How Google's Wing Drone Delivery Aircraft Works
hey everyone it's norm from tested here and we are on location we're going to a location for a shoot today and i think it hopefully it'll be a cool place one of the technologies we've been covering at tested for the past decade is the emergence of drone technology like hobby rc drones you know from companies like dji and alongside that fast-paced development has also been the question of when we might see drones in commercial spaces for things like drone delivery one of the companies working on drone delivery is google with their spin-off wing we saw a few of their prototypes at maker faires in years past uh but i didn't realize they actually have been making drone deliveries today they're actually in the field in places and cities in australia and east coast in the us shipping people coffee lunch tylenol you name it what we're going to today is one of their test facilities south of san francisco where they run their aircraft that they've developed to make sure it's safe and to test the platform that they've developed and we're going to hopefully check out that aircraft learn about the prototypes that got them there and check out what it means to have delivery not in a you know two day window same day delivery but in a same hour delivery time frame so hopefully a little bit of that future that we've been promised so now here with adam woodworth adam we met before testing viewers actually may recognize some of your projects uh they've gone viral some of your rc personal projects but your day job you are the cto of wing yeah yeah i i've been involved in aviation my whole life and uh you know whether it's on the hobby side or on the professional side um i just really like flying things um and designing things yeah and here you're designing you and your team have worked on many iterations of design uh that have now are now in the field that's a thing that like blew my mind you guys have not only at this facility done the tests but people are getting a real delivery so i want to kind of hear about from the technical side in the aircraft design side how you got to this point where did you start one of the the main drivers here is uh you know at the forefront sort of the safety of the operation so you know when you're flying things you really want to make sure that those things are reliable they're safe they have a low impact on the community or operating over and so most of the design elements on this system are sort of built to meet that goal if you look at this plane sort of as a as a aircraft designer this isn't sort of what i would normally think of as as like what a plane should look like but it really is is the outcome of all of those design constraints um you know at the most fundamental level uh we're basically like if you took a you know normal multi-rotor you know quadcopter and sort of smooshed it together with a regular airplane so we can take off uh vertically hover like a helicopter and then these rotors up front we'll turn on and we accelerate up to sort of flying speed the hover motors turn off and we can cruise efficiently on the wing there and that vitol design gets you the benefits of that vertical takeoff so you can deploy not necessarily with a big runway um and take advantage of all the modernization of the the electronic speed controllers the things in the drone world rc hobby world but also the efficiencies of when you're flying fixed wings yeah so you know the wing lets you go further farther faster and you know the hovering system lets you get into you know it lets you sort of drop off a package in somebody's backyard where you wouldn't you know you wouldn't be able to sort of come in and circle and land um the other the other element of this is that you're you're basically carrying around a spare airplane with you all the time so if you think about you know i'm flying around as a plane and one of these servos brakes or one of these motor brakes or you know a part of the tail brakes you have a whole second airplane with you so you can slow down hover over like a helicopter right um and and you sort of have you have implicit redundancy in the configuration are both systems operating ever at the same time like when it's in a glide mode are you using you know the rotors that are for the vtol so under under sort of uh under normal operating conditions they sort of operate independently so uh you know once we get up to cruise speed the whole hover system will turn off and then you know all of your roll pitch and yaw commands will come from these aerodynamic surfaces uh there's a there's an interesting period between when you're hovering like a helicopter and you're flying like a plane where all these things have to be working together um but the one of the benefits of having this sort of configuration is that you have uh you have you have a very over actuated robot so like if you think about like you have these six degrees of freedom you can move around in and normally you have sort of actuation degrees of freedom you have you know 12 hover motors that can all generate forces you have four cruise motors they can all generate forces you have four control surfaces that can all generate forces so there's a there's a sort of a very interesting and complex allocation problem here of let's say this control surface isn't working right you have a bunch of other actuators on the system that could sort of lend a hand to make that happen and that's that's sort of you know one of the underlying philosophies of our whole design is this sort of like graceful degradation of performance if you have any sort of fault so let's say this breaks or you have a motor that breaks or you know you have a sensing problem you want the system to sort of gracefully adapt to that degraded condition and sort of you know just keep like be you know just keep operating and that's just a natural benefit of the design of going for a vtol design which sounds like something your team knew you wanted to do at the beginning but like you said this is configuration isn't what you would normally think of when you're as an aircraft designer but going back to like your earlier prototypes that looks like a traditional fixed swing so yeah is this where you started yeah so this is this is um one of the places we started back in around 2014 and this was called a tail sitter design so basically the the hover system is the same as the the forward flight system um so you know you take off in this orientation and then to go in the cruise like an airplane you tilt over 90 degrees and accelerate out well you lose some of that redundancy