Yesterday Jeff Bezos, Founder and CEO of Amazon, announced during a 60 Minutes segment that his company has already begun work on the future of shipping, called Amazon Prime Air. In short, orders that weigh under 5 pounds will be delivered to your door in 30 minutes, by drone.
This is definitely exciting, but exactly how much does Amazon have to accomplish between now and Jeff’s launch goal of 2015? Getting the FAA onboard will be hard enough, but what about actually getting shipments out safely, when that time finally comes? Is this even possible, or simply a publicity stunt by the e-commerce giant? They’re definitely not the first to think about doing this. Matternet has been working on bringing drone-supported shipping to areas of the world where roads aren’t common, or structurally sound enough, to handle everyday deliveries. CEO Andreas Raptopoulos talked about his vision at May’s Hardware Innovation Workshop.
If Amazon is really going for it, here are the main challenges and some of my thoughts on how Amazon will handle them:
Probably the easiest to deal with. Amazon says they’re shooting for 30 minute deliveries, which I’m assuming means 30 minutes from take-off to landing, not order to landing. Jeff says they will deliver to within 10 miles of an Amazon Fulfillment Center, which is doable if the octocopter can go at least 20mph. The challenge here is giving them enough battery power to survive the trip to the customer and back home. Carrying that much weight at that speed for up to an hour is going to require some heavy batteries.
Although octocopters can hold their own in a decent wind, sending them out in rain or snow with wet weather protection may not be worth it. Will Amazon risk sending orders out in storms, or automatically suspend shipments until bad weather passes? A powerful weather tracker that automatically adjusts order departures definitely isn’t out of Amazon’s scope.
In order to deter theft and destruction of the drones, I imagine they’ll come equipped with at least one camera. This could be used along with GPS to locate any thieves, once the unit leaves it’s pre-planned path or loses power. Video footage will also be useful in case of a legal situation and can play a big role in the next challenge: navigation.
Taking off and heading for a coordinate is easy, but when it comes to landing in a residential (or urban) area, pure GPS won’t be enough to keep the drone, and everyone in the surrounding area, safe. To reliably land on a customer’s doorstep, as seen in the video above, Amazon can use cameras and computer vision to make sure they aren’t land on a tree or person. But where will they land for customers that live in apartment complexes, high-rises or dormitories? Will they send out teams to “pinpoint” ideal landing locations, similar to the way Google does Street View? Personally, I’ve love to climb to the roof to receive my package – maybe I’ll even get a text when it’s a mile away so I can run up there and pretend to wave it in. That would be fun for me, and the most convenient landing spot for the drone.
The most important aspect of this puzzle. The instant one of these falls out of the sky and lands on or near someone, it’s not going to be good for Amazon. How do you think they’ll prevent this? By making sure everyone in the surrounding area is absolutely aware of the drone’s presence? Hefty safety barriers around the blades, backup power, emergency parachutes?
The one thing Amazon has the least control over. Jeff says his 2015 goal is the earliest possible arrival of Drone-related FAA Rules and Regulations, but even that’s optimistic. If I can use this service by 2020, I’ll be happy. When do you think the first real Amazon Prime Air shipment will leave the loading dock?
Of all of these, which do you think will act as the largest roadblock in Amazon’s push for a sky-filled future in shipping? Don’t get me wrong, if there’s one company out there that’s in the best place to do this within the next few years, I think it’s Amazon. Who knows, maybe the drones will outnumber the trucks soon.
96 thoughts on “Pie in the Sky? Technological Hurdles for Amazon Prime Air”
I’m so glad they are working on this… but if you live within 10 miles of an Amazon Fulfillment Center you should just go pick it up yourself when you are out and about ;-)
Unless they had fulfillment centers every 15 miles I don’t see much value in this other than publicity.
Yeah, agreed. It will only be able to service a small percentage of the Amazon customer base at launch. They are constantly working on new fulfillment centers (they’re building one in NJ right now), but again, not enough to reach a bunch of people if the range is only 10 miles.
