Boston Dynamics rocked the world with the robots like Big Dog, Cheetah and Atlas. As one of the company’s latest creations SpotMini comes to market, CEO Marc Raibert will discuss the company’s journey to productizing.
What are the biggest pain points for Boston Dynamics as far as actually Bringing these products to market so First of all let me set the record Straight we weren’t trying to monetize The robots all those years in fact we’ve Always been a money-making organization Even in the early years although we Weren’t doing it by building robot Products but we have been building Robots for about the last dozen years And it is true that we did you know Contract work at first and then we were You know I’ll call it self funded by Both Google and now Softbank and now we Are adding a branch to the things we do We’ve been an R&D company for a long Time we love being an R&D company and Now we’re adding being a product company In a couple of areas spot I would say Spot will be our first product but as We’re going to discuss we did in that Position recently who already has a Robot product out there but spot will be The first product developed at Boston Dynamics starting this summer and you Know maybe a year or two a year and a Half later we’ll have handle as a Product as well so you asked what the Pain points are I you know obviously the Skill set required to making a product Are have additional things in addition To the engineering that you do to Develop but new the first one of
Anything so for instance we’ve been Worrying about passing emi certification Tests safety for the robot supply chain And getting manufacturing all in hand And all of that stuff has been a Learning curve I wouldn’t say it’s been A pain point but you know we’ve Certainly learned a lot added some new Styles of people to our to our team in Order to do that okay so you looted to Handle before so actually let’s let’s Start with that video let’s play that Handle video and I want to say right off The bat I think this is the one thing That we’re showing today that probably Everybody in this audience is already Sedin but this will give you some some Context let’s go back to that video he Just has there we go okay so so this is Handle this is a robot that we’ve seen In the past but now it’s actually doing A very specific job is this was this a Robot built to be a logistics robot It’s the closest we have to a Purpose-built robot so when we started We’ve been interested in making a Wheeled legged combination for a long Time and never had quite the right Sponsor for it But about two years ago we just decided We were going to do it and we built the The handle you’ve seen before which was Partly atlas parts and partly new which Was the legs with the wheels and once we
Got it going we started being interested In the logistics opportunity and we Redesigned the whole thing with this Particular task in mind that is moving Boxes stacking them in pallets and D Stacking them from pallets and there’s Just a huge amount of this kind of work Going on around the world we estimate That there’s about a trillion cubic foot Boxes moved around the world every year And most of its not automated and so There’s really a huge opportunity there And of course this robot is great for us Because it includes the DNA of a Balancing robot and moving dynamically And having counter weights that let it Reach a long way so it’s not different In some respects from the robots we’ve Been building from years on the other Hand some of its very focused on Grasping being able to see boxes and do Tasks like stack them neatly together so It’s it very much is a Boston Dynamics Robot from the stand point if it is it Is quite complex in a lot of ways I was Out in at Amazon Fulfillment Center in Staten Island a couple of months ago we Saw the Kiva now Amazon robotics robots And those are at their base they’re very Simple and most of these factory Automation fulfillment robots we’re Seeing are very simple this is this is a Lot more complex so how do you actually Sell a company on such a complex
Sophisticated robot so this robots about Half as complicated as our most complex Robots but you’re absolutely right it’s Much more complicated than most of the Things you find in warehouses on the Other hand there’s nothing that can do The functionality that this robot does You know in in terms of reaching across The D palette or reaching to the top of A tall pallet there’s no machine out There that can do it and so we had this At Pro mat we didn’t have the robot we Had video of this robotic from Pro mat Which is a Logistics conference the week Before last I think and there was Intense interest in getting that Functionality available We know that combating the complexity is Going to be one of the things you know Making it reliable enough is one of the Things we’re gonna have to work on I Don’t think the cost is going to be a Problem because you know the cost isn’t Going to be that out of range and the Functionality of the robot is really Offers a lot and bas of Amex made really Your first major acquisition recently in Order to basically complete this robot Can you talk about that sure the robot The vision system used in the video you Just saw is used to both locate the Pallets using markers but then when it Gets to the pallet it looks for boxes And looks for the shape of the box and
Finds the one that’s supposed to pick up And picks it up and it works pretty well But we decided that we really needed to Up our game and so we acquired Kinema Systems a company located in Menlo Park Who’s got just an extraordinary 3d deep Learning based vision system that we are Actively working on porting to handle We’re also going to sell the product They have which is a fixed manipulator That’s doing deep palletizing and if you Go on the web you can see their their Videos but we’re also going to take that Vision system and adapt it for use with Handle we’re really excited by that We’re also going to adapt their vision System wherever it fits with all of our Robots because it’s really remarkable They got an award I think last year from Nvidia for one of the best deep learning Based applications out there I think They were one of two top winners I was Gonna ask you about that so obviously a Big part of the reason why you made the Acquisition was specifically for this Logistic so you know spotting boxes on The shelves things like that Again the robot that we were talking About last year and that we’ll be Talking about in a bit as a spot mini Which is the first robot you and now Will be commercialized does the Kinema Technology make sense for a spot many I I don’t know exactly about spot many you
