on the base to detect where those boxes are to clean up its mess. But we can't tell or distinguish yet the difference between a human and a box falling in its path. So providing those additional limited safeguarded outside of the container is really important to make sure that it can be productive and safe for people to work around the robot. OK, let's see here. There's a live question that we can answer, Mike. So there's a question on the live chat for what's the planned roll out for Europe and Germany? So I hope we have exciting announcements to make soon. But right now we're currently only available in North America. And I think for more details on the rollout plan, you can reach out to our sales team, and they'll be happy to share some of that on a case by case basis. Great. Let's see. We have time for a few more questions. Let's see here. What's the speed of the trailer unloading compared to using labor? Spenser, you want to take that one? Yeah. I think the labor numbers are all over the place. We haven't asked a single customer and gotten the same answer between all of them. I think the best thing we can provide is that the rates of Stretch are consistent and predictable. So like we said today, depending on freight profile, we're seeing between sort of 400 and 600 as the average across all of our production deployments, really depending on that inbound freight. And then we see associates go anywhere from 180 an hour all the way up to 1,000 an hour. We just see these huge swaths of numbers, and they vary throughout the day. They get lower as the day goes on. And that's the best data point we have to share. Great. We've got some questions on variability and uniformity of box sizes, and I'll take that one. So Stretch, as mentioned earlier, can handle payloads of up to 50 pounds, or 23 kilograms. And our smart gripper, with individual sensing and individual vacuum cup control, allows the robot to pick up different sized boxes, as well as slightly crushed boxes as well. So the gripper is adaptive. Of course, it depends on how crushed or damaged that case is, but it is able to handle a pretty wide variety of conditions of boxes, including some that are slightly dusty as well. Let's see here. What else do we have? And there's one here about-- so it says typical fashion containers from overseas are filled with cartons that are stuck to the container, which I kind of assume to mean really tightly packed containers. And it says, how do you handle this? So assuming it means really tightly packed containers, this is actually one of my favorite topics. So what we see in practicality at production facilities is really similar to what you would see with an actual associate in the trailer. So getting those first few boxes of a new facade out is hard, and that's also kind of where we see most of the box avalanches, if you will, on really tightly packed containers. So with Stretch's gripper, and the force it's able to put on a single box to pull it out directly vertically, it actually sometimes has an advantage over associates because we don't need that dexterous manipulation to sort of get the fingers in between the ribs on the container and try and get that box out. Sometimes they'll pull one out from a middle just because it's easier to reach. And so those are areas that Stretch actually excels at, and sometimes we see it performing better than associates with, especially with big, heavy, tightly packed boxes. All right. We have time for one more question. Does Stretch have the ability to load containers? If so, does it use AI to optimize loading within the container? We briefly touched on that with our vision of the warehouse in future applications. So not yet. We do have it on our roadmap. And the important thing is that the form factor and the design of the robot is designed to handle all of these different applications in the warehouse. The robot does use machine learning and perception with the 2D and 3D vision sensors on the perception maps, as well as behaviors to operate autonomously inside of the container. So no need for connections to outside systems. It's all just built in to the robot. So we can't load containers today, but we will in the future. And it will all be done through a software update. Oh, we have a few more questions coming in, so sorry. We can keep going here and make sure we get some more of these questions answered. Yeah, I see a live question on the line about UPH and staffing levels on the dock. I think we've talked about UPH. Most of our production sites are seeing between 400 and 600 today. Staffing levels is a fun one because we are currently pushing the boundaries here. We think the right number is that one associate would be operating up to five or six inbound dock doors, meaning five or six Stretches at a time. And that's kind of the area we're looking to push the boundaries over the next year with our customers to see what we can get there. Good. There's a question here about offering financial options, such as robots as a service or leasing options for a rapid deployment. Not yet. Since this is a brand new product that we've launched this year and we're getting through our initial deployments, our backlog and our demand is pretty strong through this year for customers that are acquiring fleets of robots. So we don't yet have a RaaS model, but we certainly will consider that in the future. See a couple of questions on what's the maintenance model and what are your current critical issues, specifically related to maintenance? I know our sales team would be happy to discuss the support plans and how it works, but I'll answer what are the current critical issues. So we believe that Stretch is potentially one of the most complex mobile manipulation robots ever fielded at scale. So Stretch is early in its product lifecycle. We're maturing with real world deployments. We're going live with new customers seemingly every month. That could just be my mind. And I think our main thing right now is we're focused on building out a robust reliability program so that Stretch is on par with most of the traditional automation equipment our customers are used to. We have a couple of questions about refrigeration, cold storage, different environments. So today we're really focused on a typical inbound operation of dry goods usually, not cold storage yet. It's certainly something that we're hearing from customers, and there's a lot of that out there in industry that we'll consider in the future. OK. Mike, I see some questions on charge time. Yep. Yeah. So to go back to the key specifications, I mentioned earlier that the robot has a lithium ion battery capable of being able to run continuous operation for up to 16 hours, or two shifts of operation. With a fast charger, you can recharge the robot in just over three hours on the fast charger. And also, as Spenser mentioned earlier, when it's used in the accordion conveyor type of configuration, where the flexible conveyor is attached to the base of the robot, you can connect to shore power and run continuously off of that shore power. All right. How about time for one more question. There's a question on moving towards the parcel and loose load segment. Spenser, do you want to recap what you mentioned a little bit earlier, just to reinforce our focus here? Yeah, I think the loose load market, and I said it a little bit earlier-- we believe there's two market segments here. We believe there's your inbound cross-dock transload operations that fuel our global supply chain system. And then a little bit further downstream there you have the all right, now I'm into my individual [INAUDIBLE] or-- e-commerce is the wrong word-- but at the parcel level. So I'm sending it through postal services or private parcel services or even RMA return processing. We think these are two separate large industries to tackle, and we're currently focused on the inbound general case handling market, or the conveyables market, if you will. And we think the future for the loose load parcel poly bag market is there, and you might see a little bit of a different solution for that in the future. Awesome. Good overview. Thanks very much, Spenser, for joining me today. We will wrap it up. Of course, we have a few more questions rolling in, but we're running out of time. We will reach back out to you to answer some of these questions in the future. And of course, if you're interested in learning more, please visit our website, contact our sales team, and you can visit the links that are shown on the screen. Thank you all so much for joining Spenser and I today. We really appreciate your time, and thanks a lot.