A single distribution center could service almost all of Philadelphia and a significant part of it’s suburbs. Two could service the majority of the NYC region and the surrounding cities like Newark and Jersey City. LA might require three or four would get you the majority of the greater LA area (which is what, 18 million people?). It’s unlikely that this will support rural areas anytime soon, but most urban and many suburban areas would be pretty easy to cover. Not to mention that this would not be entirely exclusive of the delivery truck model. A truck could go out on it’s delivery route, and have 4 of these launching from and returning to the roof for battery and payload replacement. All light packages could be automaticaly delivered along the route, while the driver delivers the heavy items.
Keep in mind amazon has a HUGE amount of info on people’s purchasing habits. Imagine a UPS sized delivery ruck full of items for whatever Amazon predicts a given city will buy in the next 24 hours. Add 8 UAVs that take off/land to/from the truck. That single truck might well make quite a few deliveries within a day. Sure less popular items won’t arrive within 30 minutes, but quite a few things would.
I can’t speak for any other major cities, but I can tell you that here in Chicago ordering the item; walking to where my car is street parked; driving out to the a distribution center ~10 miles away; waiting in line at the center; driving back home from the distribution center ~10 miles away; finding a street parking spot; and walking back to my apartment can, literally, take hours. 20 miles round-trip driving is no joke in a major urban center (to say nothing for all the other stuff that would also be eliminated).
I had no idea Amazon even supported will-call at their fulfillment centers. How smooth is the pickup process once you actually get there?
“Wow, Amazon delivered the latest issue of Make and a whole bunch of robot and copter parts! Let’s start hacking!’
Seriously though- this is Jetsonian futurism at it’s worst. This is just a publicity stunt to make us more comfortable with increased drone traffic. Just like the ‘convenience’ of texting ushered in a new, easier way to monitor our communications, this is just techno showboating with an ulterior motive.
But, it really *isn’t*. The current order fulfillment model is terribly inefficient, unsatisfying, and DANGEROUS. Look at the traffic fatality statistics. Look at the traffic damage statistics. Look at auto insurance rates for commercial vehicles. Piling a bunch of expensive boxes into a truck that drives down almost every street in the US is stupid, and the only reason we do it is that we don’t currently have a better model. This is a better model.
Incidentally, 10 miles in flight is a lot more than most people assume it is. I live near Philadelphia, in that I can get to the center of the city in a bit less than an hour under good traffic conditions. I’m just about 10 miles from the center of Philadelphia as the crow flies. A fullfilment center in Hoboken could deliver to all of manhattan, most of brooklyn, some of queens, and some of Staten Island too (and if they had downtime, 10 miles into Jersey). I’m pretty sure that’s considerably more than your standard delivery vehicles coverage area.
Well at least we are using technology to feed the hungry, deliver life-saving medicine, ….or something
Well of course there are people using drones for socially responsible things. Matternet, which Eric mentioned in his post is an example.
Then we have a series of posts about aerial drone use to help park rangers in South Africa police game poachers.
People shouldn’t expect every use of aerial drones to be either altruistic, or on the other hand to be offensive and intrusive (military drones and police surveillance). In between there are companies that just want to make a buck.
I can’t imagine how they will zing around!
I will be SHOCKED if battery technology is up to this in less than 10 years.
Pleasantly shocked, of course.
The power cell has to be capable of much more than a best-case trip. Consider that the buildings in major cities tend to funnel and intensify winds. Next, it might be pretty common to have to route around your least-time option because of emergency services. Then, you have to be able to carry the package back, in case the recipient can’t acknowledge receipt (“I’ve fallen, and I can’t get up! How ironic, Amazon is delivering my emergency pendant right now!”). And, in that case, a 10-minute flight, 10-minute wait, and 10-minute return is plenty of time for weather conditions to change enough to require a change of route.
On the other hand, maybe a tiny internal combustion engine could run a miniscule generator long enough to overcome those restrictions.
Thanks; all very good points. This is why landing on rooftops of large buildings would simplify the process for the drone’s flight path planning, although definitely not for regulations.
I imagine Amazon will have to have teams of people on the ground ready to pick up fallen drones, as well, for the other reason you mention – who knows what sort of adverse conditions they’ll have to deal with at a moments notice while flying between buildings, etc.
Why would you need to acknowledge receipt? I haven’t signed for a package in ages, and my house orders from Amazon all the time. They just get left at the doorstep, for me to find when I go outside next. At least with this there will be the potential for a text or email at the time of delivery.
