How Amazon’s Drone Plans Could Work, According to An Expert

“We’re convinced that it’s going to be the next big paradigm in transportation.”

Amazon CEO Jeff Bezos addresses a press conference to introduce new Amazon and Kindle products in New York, September 28, 2011.
National Journal
Alexis C. Madrigal, The Atlantic
Add to Briefcase
See more stories about...
Alexis C. Madrigal, The Atlantic
Dec. 2, 2013, 5:36 p.m.

Two and a half years ago, An­dreas Rap­to­poulos foun­ded Mat­ter­net, a com­pany de­voted to cre­at­ing a net­work of drones that could de­liv­er light­weight pack­ages. It’s start­ing with med­ic­al ap­plic­a­tions, with plans to ex­tend from there to “bring to the world its next-gen­er­a­tion trans­port­a­tion sys­tem.” To hear Rap­to­poulos tell it, when the his­tor­ies are writ­ten in a few dec­ades, people will think: elec­tric grid, road in­fra­struc­ture, tele­phone lines, In­ter­net, mo­bile phones, and … tiny fly­ing drones.

“We think about it not just as a point-to-point de­liv­ery, but as a net­work. What can you do if you have many sta­tions of these fly­ing drones?” Rap­to­poulos said. “What can you do with a sys­tem like this in the de­vel­op­ing world, in our cit­ies, in our mega­cit­ies? We’re con­vinced that it’s go­ing to be the next big paradigm in trans­port­a­tion.”

Of course, last night, Amazon’s Jeff Bezos re­vealed Amazon Prime Air, his com­pany’s plans to use drones at some point in the fu­ture to de­liv­er pack­ages to cus­tom­ers.

It all sounds a little crazy. And we can all think of many ob­jec­tions to drone de­liv­ery net­works. They won’t have enough range! People will shoot them down! What if they crash! They can’t op­er­ate in places where you can’t get a steady GPS sig­nal!

Giv­en that Amazon seems un­likely to give real an­swers to these ques­tions, I con­tac­ted Rap­to­poulos, who has spent the last sev­er­al years deeply en­gaged with these prob­lems since work­ing on a pro­ject at Sin­gu­lar­ity Uni­versity in 2011.

First off, why cre­ate a net­work of fly­ing drones at all?

“You have the tech­no­logy that can help the most dif­fi­cult part of de­liv­ery: The last-mile prob­lem. You have a light­weight pack­age go­ing to a single des­tin­a­tion. You can­not ag­greg­ate pack­ages. It’s still way too com­plic­ated and ex­pens­ive. It’s very en­ergy in­ef­fi­cient,” Rap­to­poulos said. “UAVs or drones deal with the prob­lem of do­ing this very ef­fi­ciently with ex­tremely low cost and high re­li­ab­il­ity. It’s the best an­swer to the prob­lem. The ra­tio of your vehicle to your pay­load is very low.”

Part of the ar­gu­ment is that our cur­rent last-mile de­liv­ery sys­tem can seem kind of ri­dicu­lous, at least from an en­ergy ef­fi­ciency point of view.

As Rap­to­poulos put it: “In the fu­ture, we think it’s go­ing to make more sense to have a bottle of milk de­livered to your house from Whole Foods rather than get in your car and drive two tons of met­al on a con­ges­ted road to go get it.”

Of course, we could also build walk­able neigh­bor­hoods that don’t re­quire driv­ing as of­ten as we do, but walkab­il­ity re­quires dens­ity — and even places like San Fran­cisco some­times balk at the sorts of build­ings that en­tails. And we’ve got a lot of low-dens­ity in­fra­struc­ture in place that isn’t go­ing away any­time soon.

How quickly could this all hap­pen?

The tech­no­logy is get­ting there. It is not as good as people as­sume. There is a lot of hype around what drones can do today.

Amazon has said their timeline is de­pend­ent on rule­mak­ing for ci­vil­ian drone flights by the Fed­er­al Avi­ation Ad­min­is­tra­tion. “We hope the FAA’s rules will be in place as early as some­time in 2015,” their web­site con­tends. “We will be ready at that time.”

But even Rap­to­poulos, a boost­er of the tech­no­logy, is skep­tic­al of that timeline.

“It’s not go­ing to hap­pen in the U.S. in the next two or three years. Even if you’re op­tim­ist­ic, it’s not go­ing to hap­pen be­fore three to five years,” he said. “Our as­sump­tion is that this may hap­pen in oth­er places in the world first. It may hap­pen in the glob­al south in coun­tries that are de­vel­op­ing and don’t have al­tern­at­ives. There, it’s not about cost re­duc­tion but giv­ing ac­cess when you don’t have ac­cess at all.”

Is this tech­no­logy any­where close to ready for mass de­ploy­ment?

