Even if the Feds Let Them Fly, Amazon's Delivery Drones Are Still Nonsense
wired.com
wired.com
Imagine if they put a small warehouse in a big city, filled with their most commonly sold items in that city. (For now we'll pretend the warehouse/dronepad is on the roof of their Seattle headquarters, right in the center of Seattle)
I can SEE Amazon's buildings from my apartment right now. It's probably a mile or two as the crow (or drone) flies. That seems well within the range of what Bezos talked about for Prime Air. I also have a rooftop deck on my building that only residents can access. It's pretty big and would be a perfect place for a drone landing pad.
There are probably a hundred similar apartment buildings all within a 10 mile radius of Amazon. They would all be able to easily handle drone deliveries from this small, centralized, micro-warehouse.
Can commercial drones handle straight line flying with building sized object avoidance over a space of a couple miles? I think so.
Will they "in a few years" just like Bezos said? Absolutely.
Just because it won't work in rural south dakota or the suburbs of Detroit Michigan, doesn't mean that it won't work somewhere.
Like the author said, Amazon trialed fresh for FIVE years before taking it out of Seattle. What's to say they won't do the same thing with Prime Air?
(Just in case anyone is wondering, I don't work for Amazon)
Warehouses -- even small ones -- also need to be able to take goods IN. They need truck-accessible streets, loading docks, and they need staff who store, and then pick/pack/ship your goods. Small warehouses that handle high-turnover goods need even more staff, since they need to constantly replenish themselves (since they're small, and can't just take a hundred thousand units in once a month, and since we're stipulating that the goods are high-turnover and leave the warehouse quickly).
It's expensive. Amazon has wonderful warehousing and distribution, which only goes to show that they aren't actually naïve about how expensive it is. Which goes to show that the drone delivery thing was a publicity stunt.
Too small a warehouse, and you don't deliver the customer satisfaction that is presumably the point of a drone delivery, because a trivial number of goods actually get delivered to customers. (Especially since we're talking very small goods in the first place).
They don't have to be downtown.
Regardless, you're thinking about this the wrong way. As a service that must serve everyone and deliver everything instead of as an opportunistic service, which is a smarter way to approach the problem from a business perspective.
Amazon already has warehouses, fulfillment centers, in several major metro areas.
Let me state that again, because it's hugely important. Amazon already has warehouses in some cities. They have to have their FCs somewhere, and the places where they are, like Seattle, already have a comparative advantage when it comes to fulfillment of commonly stocked items.
This changes the problem dramatically. It transforms it from: buy high priced land in major cities, build FCs there, set up drone operations; to: set up drone operations in a few cities where you already have large, well stocked warehouses, and then pilot the program opportunistically in those cities. Then expand the program to other cities depending on how successful and profitable the pilot program turns out to be.
This is business 101, you leverage what you've already got to mitigate risk and reduce development and acquisition costs.
Seriously though, every retailer has figured out this problem. Drive around in Brooklyn, Queens and the Newark area and you'll see hundreds (if not thousands) of working warehouses. Hell, the New Jersey turnpike feels like a corridor in the middle of a giant warehouse.
In my city, a fruit and vegetable distribution place setup a small warehouse for restaurant deliveries about 5 blocks from the central business district. It added a little cost, but now he can deliver a case of fruit to a restaurant in like 20 minutes vs. 90.
1) It's not that expensive to send a Semi filled with goods several times a day to a small warehouse, again that's the norm.. not the exception in retail.
2) Packages are already being delivered from "small warehouses" with all Amazon orders - every city has UPS/Fedex/USPS distribution warehouses that split the big trucks into smaller trucks that then drive to you. Amazon is trying to use the drones to replace the delivery drivers.
Compromise that's doable with today's tech - Amazon could utilize a network of Amazon Lockers in outdoor public places and have the drones deliver the packages to an opening in the top of the locker which the locker then takes and loads into the right slot. The customers can then pickup the packages at their leisure like any other Amazon Locker. The drones would not need to deliver to every location, just the preset lockers which would be in areas with safe and easy access for the drones.