Sure, that’s a point, but why bother with instant drone delivery if you are just going to leave it on the front step? I am assuming this would only be used for high-value or time-sensitive items. In other words, exactly the kinds of things you would want a delivery confirmation for.
You mean important like the skate key that the guy in the video bought? drones would likely be cheaper than currently delivery systems, and more importantly, it would allow amazon to tap into the immediate gratification that buying from the store gets you.
Let’s say that you’re working on a project and you need a part. If you had thought ahead, you could get it at amazon, but since you didn’t you need to drive to a brick and mortar store. With this, amazon becomes more convenient than the stores in your town.
Piracy will be the biggest issue, though how one would know what one was pirating is an interesting question. Still, it would be nice to snag an octocopter for free…
I like what I saw in NYC with regards to eBay. They had a box service in some convenience stores. Order something and have it delivered there. If Amazon takes this approach, they could have the drones land on the roof, and the delivery boxes, (QR coded or radio tagged) would slide down into the appropriate delivery box for secure customer pickup. Combining both drone delivery and eBay centrally located boxes would be the best solution.
Thanks for sharing. I wasn’t aware eBay was working on something like that in NYC – very cool. When thinking about what other companies can accomplish this huge task, other than Amazon, eBay did pop into mind…
Eric, I stand corrected by my wife who reminded me that it was actually an Amazon mailbox center (AKA Amazon Locker), not eBay, in NYC. I had my son take a picture of it, which I have mailed to you separately.
The infrastructure, it appears, is already being built.
No problem – thanks for sending the photo! That’s pretty neat.. had no idea that sort of thing existed.
I hate to be the fear-monger here, but isn’t there a security issue wherein you could just drone a bomb to someone’s house and they will be acculturated to trust and open it up? For a sophisticated attack, you’d have to know that they ordered something from Amazon, but it could be super dangerous even if you didn’t do any hacking at all, just guessed that someone from the home would assume a purchase had been made.
Same could be said for current methods. Would someone question if you put a real looking shipping label on a box and left it on their driveway?
Or just, you know, mailed a bomb to someone?
At the same time, being a New Yorker, I think there’s an awesome potential to use a delivery service like this to coordinate your possessions for when you’re out on the town. Take my laptop home so I can go out / bring me nicer clothes so I can go party / bring me healthy food so I can keep working would make city living twice as cool.
I completely agree. This is a perfect example of something that has soooo much potential, on both ends of the spectrum. Sure, someone could load up a bomb for an unsuspecting customer to unbox or some nice clothes for unprepared party-goers.. all the more reason why this is such a massive undertaking. We’ll see what happens :)
Thanks for commenting.
New sport for RC plane/glider hobbyists: hunting Amazon drones.
Seriously, the pay off could be huge, and the risk to getting caught low. Video won’t be able to find a potential pirate drone’s pilot, just hide behind a tree or car until your RC plane crashes into the drone, then go pick it up, disable the GPS (with your heel), free drone, free goodies.
There is a whole sub-hobby of RC Glider enthusiasts that just love to crash into each other, they’re already really good at air to air slope gliding combat. Those skills will translate perfectly to dive-bombing Amazon drones, which will be slow by comparison to combat gliders. The effort it would take to program in evasion protocols, plus the extra energy needed to execute them, will make the whole thing fail.
Someone needs to take the CEO of Amazon out to their local slope-gliding hotspot with one of their prototype drones, see how it stacks up against the real-world. Good luck, Amazon, this could be really entertaining.
If they honestly do not think that people would steal this they are stupid, sorry. Net the copter then use a similar device to block the GPS. Their new toy is now someone else’s. They should spend more time helping employees.
Sounds too sci-fi to me.
So you are assuming the amazon drones don’t use visual way points, dead reckoning, or cell tower based positioning. Then there’s an attack, you successfully take out the done, then steal the box. To steal the box you likely need a $5k and up drone (8 rotors and can pick up a 5 pound box) and assume that the victim drone doesn’t call it’s friends to follow your drone home while sending it’s gps coordinates to the police.