but the one that was so some of the issues here are you know you you don't have that sort of inherent redundancy in the configuration and you also have to you have to size this propulsion system for both cruise and hover efficiency so like those are from a from airplane design perspective those are very different sort of performance regions um the requirements for a hovering system are usually much different than the both the power requirements and sort of the efficiency trades that you have when you're cruising and so for a configuration like that on the vital side where you have a sort of separate lift and thrust setup you can optimize each one for their their relative environments and then that transition part that's where the yeah and magic happens yeah and one of the one of the sort of other benefits of going to uh the configuration we have now is that the transition event is one of the sort of more dynamic events in vital flight so you know for something like this you know you have to pitch over 90 degrees all of your you know the sensor sort of orientations change the orientation the package changes um it's a relatively dynamic maneuver uh for for a configuration like we're flying today you more or less just sort of accelerate so if you think about you have sort of like an infinitely long runway now um that's up in the sky and the plane can just sort of you know it can take its time to get up to speed where we're going we don't need roads yeah there's no you know you don't have to change the orientation of the gps antenna you don't have to worry about you know is the is the vision system that's helping with navigation does that need to change orientation and you know for a lot of the things that we're carrying a lot of is prepared food um you don't want like right you know you don't want your coffee cup to have to do a 90 degree coordinate transform or build in something like a gimbal storage system and to go from there to this configuration yeah is it simulations in cad yeah a lot of it you know uh i always joke that the the way that you design airplanes is spending a lot of time you know drawing all the airplanes you're never going to build yeah um so you really do sort of you know go to the whiteboard and everybody sort of puts up their post-it notes of like well what if it looked like this what if it you know had 50 rotors you know 12 what if it you know what if it changed shape what if you had multiple airplanes like there's there's all these you sort of fully exploit the design space and then you start applying some of the operational constraints to it so you know an example of that would be like okay well we want the we want the package to maintain a fixed orientation so you can take a bunch of the concepts you've drawn and then you can sort of push them off the side uh you want we want redundancy and hover so you can take a lot of the the components that you know can't absorb a propulsion system you can push them off the side and eventually just sort of on first principles you can narrow down the design space from you know something that looks like there's thousands of options into just a handful that you can go and do sort of the deeper math trades on like for this system in particular um you know a lot of the design work was around you know having the margins to be able to absorb system failures so if you if you think about if you think about a you know a regular camera drone like a regular quadcopter if you lose one motor um it's very hard to maintain sort of attitude and enroll pitching control um for something like this if you if you lose one motor you have a bunch of other motors to make up for it but there are some combinations where um you know even if you're even if you have six motors you you might be able to lose one and still maintain steady low flight but you can't complete the mission um we found in a lot of the design trades the the sort of reliability case closed around eight motors so so eight hover motors would be sufficient uh to you know absorb a propulsion system failure or sort of 80 percent more here that's a good sweet spot and the the interesting trade uh that we discovered was uh you know you're you're designing sort of primarily to the the safety and redundancy uh target but then you know since you're offering a commercial service a cost is a large element um and sort of general complexity of the build becomes a large element of it and by going from eight up to twelve motors uh you can shrink the size of the motors so you can use uh you know less expensive materials for the the lift rotors like these don't have to be carbon fiber they need to be plastic um it also moves you into a size of motor that's uh much more popular in the hobby space so you have the sort of volumes of production scale there that affect the cost there and uh you know in our case uh the smaller routers we don't have to we don't have to position a line in any particular orientation so if you had if you had a big rotor on here and you wanted to stop it to hover it's acting like a little wing yeah and that can cause a lot of drag and i can also sort of make the controls problem harder so by going by going from eight up to twelve it let us shrink all the individual propulsion system elements and in many ways like be able to ignore some of the second order problems and then you describe this as having like two flight systems one basically two airplanes in one physical design but then you're also carrying a payload a payload that's not attached to the body it's hanging right at us at a set length at the turbine length with a package can you talk about how that factors into the flight characteristics and the design of this yeah so this is um this is a this this box in of itself is a very interesting design case study um so you know where the the airplane on its own is about four and a half kilos and the the package is about another kilo and a half so you know a third of this weight is in here so if you look at sort of normal payload fractions for a hovering thing that's that's a lot of weight to be carrying around um and it's also it's also relatively big compared to to the airplane like if you look at this like the size of it compared to the wing this is a lot of surface area to be dragging around um for all those reasons we went to an external payload so if you think about it uh if you size the fuselage to carry this inside you you carry that drag throughout the whole fight um in our case like we're dropping off the box so if you carried on the outside you only need to take that that drag penalty on the way there and you can fly back