Know spot mini is designed as a platform Where it’s got a vision system built Into it that’s doing a good job but then There’s an opportunity to add stuff Which we expect in third parties to add And when we get the robot out here I’ll Show you a little bit more about that I Don’t think we need a kinema system Right now but I’ll tell you more broadly At our company everybody’s fighting over Who’s going to get first shot at using Kinema Zoology for their application and We’re you know we’re trying to keep it Focused on the immediate need which is Handled but everybody’s lining up to Both take care take advantage of their Knowledge and the capability of the System okay so let’s get to our next Video it’s just something you showed me Backstage which is absolutely amazing It’s a behind-the-scenes look that I’ve Never seen before and I think most of The folks haven’t seen it before so These are the components of spot mini That we’re seeing being tested in the Laboratory so obviously we care about Making spot mini reliable and have long Life when it’s out there so we’ve been Doing a variety of testing in our lab Where this is the component testing Where we go is obviously the legs we Have the arm which I think you’ll see Next we have the shoulder and hip motors And all those actuators here’s some
Other internal drives all being life Tested in our place but we’re also Testing the whole assembled robots so we Have this thing that sort of looks like A Corral and there’s automation software Which is the software it uses to Navigate and it’s just on an endless Testing run on the right there there’s An obstacle so the robot sees the Obstacle and avoids it here we have two Different floor types so the robots Dealing with that I don’t know what’s Happening in this Lane maybe just life Testing and there’s a stairway in number Four and basically we run about 200 Kilometres of travel a week currently we Get bug reports and the software people And the hardware people are Working on resolving those bugs this is Another lane of testing where we have Both inclines and you can see there’s a Little rocky area there all these Different things require different Skills on the part of the robot so can You actually talk about some of the some Of the applications that the company has Been looking at again this is a very Different model from the standpoint of Handle being a relatively purpose-built Robot as you said and this seems to be a Case wherein now that you’re starting to Bring it to market You’re figuring out some of these Applications so let me preface that by
Saying we think of our three robots There’s Atlas which is an R&D robot and We have no plans to commercialize at any Time soon then there’s spot which we Think of as a general-purpose robot Which means we don’t have in mind that The specific application that’s going to Just knock things out of the park but It’s got a lot of versatility and we’ve Designed it to have versatility so People can customize it and then there’s Handle which is a purpose-built robot And it’s going to do really just one or Two things and we’re optimizing it for That but going back to spot spot is a Platform where you can add mechanical Components and there’s an API so you can Add your own software and in fact we’ve Been developing components in order to Get the project out of the gate where We’ve added a manipulator which is an Add-on we’ve added a 3d a 360-degree set Of cameras that also has a very low Light camera which you’ll see why do you Go to the next video and I can yes I Should go to the next video right now so We can actually see it there we go see What what is it doing what’s the purpose Right now this is being tested at the Massachusetts State Police and this is One of their testing places where they Worry about hostage situations or bomb Scares and the idea is to avoid having To have a person go through that door to
See what’s going on you know there’s Somebody inside the door opens up and Boom there’s a robot opening that door Right and the idea is that the robot can Open up the door there is someone Teleoperated it from from a safe Location we have been experimenting with Tele operating over longer and longer Distances And we’ve done it from California to Boston I recently did it from an Airplane flying over Chicago to wherever The data first goes to Boston the Wi-Fi On the place on the Wi-Fi in the plane Do you all know how terrible a Communication that is but the key Technology is having enough autonomy on The robot so that some simple high-level Input from the user can can be an Effective solution and that’s what we Tested there what do you go to the next Yell let’s go let’s play the final video And I want to ask you real quick again Last year will we’re on stage you Mentioned a time frame I think of this Summer we’re aiming for July or August These will start so we are we have these Things coming off an assembly line now But they’re all betas and we’re using Them for testing like you just saw and We’re also doing a lot of redesign and Then the production will really start in July here’s at a construction site in Tokyo where the robot has previously
Been taken through the space and we’ve Mapped it and here the robot is Autonomously going on a route collecting Data you see it’s got that tall thing on Its butt that’s a set of 360 cameras That takes very high-resolution imagery We have software that lets you treat the Imagery pretty much like Street View Except its Street View for these cases And this lets the construction company Get an assessment of progress at their Site you might think that that’s a a Low-end task but these companies have Thousands of sites and they have to Patrol them at least a couple of times a Week to know where they are in progress And they’re anticipated using spot for That so we have over a dozen Construction companies lined up to do Tests at various stages of testing and Proof of concept in their scenarios this Is not a skyscraper this is more of a Hotel where the spaces are tighter and So the navigation has additional demands On it as compared to the other one so We’re a couple months out from from July At this point are you able to talk about Pricing it not quite yet this summer Okay so we do have a very special guest Can we actually bring Stage this is Lily Cohen and beta what Number is it the 24 I can’t quite see 26 So if you could actually start by Walking through this is a close to