Given they my current phone has 2 cameras (front/back), 3 methods of finding it’s location (cell tower, GPS, and GLONASS), 3 wireless networks (bluetooth, wifi, and 3g/4g/lte), and 3D compass and 3D accelerometer and sold for $350 (off contract) I don’t really see it being particularly easy to target, sneak up on, successfully attack, and successfully get away with a 5 pound box from future amazon drones.
Seems easier to just drive around the suburbs and steal boxes from front doors.
Given that stealing an amazon box and selling it “used” isn’t going to fetch a particularly high price. You really think that trying to pick off amazon drones will be profitable and not land you in jail?
Batteries, flight times, carrying capacity are all doable. FAA approval will be tricky. I expect restrictions of flight under 500 feet, no flight within a certain radius of airports, failsafe & redundant systems, etc. At first, they may be required to fly “human in the loop”, that is a person monitoring via remote camera in real time. Once the concept is proven safe, they may be able to allow autonomous flights. In some ways, autonomous, pre-programmed flight is safer, as there is no control signal to be hacked into.
Real time telemetry over those distances may require radio repeater infrastructure, especially in hilly terrain or where there are many tall buildings.
Navigation is the most difficult aspect, in particular the last 200 feet. GPS can guide you to an address, but as the article says, picking a landing spot will be challenging. What I expect is that to use the service, you will have to sign up. You will then be sent a tarp with a landing target on it. Perhaps the target will be something like a QR code that identifies you. It will then be your responsibility to place the tarp in your yard in a safe landing place, away from trees, overhead wires, etc. Part of the deal can be that you will be responsible for damage to the drone if it results from you picking a bad landing spot. GPS would then guide the drone to your general area, where a camera system would spot the landing zone tarp and land.
I don’t think theft will be a major issue. The drone only has to land momentarily, drop its package, and take off again. Running up to grab it with 8 whirling props would be foolhardy. And yes, I expect they will be tracked by GPS in real time. Cameras may be a political issue – the privacy implications of an air force of Amazon spy drones would be a difficult sell, even if it was argued that they are necessary for safety & security.
Given how I see the technology working, I suspect the only viable market to be smaller towns & suburbs, where people have adequate yards for landing areas. I don’t see dense cities, apartment buildings, or office buildings as viable for the flying drones.
However, in those environments, driverless cars may be a better alternative. Think something the size of a Smart Car, with mailboxes on the outside. It drives around town, stopping at each delivery spot. It sends you a text message with a mailbox number & a code. You head outside, go up to the car, find your mailbox, punch in the code, and take your package. The car then heads off to the next spot. Navigation for this model is actually much easier (though parking could be a challenge), and you don’t have the capacity restrictions, or the worry that something’s going to fall from the sky.
Thanks for the thoughts, David. I really like the QR target idea, and appreciate you bringing up driverless cars at the end there. Those are easy to forget about amongst this new excitement, and much closer to “release”.
Did amazon promise battery powered? Why not say alcohol? Natural gas, or anything else that burns relatively cleanly and has 10x the energy density of batteries. After all military drones are rarely electric. Quad rotors vary RPM which depend on electric motors and short sensor feedback loops. But it’s just as feasible to use constant RPM and vary the blade pitch. See recent news items on quad rotors from MIT. Blade pitch quad rotors can work well with an arbitrary constant RPM motor of sufficient power/weight.
Weather seems easy, when your address doesn’t have good weather amazon won’t display the “deliver in 30 minutes button”. UAVs could always abort if there’s a surprising change and your phone would notify you to enable you to cancel or reschedule.
As for theft, amazon could have you order from your phone, enable GPS, and go to a place with a 10 foot x 10 foot clearing. Then they can display a cute little progress animation on a map. If the drone doesn’t detect safe conditions and that you are within 10 meters it would give up and go home.
Navigation is pretty easy. They need a map of the highest point relative to whatever FAA requirements are (like nothing above 500 feet/not near airports/etc.) and a sensor (like a kinect, lidar, even ultrasonic range finding) to see if it’s safe to approach. Then start descending when the customer is detected as nearby (via an app running on their phone). If you are in an apartment building head to the roof, nearby park, or equivalent open area.
I don’t see any technical limitations.