“The tech­no­logy is get­ting there. It is not as good as people as­sume. There is a lot of hype around what drones can do today. We see it in bi­o­tech­no­logy. We see it in ro­bot­ic tech­no­logy in par­tic­u­lar,” Rap­to­poulos said. “We need to re­solve a lot of things be­fore we can get to the point where it is re­li­able or ef­fect­ive.”
Mat­ter­net has de­veloped half a dozen drone pro­to­types and tested them in Haiti and the Domin­ic­an Re­pub­lic. “The next step is to op­er­ate a net­work for a month in a real loc­a­tion where it solves a real prob­lem,” Rap­to­poulos said. “The next big item on our cal­en­dars is how we can get that tri­al — and we think it’s go­ing to hap­pen in the first half of next year.”

But what about the range of the tech­no­logy? The bat­ter­ies aren’t good enough, are they?

“We star­ted at 10 kilo­met­ers and got to 20 kilo­met­ers. Even without as­sum­ing a bat­tery break­through, we see a 5x in­crease in the range. If you factor in some ad­vance­ments to bat­tery de­vel­op­ment, you might see an­oth­er 3x in­crease to 300 kilo­met­ers,” he con­ten­ded.

In the near term, Mat­ter­net is still try­ing to get to 100 kilo­met­ers by op­tim­iz­ing their sys­tem and sub­sys­tems. But Rap­to­poulos is op­tim­ist­ic that it will hap­pen. “There are quad­copters out there that can do 50 kilo­met­ers a day, but they cost 10x what our tar­get cost is,” he said. “How can you get the tech­no­logy bet­ter while keep­ing the cost down? Tech­no­logy is pretty good at that. It’s in­ev­it­able it’s go­ing to hap­pen.”


But what about re­li­ab­il­ity?

“We need to design these vehicles to make sure they don’t rep­res­ent a pub­lic risk. If we’re able to do that, we’re ready for prime­time,” Rap­to­poulos re­spon­ded. “The way to un­lock reg­u­lat­ory ap­prov­al is to show with really good data, 99.9999999 — sev­en nines — per­cent re­li­ab­il­ity. Then, of course, you’ll have reg­u­lat­ory ap­prov­al.”

So far, he doesn’t think that any of the burrito or pizza de­liv­ery stunts qual­i­fy as any­thing close to a real solu­tion to the de­liv­ery ques­tion.

“People say­ing, ‘We’re do­ing this kind of de­liv­ery in China.’ Or talk­ing about burri­tos, pizza, ta­cos, whatever. All this stuff is BS. In or­der to get the de­liv­er­ies work­ing as a sys­tem, the drones need to be re­li­able. Cars are re­li­able. Planes are re­li­able.”

He con­tin­ued, “There are three things you’re try­ing to op­tim­ize for re­li­ab­il­ity: time, de­vel­op­ment, and keep­ing the cost per vehicle down. The more time and money and cost per vehicle you al­low, the bet­ter the re­li­ab­il­ity. For Amazon’s ap­plic­a­tion to make sense, the vehicle cost should be be­low $20,000. If its $100,000, it’s not cost-ef­fect­ive any­more.”

But he saw re­li­ab­il­ity as far from an in­sur­mount­able prob­lem. “It’s the same thing we have with every tech­no­logy. We know we’ve been able to build much more com­plex ma­chines. A 777 has thou­sands of mov­ing parts, versus eight for a quad­copter. But the ques­tion is how quickly, for what level of money, for what re­li­ab­il­ity. These are the com­pet­ing factors.”

But won’t you get sued if one crashes?

His com­pany’s plan, too, is to start de­ploy­ing in places where the reg­u­lat­ory and lit­ig­a­tion risks are lower. “The ap­plic­a­tion changes your re­quire­ment of re­li­ab­il­ity,” Rap­to­poulos said. “The FAA may re­quire an­oth­er level than au­thor­it­ies in Haiti. If you lose a vehicle in Pa­lo Alto, you may be sued for mil­lions of dol­lars. If you lose a vehicle in Haiti, you may not be sued at all. “

But maybe, Rap­to­poulos con­tends, there are ways to in­teg­rate drones in­to the air­space that would present a lower risk to every­one. “Maybe there is a way to fly these things on routes where you don’t risk any­thing where you lose them. It will take that kind of in­nov­a­tion. [To us] it makes sense to start this first in rur­al places and maybe in the third world. Then once we fig­ure out how to do this at scale, we can bring it here.”

What are the spe­cif­ic things that can be done to in­crease re­li­ab­il­ity?

