Okay, look. I know that the idea of having drones deliver stuff sounds like the future. And there are a lot of companies making promises about it, including, this weekend, Amazon. "We'll be ready to enter commercial operations as soon as the necessary regulations are in place [in 2015]," Amazon says. The reason that Amazon can say this, and that it sounds so exciting, is that it is technically possible to put together a demo like the one they've just unveiled, where you have a drone autonomously transport an object from one place to another.
However, once you get past the wow factor and start thinking critically about what it would take to make something like this work in practice, it starts to look a lot less realistic very, very quickly.
Here's the video, just so you know what I'm getting all worked up about:
Amazon, for the record, does not say that their technology is ready to be implemented as-is. They're going to have everything ready by 2015, when new airspace rules from the U.S. Federal Aviation Administration will allow for some sort of commercial drone operations. That's really not that long from now: it's going to be 2014 in a month. Notably, Jeff Bezos himself seems to be a little more pessimistic than the Amazon Prime Air page is, as he expressed to Charlie Rose on 60 Minutes last night:
“I know this looks like science-fiction,” Bezos tells a jaw-dropped Rose. “It’s not. It’s early; this is still years away . . . we can do half-hour delivery, and we can carry objects—we think—up to five pounds, which covers 86 percent of the items that we deliver.”
Bezos says the current generation of test drones have a 10-mile radius from a fulfillment center. Given the growing number of Amazon warehouses out there, that would cover quite a substantial portion of some major metro areas.
“I know it can’t be before 2015 because that’s the earliest that we could get the rules from the FAA,” Bezos says. “My guess is that’s pretty a little optimistic. But could it be four to five years? I think so. It will work and it will happen, and it’s gonna be a lot of fun.”
That sounds great, but here are just a few of the existing issues that an idea like this is going to have to robustly solve before urban drone delivery on a commercial scale will work:
Navigation: Drones using GPS (differential or otherwise) have no trouble navigating around open areas or landing in fields. We're probably dealing with an accuracy of somewhere between 3 and 10 meters, which is fine for most applications. But it's not fine for landing on the front steps of a house as shown in the video, especially when there are obstacles (like trees) all over the place. Furthermore, how does the drone know where to go? You can't just give it an address; it needs precise GPS coordinates. Aerial maps usually aren't updated frequently enough to show obstacles, and if you leave it in the hands of the consumer (through, say, a GPS-enabled app), you're going to get a bunch of people giving your drone the coordinates of their homes with 50 meters of error.
Obstacle avoidance: Even if you get all of the navigation bugs worked out, the only way it's going to make sense to make deliveries like this is if you're in a high-density urban area. That implies lots of buildings and power lines and telephone poles and pigeons and trees and stuff all over the place, and your drone is going to have to safely avoid all of them by itself. This is certainly possible, but doing it dynamically (in real-time on board the drone) is going to require a lot of computing power and some relatively sophisticated sensors, like (at a bare minimum) a camera that's high-resolution enough to pick out black power lines against black pavement. And then once the drone gets to where it's going, it's going to have to locate a safe place to land and drop off its box, which brings us to safety issues.
Safety and liability: These drones are not toys. In order to lift a 5 pound payload for 20 minutes with redundancy, you need big motors with big props. Even small motors with small props hurt (try sticking your finger into the prop of an AR Drone while it's hovering), and big motors with big props can likely cause significant injury. The amount of interaction between the drones and humans is intended to be minimal, but these things are autonomously delivering packages to ground level while fully powered. Adult humans probably (I should stress that "probably") have the sense to stay away, but what about kids? What about pets? And this doesn't even get into what happens if one of these things actually crashes for whatever reason, which they almost certainly will from time to time, if for no other reason than weather is unpredictable.
Legality: It's not likely that the FAA is going to suddenly open up airspace into a wild unmanned free-for-all in 2015. What's more likely is that new rules will allow unmanned aircraft in public airspace, which is generally 500 or 1,000 feet away from all obstacles (including the ground). In keeping with a 1946 Supreme Court Ruling (read more about this here), operating aircraft (or anything else) too low over private property constitutes a violation of the property owner's rights. What this means is that it's probably not going to be legal to fly unmanned drones over private property below 500 feet, which makes it a lot more difficult to make deliveries.
Cost: At the end of the day, the question is really whether it's going to be worth it to (attempt to) implement a system like this for anything but the novelty factor. With companies like Google and its partners already offering same-day delivery (within a three hour window) for all sorts of things of any size and weight in some urban areas, is there really that much value to attempting to use (by definition) expensive and problematic robots to narrow that three hour window to 30 minutes?
I'm probably coming off as some sort of curmudgeon here. Honestly, I want more than anyone for this to be the future, I really do. And most, if not all, of these problems are certainly solvable, and will be solved at some point, especially in non-urban contexts like agriculture. However, solving them in a year or two (or even five) looks exceptionally difficult, even with resources like Amazon can bring to bear. Getting something like this to work autonomously even in a carefully controlled environment would be quite a challenge, to say nothing of an actual urban environment.
The point is, I just don't see urban drone delivery happening in the next few years, and I feel like when companies (especially big companies like Amazon) make claims like these, it ends up creating unrealistic expectations for robotics as a whole. We don't want robotics to turn into the next flying car, where the promised fantasy encouraged by some cannot possibly match the reality that many others are trying to achieve. Again, I'd love to be proved wrong about all of this, but until I am, I feel like it's important to keep on taking a critical look at ideas like these to make sure that they stay grounded in reality. What do you think?
[ Amazon ]
Evan Ackerman is a senior editor at IEEE Spectrum. Since 2007, he has written over 6,000 articles on robotics and technology. He has a degree in Martian geology and is excellent at playing bagpipes.