as a sort of cleaner smaller airplane um but carrying it on the outside makes this sort of part of the airplane with all the vector forces yeah so you know all the loads that all these other parts of the plane have to take this has to take as well and this also can't impart sort of any strange forces on the plane um so everything about sort of the shape of the box the construction box is all to address that and if you look at it sort of from the front it it it forms a relatively aerodynamic shape not you know it's a super aerodynamic shape but um relatively so and because we don't control the orientation that goes up into the bottom of the plane it needs to be symmetric so that like if you design this this nice sort of aerodynamic body one way and then it locks the other way right you might have some problems um additionally you know the assembly of the box needs to be relatively straightforward um you know this is part of you know as we as we sort of partner with merchants as we as we carry other folks goods this is like part of their their operation so they're you know having these things at the restaurant somebody's gonna put together um you gotta make sure that when you put it together it will still fit into the airplane um it can't take up a lot of space um so there's a there's an interesting uh sort of paper folding problem associated with all this of how do you get the shape to fold flat to fold open to be rigid when it's connected and to fit up into the bottom of the airplane it's a combination of aircraft design and packaging yeah yeah it's one it's a very interesting one do so i'm here with tony who's a flight test engineer who runs the flight testing operations here tony this is where you and team have been running test non-stop for the past couple months um to to get certification from faa that's right that's right we set up this facility actually about a year and a half ago and spent most of 2020 preparing for type certification which required us to do uh over a thousand flight hours in in multiple cycles with our aircraft system here and so this was set up to help facilitate that and then earlier this year we operated the system uh through the end of the the type certification requirements and were uh able to uh produce the outcomes that we wanted so what i saw earlier today was an example of what would have been going on with the aircraft landing taking off having these the dummy the packages that were sent you know a kilometer away or so which is is that representative of what a deployment in the field looks like it's you know in australia or virginia where you have this is it somewhat like this with these pads that are for charging but also with these giant qr codes that what it is it is it is uh while you know we don't have residential neighborhoods around here we've designed this to be as representative as possible so the distances that we have the infrastructure that's in place a lot of this mimics what our situation looks like in the production environments in australia so up to this is 50 aircraft but it could scale right it could be smaller it could be bigger their space this much apart give me a sense of what that cycle looks like what what is this aircraft doing now and what does it do you know in that cycle of takeoff and landing sure so it's obviously on the chart or on the charging pad right now this is where it's uh being charged up the batteries that are inside the aircraft and then once a customer places an order uh the aircraft receives a mission to begin its its uh delivery mission to this customer and so it'll take off from the charging pad it will go and meet the merchant at a specified location where the merchant loads the package onto the aircraft the aircraft then travels to the customer's front yard backyard wherever it may be delivers the package and then returns home to charge again wow so up to two stops right going to the merchant uh with the operational range of how many miles or kilometers uh we're looking at 10 to 15 kilometers or so okay so that's that's pretty good and also that gives you a sense of like distance from merchant uh a user an end user of like what type of speedy delivery they're looking for right the same half hour delivery time frame um and it's charging for how long typically uh charging depends on how far it flies so if it does a very short mission then obviously charging time is less but um generally we're looking at anywhere of uh 10 to 30 minutes or so wow and at the scale like i think the thing that we're bearing lead is it's all autonomous right like you have people watching the computer systems that run this uh but they're the missions are sent via back end and up to these 50 of these aircraft here are all independently going on those missions that's correct and it's it's really strange to to think about this because as the pilot in command you're responsible for all these aircraft but you're not actually flying them right so being a man pilot switching into the you know autonomous space was a bit of a a shift that took a while to get used to but now that we're here it just makes a lot of sense it's much safer it's much more efficient and ultimately i think this is the way that we scale these types of operations is there any way to directly tap into these aircraft for direct you know rc control no it's so for for the production operations which would be our in our business environments there's no way to tap into the aircraft itself we're here at a test facility so we do have some extra capabilities here and some of our aircraft you may have seen have a little rc receiver on them and that is for if we need to do any type of overrides while we're doing specific test campaigns and then kind of how aware are these aircraft of each other in the other sensors on them that allow them to know what's around them outside of just the fiducial markers or is it all telemetry so there are some sensors on board the aircraft that allow it to understand the world around it it could see trees it can see obstacles but when it's in the air it's not aware of other wing drones as it's flying and all of that is managed by our backend system where it's able to plan all the different routes and keep everything deconflicted that sounds like that's one of the big like technology hurdles is creating that flight controller system what would be an air traffic controller but for a you know a scalable number of drones yeah and actually i had my undergrad in air traffic management and so it's really cool to be in the space where you know commercial aviation has manned traffic uh air traffic control where here