“There are a lot of oc­to­copters and a lot of quad­copters, but how do you design one that has the right re­dund­an­cies? Should the vehicle have a para­chute so when it has a cata­stroph­ic fail­ure, it doesn’t just fall out of the sky? If you have one fail­ure, can you dia­gnose and get it to a land­ing spot?”

Some of those prob­lems may be solved by in­creas­ing the soph­ist­ic­a­tion of the ana­lyt­ics they have on each drone. “How well can we pre­dict fail­ures? If we’ve flown 2.5 thou­sand hours and we have this kind of tele­met­ric data, I might know I should re­tire the vehicle.”

And each en­vir­on­ment brings its own chal­lenges.

“You have to worry about spe­cif­ic prob­lems in spe­cif­ic en­vir­on­ments. In Haiti, you have to worry about dust. If you want to work in San Fran­cisco, you have to have worry about GPS sig­nals be­ing lost be­cause of the ter­rain.”

As­sum­ing you can work out the tech­no­logy, why won’t people just shoot them out of the sky?

“They fly at 400 feet between 45-65 kilo­met­ers an hour and they are very small. At that height, you can barely see them. You can­not hear them. It’s like a tiny dot mov­ing in the sky. That’s the prac­tic­al as­pect of the ques­tion. It’s not go­ing to be a bunch of kids do­ing it for fun,” Rap­to­poulos said, rain­ing on every kid’s parade.

“The second point is that it’s il­leg­al,” he said. “The reas­on we’re not shoot­ing oth­er mov­ing things with guns is be­cause it’s not something that’s leg­al. It is more chal­len­ging to rely on that frame­work in a place like Haiti or Kenya or Mali. The risk there is high­er.”

But couldn’t the drones get taken out when they land?

“As you poin­ted out, the vul­ner­able part of the mis­sion is when they come down,” he re­spon­ded. “In our case, they do a ver­tic­al des­cent and then they go out again. And those loc­a­tions need to be pro­tec­ted.”

Think­ing about the de­vel­op­ing world con­texts where Mat­ter­net is work­ing, he con­tin­ued. “You need to have them owned by people who use the sys­tem, and then you tap in­to the so­cial dy­nam­ics. We’re not plan­ning to set up the net­works in loc­al places. We’re just provid­ing the tech­no­logy. So, they have to be owned by people in the de­vel­op­ing world that have the right so­cial status,” he said. “It would be people on the ground who un­der­stand how their loc­a­tion works. Those people are the ex­perts on the ground. They know how to read the coun­try and pro­tect their as­sets.”

What do you think of the reg­u­lat­ory hurdles in the U.S.?

“We’ve just had a pub­lic state­ment from a big com­pany they want it to hap­pen. Pub­lic ac­cept­ance goes hand in hand with reg­u­la­tion. There are many reas­ons that the pub­lic will see these as the wave of the fu­ture,” he said. “But we can­not [make that case] that if we can­not guar­an­tee to the pub­lic that this is a safe thing to be fly­ing over our heads and our chil­dren.”

So, let’s say you can fly a few drone de­liv­er­ies, does this ac­tu­ally work as a big busi­ness, the way Amazon seems to be ima­gin­ing?

“Scale is a chal­lenge in it­self. For Amazon to do this, they don’t get to do 10 or 100 de­liv­er­ies a day, they get to do thou­sands or hun­dreds of thou­sands of de­liv­er­ies a day. How you re­solve the scale is­sue is a ques­tion,” Rap­to­poulos said. “But we’re pretty good at solv­ing those chal­lenges as a tech­nic­al civil­iz­a­tion.”

What We're Following See More »
Gen. Kelly Rips Rep. Wilson for Criticism
9 minutes ago
Bush Slams Trump, Implicitly
23 minutes ago
Senate Rejects Effort to Nix SALT Tax Changes
1 hours ago

"Senate Democrats on Thursday failed in their first attempt to save the state and local tax deduction, which helps many residents of California and other high-cost states reduce their federal income tax bills. The Republican-controlled Senate voted 52-47 to reject an amendment that would have prevented the Senate from considering any bill that repeals or limits the deduction as part of a planned tax overhaul."

Trump to Hill Next Tuesday
2 hours ago
Longtime Progressive Members Pushed Out at DNC
3 hours ago

"A shake-up is underway at the Democratic National Committee as several key longtime officials have lost their posts, exposing a still-raw rift in the party and igniting anger among those in its progressive wing who see retaliation for their opposition to DNC Chairman Tom Perez. The ousters come ahead of the DNC's first meeting, in Las Vegas, Nevada, since Perez took over as chairman with a pledge this year to unite a party that had become badly divided during the brutal Bernie Sanders-Hillary Clinton 2016 primary race."


Welcome to National Journal!

You are currently accessing National Journal from IP access. Please login to access this feature. If you have any questions, please contact your Dedicated Advisor.