In other cities I doubt it would be too hard to find some buildings that would let them put a shipping container on the roof stocked with some essentials for a reasonable fee.
Stocking stores is a solved problem in big cities. I don't expect a drone warehouse to carry more than a typical drugstore. They could even partner with local stores for order fulfillment.
Remember, this doesn't have to scale out to all places and all items immediately. This could just start off as a way for tech hipsters in Seattle to order condoms on demand. If it becomes successful, they'll build it out as they need to.
1. Last time gas prices went up, many of these rural folks were stranded without food because they couldn't afford the gas to get into town, especially fixed-income older people.
2. There's not as much competition out there. If the local Walmart shuts down because they can't compete with Amazon, the drones might start to look more appealing.
(And, having myself been raised in rural Mississippi and toted a Daisy air rifle, when at play, throughout most of my childhood -- you'd better believe kids would shoot BBs at Amazon drones they saw flying past! That's the kind of fun even a kid who doesn't shoot at songbirds can enjoy.)
these drones will need to have anti-projectile and defensive maneuvering systems from day 1.
Here are a few ideas that could make this work easily. Self driving vans via google plus drones in the back that fly out when they pass the delivery location with the package delivers it and returns to the truck. That is a great starting point and then when drones become better, longer lasting, you start sending them from UPS delivery locations at airports.
This is like that Article about how Amazon is a dumb idea from the 90's. So narrow minded.
You can get an article circulated by badmouthing the hottest tech news, but in this case, it's just a bad misleading article with little basis in truth.
Also the government just opened 6 commercial testing zones, and the FAA just laxed some of the regulations regarding commercial UAV experimentation and flight. Note that these moves happened at lightening speed for the government. Just weeks ago it was illegal for a company to fly with out FAA clearance and a pilots license. This is going to happen. I believe Amazon or Google are the most well positioned to make it happen.
Not only is Amazon talking about introducing drones, it's also entertaining the idea of making UPS or FedEx strictly Business2Business enterprise, as Amazon would be handling the final leg of delivery ultimately to the customers door. I haven't heard anyone mention anything about this. Personally, as bicyclist living in NYC, I can't wait until those disgusting aggressively driven diesel delivery trucks are moved to the strap yard, replaced by the distant buzz of a battery operated drone army.
Also to correct the inaccuracy presented in the article. The drones will be the final leg of the package journey, not the mid term. They will fly maybe 10 miles max, with a top speed of maybe 60 mph max within a city, not, 120 miles from city to city. It will start in one city like Amazon Fresh, and then 2 or 3 more years after that. Not every city all at once like the article insinuates then points out would be impossible. One city with in 5 years still makes Amazons plan a success and it makes this article completely wrong.
I research, build and fly "drones".
For the type of drone you are best informed about, what is the maximum round trip flying range, assuming you adjusted the weight of batteries to maximise flying range and there was no wind and no worries about radio range or line of sight?
What about if you were carrying a 1 lbs / 0.5 kg payload in the outbound direction?
What would the round trip time and average flight speed be?
What would the retail cost of such a drone be, if I wanted to order a kit today?
I don't know if ~tripling that performance is in the realm of non-breakthrough engineering improvements but it seems at least order-of-magnitude close.
[0] http://www.uasvision.com/2013/06/25/unmanned-microdrone-cros...
- Can you ensure safe failure modes for your in-flight drones?
- Who assumes liability when your drones crash?
- What is the blade tip speed during operation? What precautions do you take when flying in populated areas?
- Add rotors. The demo video showed an 8-rotor drone, I can imagine that building it as a pair of parallel quad rotors from the power supply point of view would give you quite a bit of control even at a 50% loss. A single-engine loss is obviously fairly small beans. I'm emphatically not an expert so this is armchair engineering.
- I don't know why anyone would ask this question. The answer is trivially obvious (the operator of the drone, whichever legal entity that ends up being modulo whatever insurance they certainly have). Obviously "it could crash" is hardly novel in other delivery services and yet they seem to survive the liability question without issue.