Production unit if you can start by Walking through what’s new about this Spot many you know we’ve we’ve Redesigned many of the components to Make it more reliable to make the skins Work better and to protect it if it does Fall this robot has I’m going to go out There and point it a few things this has Cameras it has two sets on the front and One on each side and one on the back so It can see in all directions I gotta say – it looks a little more banged up I Think that I’ve seen a lot of your Robots it looks like it’s kind of been Through the wringer I mean we’ve been testing these things Relentlessly as I say they usually go Through many hours of of testing a week They do fall still this is also designed Use these are like roof racks so that When you build your special platform Like maybe the previous speakers will Make a a deck that you can take a drone Off from then they’ll just attach that There’s connections for for an Ethernet So that they can communicate and then we Have our IP API this also has a special Radio on it it has built-in Wi-Fi but For an event like this where you all Have your phones out and you’re texting And stuff Wi-Fi can be unreliable so We’re using a spread spectrum radio and Of course it’s got the arm why don’t we Show off the arm one of the great things
About a mobile robot like this with an Arm is that the workspace is essentially Infinite and the motion of the base Contributes to the motion of the arm if You have a fixed arm they’re usually Very limited to what they can do and of Course we’ve coordinated the arm with The body so Lily is actually just Driving the hand I know you think it’s a Head but it’s really a hand and she can Drive it for and aft and sideways and Now she’s put it into a mode where the Hand is stabilized in space we love to Call the I love to call this chickenhead Mode And really people have the capability of Stabilizing their manipulators and their Heads while the rest of it moves and it Really facilitates the ability to handle Objects and work in the world as the Robot travels around in the world can You talk a little bit about the control Mechanism so this has onboard computers That interpreting the data from the Vision they also interpret data from the The load cells in the in the link so it Can feel the environment it’s balancing Itself I won’t kick it we’re not allowed To do that anymore but I’ll show that It’s stable and if I push on it if I Push hard enough it’ll step and you know It’s pretty robust if it does fall over It can get back up so there’s the low Level stabilization the robot and then
It has vision that it can use to Negotiate obstacles like this block and Here Lily I want you to run me down with The robot they won’t do it huh I wouldn’t promise that it’s never gonna Run into anybody but it’s using its Vision to treat me as an obstacle and Then even though Lily’s just saying go Straight ahead it goes around and later On where I think after the show closes We’re gonna demo this out front you can Drive it and see you know you can see if You can run one of us down with it if You want and then on top of that we have Navigation software that can Autonomously navigate around space and Eventually you guys will all be writing Apps that run in and interact with the Controls on the robot so what’s the What’s the purpose that of having it Mainly controlled because it seems like The vast majority of the time it’s going To be moving around autonomously I Thought I don’t know there’s many people Who have applications so there’s the Police application where they really Want to drive it although again the Autonomy is always running at a low Level we have people who want to do gas And oil applications where you’re at it In the oil rig or something like that And they’re happy to have a person in The loop they just don’t want to have The person out on the oil rig and we
Other energy applications where it Doesn’t really need to be completely Autonomous in the fact that there’s no One there and then we have lots of Applications Autonomy is a thing and I think the Kinds of paradigms used I hope will Unfold as we get the world building on Top of this we want it to be like the Android of robots where people are Developing their own apps at the Android Of androids the Android of androids yeah So unfortunately we’re running out of Time right now but I guess it really a Good question too close on is what were The next few months look like for a spot Many you know what’s the road to Production so we’ve we’ve we’re in Feature freeze and the software that Happened I think two weeks ago so this Is this is pretty close to it I don’t Know if this is the if the frozen soft This is probably a previously frozen Demo release but it’s very much like This and we’re just you know running Down bugs making sure that we comply With what the targets were so you know Things are getting more and more Narrowed down in our process you know Our company is is most you know our History’s our R&D company has things Much more freewheeling so it’s a new Process to have a feature freeze and Then have this thing gradually go we’re
Working on certification we just passed EMI certification safety you know safety Is a moving target because the simple Things you can do with safety are is Just turn something off but if you have A robot on a stairway and something’s Going wrong you don’t necessarily just Want to turn it off and have it tumble Down which can cause other problems so We’re working with the the standards People to try and come up with rules That make sense in the context of robust Like this great unfortunately as I said We had a lot to talk about and we are Actually all at a time but as Mark Mentioned earlier the robot you’re gonna Be sticking around for a while and after The panelists are over you can actually Check out spot me in person up in the Front mark thank you so much for joining Us and yeah what we can leave together Okay coming up next we have another Fireside chat with anthony levin dosti Of pronto AI and your moderator Kearson course I’m mark thank you so Much [Applause]