it's fully autonomous and you can imagine a world where maybe our air traffic control system could expand and potentially provide benefit to other airspace cases so brandon tell me a little bit about the the way these aircraft are controlled what's the system like that runs them the system when someone orders something the system takes that order and when the merchant is ready to load the package flight path is automatically created so our systems take into account the aircraft's capabilities the geospatial knowledge in the flight area any airspace restrictions or any other limitations in the exact flight area and a four-dimensional path is generated so that's three dimensions in space and then the other dimension is time right and those volume reservations are like the contract between the unmanned traffic management system and the airplane and the airplane flies within that reservation volume to deliver your package and all these reservation volumes are actually usually unique each order creates a custom plan that's optimal for that time and place to get that package to you and that's taking consideration everything else going on in your system yes you know what other flights the things that may need to get prioritized uh you're saying there's information that's hard coded in terms of maybe geography is weather or something that's factored in yeah we take into account the the weather and the winds um because the airplane does fly uh in airspeed um to get to deliver the package so depending on if there's a headwind or tailwind those timings could be different but we take all of that into account and getting to the point where the system works you're taking like the basically what a air flight traffic controller would be doing on their day-to-day job but applying it to maybe go you know this this these parameters for delivery yeah and it's fully fully autonomous right and we've created the the algorithms um and the system to test and evaluate those algorithms is actually quite expansive we do millions of simulations every single day and we have other algorithms that can look at that data and look at the optimality of of our software to make sure that if there's any issues then we can fix them pretty quickly is it fundamentally a closed system and that it's your your aircraft information that it's factoring in or other external other telemetry that you're also considering we've built we've built the system from the ground up uh to be interoperable with other unmanned traffic management systems um so in the industry there's a drive towards having these uh service suppliers that can supply unmanned traffic management solutions and we have the capability to share those four-dimensional reservations with other systems so we can all deconflict together is that a trend that that industry is moving toward like yeah we see this i see it working reliably right now for delivery but is that not just scalable number of aircraft but size and type of aircraft and this fundamentally this type of system is what how aircraft how the skies will be kind of traffic controlled going forward yeah exactly you know we really view drone delivery is a new form of transportation and i think uh there's going to be many many use cases that we haven't even contemplated and we all need to be able to share the airspace together to really bring this new this new form of transportation to everyone we think of traditional delivery services using vehicles right they work on existing infrastructure there's an understanding between community and you know in government and regulation of what that is here it's a lot of new ground or air right so what is that like to work with different countries or different communities yeah and having them understand that airspace becomes part of that you know delivery experience yeah i think there's a um i think that the way that you framed it is it really is a new thing right so it's it's a new bit of that infrastructure and um you know we try to really respect that right like you're introducing something fundamentally new um you don't want it to be disruptive you want it to be additive to the experience um so you know we do we do a lot of community outreach before we go into an area to offer the service um we do a lot of sort of surveying of the folks who you know are the are the relative few on the planet who are getting to sort of experience this in reality like the feedback of sort of what people are experiencing is really important helps feed into uh what the next versions of stuff like this will look like um and i think generally you know i always look at like reality is the ultimate cure for the what-ifs so when you have when you have an industry you have a technology that the idea of has existed for so long like there's a lot of time to imagine all the things that might happen there's a lot of time to sort of like you know plan out all the scenarios of like oh well this is what it'll look like or these are the risks or you know this is this is what scaled operations will look like but once you actually do it like once you actually go and see it in operation all of that becomes clear like you start to see sort of what the real trends are what where the real risks are where the real risks aren't um and it lets you sort of you know fundamentally reframe the problem so i'm sure one of the questions people have is when is this going to be in their part of the world you know where does this make sense where do you as companies think this makes the most sense uh and what does deployment look like when you reach places outside where you're testing currently yeah so um you know you've seen the test site today but you know this this is a reality for a number of communities uh you know across the globe right now um we've got operations uh in two cities in australia uh in in finland and in uh on the east coast in the us and you know the the delivery experience is uh you know for it for an average customer it's not that much different than you know any of the other sort of on-demand delivery services where it's just sort of become a part of their daily life um so you know you we've got a we've got a little marketplace application you pull it out you say hey you know i want this coffee or i'd like this you know we sell a lot of full roast chickens i want this chicken um and uh you know you go through the checkout and you know unlike uh unlike a regular delivery service we present a couple of different options for you know where would you like the box delivered so um you know it'll say it'll look at a a bunch of constraints around you know what the