You could also charge a battery with a generator off of fuel, as mentioned above (Tommy Phillips). Kind of a hybrid car/diesel train thing.
Google UAV fuel cell to find an article from popular mechanics on a fuel cell lasting 4 times longer than a gas powered UAV. I suspect gas UAVs last at last 4x as long as batteries.
Good points – there is a variable pitch quad for sale now, called the Stingray, and it’s very impressive: http://www.youtube.com/watch?v=TnGhEInTXYc
MIT is doing some cool work with similar tech, as you mentioned.
Very good point about the possibility of non-electric rotors. I don’t know much about speed control in that kind of system, and battery technology is one of my hot buttons, so I naturally made that assumption.
This comment stream has really expanded the borders of the assumptions, as well. Maybe instead of a tiny ‘copter barely capable of carrying a five-pound package, we might see a beefier system, with some kind of fuel-driven energy system, a generator (like an automitive system, parasite to the primary system?) to provide both backup/emergency power and power for the avionics, possibly delivering multiple packages on a single sortie.
I like the idea of a mobile base (possibly an autonomous vehicle), that could provide both street-level service and airborne drone support. Staging for air delivery of popular items? Targeted specials (our truck is in your area with size 8 black Nike Chucks: order now for 15% discount)?
Dang, this topic has a lot of places to chew on. I know this kind of thing has been a topic of discussion for a while, but I never really paid attention to it before now.
Yet another reason to want to play with a multicoptor.
The issue of a safe landing spot and final navigation to that spot is easily overcome by a cheap yet technical solution with Amazon Prime Air customers deploying a Bluetooth LE beacon on a safe, flat, unobscured landing area.
Or just a cell phone with GPS. Much like any number of apps that will relay your long/lat position real time. It could work in both directions of course so you get up to the second position of your anxiously awaited delivery.
I think the really huge obstacle is economics. Are 250 drones ever going to be as cheap as one driver in a brown truck with 250 packages in the back? I don’t think the FAA is Jeff’s problem. I think his nemesis is going to be his own accountants. As it is, I suspect the cost of the program so far is being picked up in Amazon’s advertising budget. The cost of this octocopter and the film crew is just a small fraction of what it would have taken him to actually buy as much exposure as this PR stunt has genereated over the last few days.
Edward — Marcus Wohlsen of Wired wrote about this in depth yesterday: http://www.wired.com/business/2013/12/amazon-drone/
He talks about the economic and logistic challenges of shipping same-day and how it requires a point-to-point shipping model, unlike next-day or multi-day, which uses a hub-and-spoke model. Definitely worth the read!
Read it. Exactly! Go Marcus and thanks Eric.
The Hub and Spoke problem is a significant one … IF Amazon is same-day-delivering its own stocked goods. But what if Amazon develops a pilot program, and then leases it to local stores for delivery of THEIR goods? What if Amazon front-ended Radios Shack, and then Radio Shack licensed or leased the drone delivery system. You could have your Amazon bought Arduino or cell phone delivered in 30 minutes from the Radio Shack that is in the Christmas-rush-crowded mall, five miles away. And never mind Radio Shack, this could be the re-emergence of the Mom and Pop stores, front ended by Amazon and delivered via licensed Amazon drones. Hmm…
The economics look pretty good to me. Read the wired article on the amazon drones. It’s example includes a warehouse 60 miles from SF and average UPS truck full of 120 packages. Assuming a driver makes $50k a year ($200 a work day) that’s $1.66 a package. That of course ignores truck cost, warehouse costs, manager salary, benefits, gas, etc. Something like an hour to drive to SF, then a package every 3 minutes (often a single stop will deliver multiple packages), then an hour to drive back to the warehouse. Approximately an 8 hour day.
Now image a special truck that a UAV can land on and a rack of shelves for the UAVs. At various places around SF the delivery person pulls over, walks into the back of the truck, pulls box off shelf, loads it into UAV, and pushes the go button. Seems reasonable to launch one every 30 seconds or so. Given that SF is only 15 miles on a side or so maybe he makes 8 stops. Each UAV has to fly say 2 miles to the delivery at 20 MPH and 2 miles back. That’s a a round trip of 12 minutes. So the driver launches 24 UAVs then starts handling the returns. Each one gets fueled (squirt a few ounces of fuel in) then either racked or sent out with another box.