- Probably quite dangerous. I imagine you want to fly quite high for cruising for obstacle-avoidance and noise anyway so 90% of your distance this isn't a problem. I think the answer here is "don't hit people" just like with existing forms of transportation (or more realistically "be as careful about not hitting people as your liability insurance continues to reward you for").
And with regards to the blade speed, you can make it largely a non-issue by using shrouds/wire mesh. People don't worry about the blade speed of a floor standing fan even as they brush up against it.
Well, the thing about a multirotor is that a motor failure is going to drastically throw off the stability of the aircraft. The controller would have to be able to recognize the failure, kill the faulty rotor hub, and compensate on the remaining motors in very short order to keep the airframe from getting so far off balance that it tumbles out of the sky.
Though, honestly, I'd think the controller or battery failing is a more likely mode of failure than one of the brushless motors suddenly dying.
Like an earlier poster acknowledged, the current hobby grade gear is not really engineered with industrial use in mind, but that's a fairly simple engineering problem which has already been solved by the groups who are using drones commercially (i.e. the military)
I have no doubt what so ever that Amazon, Google etc have the engineering chops to bring the software and hardware being used in drones up to an industrial grade.
Don't assume that because the first thing that you thought of upon hearing "delivery drones" is infeasible that the whole idea is unworkable. It just needs an engineer with a little more imagination
Edit: a couple of obvious solutions to your battery/controller failure mode would be a) use redundant controllers. They're small enough that you could easily have two or more running in tandem and checking eachothers integrity, and b) if you lose main battery you can autorotate down and land somewhere reasonably safe. If you use EDFs/shrouded blades they aren't a safety issue, and building a map of safe landing zones for a city is not a huge ask for somebody like Google.
Now for somebody else willing to use their imagination: what do you do about cheeky individuals interfering with your drone during delivery?
You really don't see how there's a huge difference between balancing thrust between relatively equal power sources and losing so much thrust that the aircraft is inherently unstable?
Because that's kind of a huge issue - multirotors fundamentally rely on having power at each corner for aerodynamic stability. Even brief loss of power to one is going to put you into a tumbling state with no way to compensate for it.
>that's a fairly simple engineering problem which has already been solved by the groups who are using drones commercially (i.e. the military)
What multirotor UAS systems does the military have in service for reliable flight over moderate distances?
I'd also take issue with the suggestion that controlling an aircraft which has become unstable over an axis that you no longer have control over is "a simple engineering problem."
>They're small enough that you could easily have two or more running in tandem and checking eachothers integrity
I was referring to the speed controllers that convert the battery power to AC for the brushless motors. I'd expect commercial grade ESCs to be overbuilt, but they put out a good bit of heat that has to be engineered around. They're also one of the more common modes of failure for the brushless motor systems used in mutlirotors.
>It just needs an engineer with a little more imagination
This is the equivalent to saying "we're going to magic away any problems with this idea."
EDIT:
I missed this.
> If you use EDFs
Seriously? Do you even realize how terrible a suggestion that is? EDFs are even worse in terms of energy efficiency and they'd compound stability issues since they have relatively poor throttle response.
But I suppose we can imagine those problems away too.
I've been thinking that a SawStop-type of technology might be a possible solution there.
Additionally, if you're going to allow the blades to hit anything you more or less need to accept that it will result in a crash, even if you stop them immediately. Better to just guard them from contact in the first place.
In their system, they use elevated platforms where the drones could drop off packages/change batteries. If feasible, this solves the distance problem. They estimated that to cover a 140km square area, it would cost 900k, require 50 towers and 150 drones.
[1] http://www.ted.com/talks/andreas_raptopoulos_no_roads_there_...
"I can't wait until those disgusting aggressively driven diesel delivery trucks are moved to the strap yard, replaced by the distant buzz of a battery operated drone army."
Distant buzz? Are you sure about that? Drones sound like insects, yuck.