aircraft needs to be able to you know sort of get from the environment in order to drop off the box but basically if you've got about a 10-foot diameter circle somewhere in your yard the the plane will be able to go and leave the box there so it's not like predetermined group share drop-off points it really is it's really a gps base your backyard yeah it's it's totally sort of you know like a zero infrastructure solution and and in many ways like a sort of a planning free solution so you know if if if that video property happens to meet those set of constraints to get the plane in the plane will come and you know it'll come drop off your box and you know the whole experience might take you know 30 or 40 seconds and the plane's gone and you've got you know whatever whatever things you're interested in getting um you know we're doing right now you know there's on the order of like a flight every 30 seconds or so somewhere on the planet where somebody's getting that reality of drone delivery um the cool part for me is you know having i've liked aviation forever um but aviation is a somewhat rare experience right like you think about you know when was last time you know a lot of people like you you remember the first time you flown a plane uh if you ask somebody the same question about when the first time you drove in a car it's like like like automobiles have been like fully normalized like it's just like it's just part of sort of the day-to-day life um and the cool part here is that we're doing the same thing with aviation like you're taking that experience of aviation which was previously a very rare experience and it's like for those pockets of the the globe right now it's it's it's part of the everyday um and i think it's a it's a you know it's a great way to introduce people to aviation it's also like you know it's offering a very valuable service that you know in comparison to any other form of transportation you know is fundamentally better for the environment fundamentally safer you know we target we target levels of safety for moving the box as being literally safer than any other way you could move that box so you know compared to picking up the box and walking down the sidewalk like having the drone bring it is actually like a net increase in in the safety of the whole ecosystem um and so it's cool to see you know having having spent the majority of my career in this space it's cool to see it starting to become a reality in these these these parts of the world um and you know our hope is that those parts of the world expand and there's more more places where people are experiencing this um and that's why we're you know we're continuing to refine the capabilities we're continuing to build out the test infrastructure and and really trying to sort of you know prove at all levels that you know this is something that makes sense this is something that is safe to operate this is something that is sort of a net positive wherever we we happen to bring it thank you so much adam and your team for inviting us out here to check that out it's really amazing what you guys have done and to see how far it's come since we started seeing glimpses of these at places like maker yeah yeah i remember i was probably three or four years ago now i remember showing you all around a plane that didn't really look like this yet no one could talk about it no no no you couldn't say anything about it but you know i think one of the one of the cool things is that like you know you can you can go on youtube right now and you can see videos that are like shot by customers like people who are actually experiencing it it's not it's sort of outside of that that demonstration space now it's really it's starting to become a real part of everybody's lives so that was really really neat and i think the thing that was maybe the most surprising was how how much it all just kind of worked i mean the fact that this is as a technology that they've been working for for the past 10 years has gotten to a point where it's it's out in the field and you can get standing the middle of that deployment facility 50 aircraft launching and landing was kind of like that a little bit of that magic of technology that we love a couple takeaways i had though one you know the the engineering they've done into the aircraft from the prototypes to what they ended up with incredibly amazing an amount of thoughtfulness things that i never thought or knew they had to consider from the packaging design to you know even the hook right that how how it hooks onto the package all that seems to be stuff they've been really smart and integrated on but maybe more importantly is that back end that software and the algorithms that they've made that platform for autonomously controlling a fleet of aircraft doing the job that an air traffic controller would be doing but on an airport that doesn't need to look like a traditional airport a drone port that can be shipped in a shipping container and put up and remotely controlled over the cloud that's going to be technology that's going to be not only essential as i now realize for drone deliveries but also for any type of autonomous vehicles when we think about the dream of flying cars or even airplanes in the future has could have applications there and then finally there's what adam said at the end that really stuck with me which is what they're doing isn't just you know making the science fiction reality uh the science fiction world the reality in the technology but in the practice and that's the most important part you know as engineers they can invent the technology to make this work from a technical standpoint but none of it matters unless it makes sense and it's practical in the real world and that's gonna have to be a harder thing for i think for them to uh to get over uh the communities uh accepting drones in the sky i think they've already had feedback about you know what the drones sound like as they're flying and making these deliveries because when we think of drone deliveries it's not just getting into remote farmers or to remote people who live you know off the beaten path but they're talking about getting people real items in their backyards and to make drone deliveries as common as your doordash or your uber uh and for that to happen i think they still have a little bit ways to go in both the regulation and also in both in in terms of the acceptance in society uh but good luck to them and thank them for inviting us out there to check out this tech and as novel as it may seem to me and to you their hope is that it becomes really commonplace
2021-10-29 02:30