So in a day said delivery driver might delivery say 250 packages in a day. Might need a bigger truck, but since it wouldn’t need to drive to 120 residences a day it doesn’t have to be as small/maneuverable/easy to park as the traditional UPS truck.
Now assuming the above now add that Amazon knows exactly what amazon’s most popular items are, just released computer games/books/movies/ipad/whatever. Exactly the thing that amazon want you to order from them instead of someone else because consumers want instant gratification with the hot/trendy widget. Said trucks could carry say 100 of the most popular items and include a printer, display screen, and a bell. So when the delivery person is loading UAVs if he hears the bell he reads a display, grabs the mentioned box, slaps a label on it, then loads it into the next UAV.
Amazon would of course know where the delivery trucks are, what stock is in them, and where the customer’s home address is. I suspect amazon’s volume means that dozens or 100s of trucks could be handling a city the size of SF. So when it detects you browsing the latest widget it would say “Have this in 30 minutes”.
All while delivering the package at the same or better price than UPS manages.
OK, lets look at your scenario. You still need the warehouse; you still need the manager; you still need the TRUCK and its logistical base so you still have the $1.66 cost per package to which you have to add the cost of your drones and lose the volume in the truck that could have been used for packages OR you have to have a bigger more expensive truck.
Don’t get me wrong. I’m a big fan of technology. Half of my job as a CIO of logistics organization was to talk management into adopting innovative technology that let us do things better, faster, cheaper. But the other half of my job was talking them out of implementing multimillion dollar solutions to multithousand dollar problems.
In the supply chain the cost of moving a pound of anything (from cheapest to most expensive) is — water -> rail -> road -> air — and that increase in cost is exponential: not linear. Conversely the speed and flexibility (from more to less) runs the other way — air -> road -> rail -> water. The goal of the transportation equation is to find where those two points cross. Drones have a place in that equation but its at the high cost/low weight/high flexibility end (delivering cargos in the mountains, deserts, jungle etc) not in suburbs or cities. Famously, people thought supersonic air travel was a great idea until they saw the cost.
I agree that drones cost money, and a stack of drones is likely going to take up 2.5 feet x 2.5 feet x however tall the delivery truck is. Each one is likely to be approximately 4-6″ tall.
But as you well know there are many other costs involved. Speaking strictly of physics it’s much more efficient for a UAV (weighing a few pounds) to carry a package a few miles than it is for a 10,000 pound truck, doubly so when consider traffic, parking, and that rarely is a road as direct as the flying path.
So it comes down does the cost (R&D, purchase, fuel, space on truck, and maintenance) of the UAVs make up for more than doubling the delivery rate. Needing to send less than half (truck/driver per 250 packages instead of 120) as many trucks into a city would be a significant savings. I also suspect that many people would happily a $2.00 premium to have a package in hours instead of days.
I would thing environmental groups will approve, less pollution per package. Needing less than half as many trucks is just part of the benefit. Each truck will have less start/stops and
a shorter/more direct route, and stick to the larger roads instead of entering into each small neighborhood.
It’s Kozmo.com all over again.
Let me get this straight: we should cede rights to property, privacy, security, and tranquility so that people can maybe get their baubles just a little bit faster?
How do UAV (unmanned aerial vehicle/drone) delivery widgets steal your rights to privacy? Violate your security? You think UAVs are louder than (often diesel) delivery trucks? I suspect I wouldn’t notice a half dozen UAV deliveries in my neighborhood a day, but I hear the drone of delivery trucks even a few blocks away. I’m convinced when I’m anxiously awaiting a package that delivery trucks circle my house for hours taunting me. I welcome less noise, parking issues (delivery trucks often double park), and less diesel fumes (trucks don’t have to be as clean running as cars).
Sure UAVs can see from above, but I suspect the most viable places for such deliveries will have numerous tall building anyways. Nor would I expect amazon to pay for a full video feed from them when all they really need is telemetry data and make a picture of the delivery location to send the customer to know where/when the package arrived. Not like truck drivers can’t already see over the common residential fences and into windows because they are in higher trucks and often standing instead of sitting.
Comments are closed.