Consider horses vs. automobiles, for example. Horses conveniently can be bred, so owning a horse can potentially give you more value than just having a means of transport or labor. Horses can also be trained for a variety of tasks, they can get better at their job over time. Horses are also excellent at traveling along uneven ground. They can travel along city streets or country roads and dirt paths or through forests and open fields or even through mountain passes. In comparison, cars, especially early cars like the model T, are inferior to horses in a great many ways. But it turns out that the few key advantages of automobiles are more than sufficient to make up for the down sides, which is why automobiles won out.
Similarly, if you imagine that drones must be a 100% drop in replacement for FedEx/UPS you'll rapidly come to the idea that the whole endeavor must be nonsense. But if instead drones only provide a complementary service then they can still be a feasible addition to delivery options. Even if drones only delivered high priced orders (say, more than $1k) to businesses with special drone landing spots it would still likely be a profitable and successful venture. Being able to have something delivered the same day in mere hours is a game changer in a lot of ways.
Imagine how much that would put Amazon even more into the existing business space of, say, home depot, as a simple example. If a home builder could order a replacement circular saw to be delivered to the work site in an hour without having to send someone out to a store that's a big deal. And there are countless examples of where the value of having some durable good delivered as soon as possible is immense, and would more than justify exorbitant drone delivery fees.
Your Home Depot reference made me think of the 24x7 propane dispensers that are in front of many Home Depots. Enter your credit card, put your old tank in the cage, take a fresh tank and you've got propane. It costs more, but adds convenience.
Amazon could easily partner with smaller-box retailers like CVS, Walgreen's, supermarkets or convenience stores to make store branches drone delivery points. If I need a Macbook Air or a circular saw now, I pay $25 and a 5 minute drive/walk to the drugstore for pickup vs. driving to the mall or waiting two days.
I could see something like this working for Pizza Hut though.
UPS in my neighborhood has done something similar in Christmases past--they have trucks bringing in bulk and bicycle delivery doing house by house. I didn't notice it this year; I don't know if it was a failed experiment or if I just missed seeing them.
War is Boring has an interesting post about using unmanned aircraft as a weapon-truck wing-man instead of an independent operator: https://medium.com/war-is-boring/64b87d155545
A computer can know exactly where the trucks are and when they will be in the vicinity. The delivery time can be fairly accurately estimated on the spot.
In an urban environment, bikes could do the last mile, with the robots delivered to pre-designated drop off points. A predetermined location would be easy on the computer.
But I think your idea works better in a suburban zone. The trucks exist and are constantly driving around. It is a question of how to re-prioritize the work on the fly.
Get the package to the drop off point in 30-60 minutes by quad-copter. Get the package from there by bike in 30-60 minutes. There would be uncertainty based on the availability of human messengers, but routes and delivery times can be estimated/calculated by computer. "Estimated delivery in 79 minutes. Click yes?"
What's even more impressive: they hand-picked a story that is so good that even months after the pseudo-hoax went public, you still have hundreds (thousands?) of really smart folks on HN and other forums discussing the feasibility of the idea, the economics, or if an hex-copter wouldn't be better than a quad...
And just think about the wave of CS students applying for a job at Amazon, "because I want to work on cool projects". Certainly more cost effective than driverless cars.
Amazon Prime Air is already a success. They don't need to fly anything.
1. Battery life is insufficient. This one is true, batteries are not there yet.
2. Warehouses are far from target areas. This one is true as well, in order to make drones work warehouses/dispatch centers should be fairly close by. One solution would be to utilize hyperloop kind of system to quickly deliver drones to where they need to go.
3. Problem with delivering. This one is super simple. Make drone subscription costs high and deliver to customer a drop off pad. Make a clause in prime air agreement that you must find enough open space on private land for drop off pad to serve as drone beacon and unload zone. Drop off pad can have RFID/other location beacons that would let drone know exactly where it is.
TL:DR author is right, while this can be built right now - it is simply not economical, and likely will not become economical in near future due to point #2 above.
Based on history, it's always unwise to say that something futuristic can't be done.