I believe the ultimate solution will be a combination of the two:

You have an automated road vehicle that acts as an "aircraft carrier" for drones. Something roughly the same size as a delivery truck today. It gets loaded up at the distribution center with packages and a few drones.

It picks a route that gets roughly near all of the deliveray locations. As it goes by, a drone is dispatched to take the package a couple of blocks to the person's doorstep and then returns to the "aircraft carrier".

When the drone gets back to the truck, it plugs in an recharges.

This uses drones for what they're best for: quickly and efficiently navigating small complex spaces like side streets and front yards when there is only a small payload you need to deliver. The truck addresses the shortcomings: limited flight time and payload capacity.

Actually drone delivery is much more simple than autonomous driving. Algorithm: grab package, fly up, fly directly to delivery point, drop package, all faster than driving. Compare to autonomous driving which needs to take into account road users (non-autonomous), traffic laws, visualization, etc and it's obvious why drones are more efficient and easier to program.

Assuming empty airspace, it's obvious why drones are easier to program. Assuming a single package it's obvious why drones are more efficient. Take away either of those assumptions and things get less obvious.

Those are very wrong assumptions. If a serious drone delivery program (or multiple competing ones) starts, airspace will suddenly become congested and you'll have to do heavy air traffic coordination to avoid mid-air collisions. And when you comapre with cars, you need to compare one delivery car with a hundred drones. Suddenly you see that drones are not efficient at all, which is kind of obvious, as it takes a lot of energy to keep stuff in the air with active propulsion.

I always pictured trucks driving from distribution centers, and deploying drones that cover an area. That way the drones only have to cover a small distance, reducing the number of drones that have to fly over a given area.

Mobile drone hubs seems like a very interesting solution, which could even be more energy-efficient than just trucks or just drones. Somehow I never thought of that, thanks!

Airspace is big. A central registry where drones announce their flight plans, allowing other drones to selfishly avoid them, would allow for plenty of room for drones to make sure they stay out of each other's way.

I'm skeptical on drones, and I think all this is an effort by Amazon to get in the news. But this doesn't seem like the problem.

Sort of, but as long as all drones (in this airspace) are automated, it's still an easier problem than self-driving cars. Just detecting where a road is is a feat, and then throw in other traffic, and then say that traffic is driven by people, who tend to behave irrationally by a computer's/programmer's typical definitions of "rational". Then there are traffic controls, construction, and the like.

Amazon's idea is to have airspace dedicated to autonomous drones, who can go from a to b in a straight line and mostly just have to worry about not hitting other drones (or buildings or birds or what have you). You can do that with some decent sensors and some simple physics simulations.

That's not to say that airspace is totally unregulated, but the regulations tend to be much broader (AFAIK; I'm not a pilot).

Collision avoidance is far easier in three dimensions than in two. Those extra degrees of freedom are crucial. (Of course, traffic capacity is much higher.)

It is only the severity of an accident that increases, not the likeliness.

Except that they don't need to do the R&D on the cars.

No traffic? No stop signs? No pedestrians to run in front of you? No closed roads due to construction?

There are dozens of reasons why an autonomous drone is greater than an autonomous car, use your imagination.

No traffic until people actually start using drone delivery systems. It's like with planes, the air seem so big but in fact it's very crowded and you get mid-air collisions all the time (hell, one case was just being discussed on HN earlier today). Scale that up to account for more drones in smaller space.

Apart from that, it still isn't obvious to me how this is better than an electric delivery truck. Cars are much more energy efficient than drones (especially helicopter-like ones), doubly so if you load the car to its capacity. Note that in case of drones, the mass of the package is only a small percentage of the overall mass of the vehicle, so most of the energy used will scale up with number of drone deliveries. I'm pretty sure taken altogether cars will win in terms of energy use.

I think the traffic Amazon would be worried about is human drivers, not other drones. AFAIK, Google's driverless cars have been in 14 documented accidents and all 14 were the fault of human drivers (including one incident of a human driving the Google car). Machines may be better drivers than humans.

Sure, but cars don't fall out of the sky when they get into accidents, spraying metal and death to people below.

Planes do not have the maneuverability of drones. Drones can hover in place. They can stop. They are small and light. Planes can't do this at all.

Depends on the type of drones you want to use and their range. Helicopter-like drones have very poor range and flight time, so this would limit your range considerably. And if you go to a plane-like drone, or even some kind of hybrid mode (eg. VTOL), suddenly you can't really do an emergency hover when cruising to destination.

But even if they find a way to handle the traffic, I was just saying that there will be traffic and that this poses serious problems.

> No traffic?

Not for long, apparently.

We'll have to wear helmets while walking on the street.

Or start hanging nets above roads and sidewalks to catch things dropping from the sky.

like airplanes? Might be a bit large.

At this stage, I think it may be easier to automate a drone than a car. Buildings and hills don't move around that much, and air traffic won't be as random as roadway hazards if implemented correctly.

Using delivery cars would cause a lot of traffic in densely populated areas. Cars are presumably also more expensive than drones. On the other hand, I suppose a car could do more than one delivery at a time whereas a drone would be more limited in what it could carry.

This summarizes a big problem with the congestion argument. Our economy is not big enough to cause "serious" delivery congestion when confined to 2-d road vehicles, yet its assumed if you open up to smaller 3-d capable vehicles massive congestion will result. Why?

Where I live it is not an issue, but it is theoretically possible that there might be a location on the planet right now where Fedex trucks crash into UPS trucks, and Dominos pizza cars are running Chinese delivery cars off the road, on a daily basis. However that weird situation existing would still be a workable system for perhaps 99.9% of the worlds surface.

There are also practical scaling issues. The local pizza delivery joint cannot produce fast enough to saturate a 2-D delivery system during the superbowl, just can't pump out pies fast enough. Naturally going to 3-D delivery with maybe 100x the delivery capacity because of 3-D isn't going to alter their delivery rate, the skies will be pretty empty. Even if you drive there to pick up, there will be a 2-hr wait for pizza. Drones and trucks are expensive and the finance guys would prefer the absolute minimum be owned and deliveries be made 24x7. That would imply we're not going to get drone congestion until we have deliveries at 3am. I'm just not seeing people order enough raw "stuff" to keep delivery services busy enough.

How much more traffic would there be if each delivery service had to drive back to their distribution point because they could only deliver one package at a time?

As others pointed out, combine truck and drone.

...because there's a secondary step in getting it from a car to the doorstep?

Will drones or self-driving trucks include Magic Doorbell-Tropic Technology[tm]?

How will they put deliveries into letterboxes? Or will there be a special Drone Nest Drop Off Utility Safebox[tm]?

And so on. The more you think about the idea, the less practical it gets.

I'm sure the problems can be solved. Obviously, you could retrofit your doorbell to include a drone code. Or make an app (it's always an app...) that alerts you when the drone is near.

This is perfectly fine, unless you don't want to change your doorbell and/or don't have the right kind of phone and/or don't want your stuff stolen.

There's a whiff of "Reinventing Yesterday's Technology for Tomorrow" about the idea. It looks a lot like a high school science show-and-tell trying to turn into a delivery system that can compete with 75 years of professional logistics expertise.

I don't usually do "not invented here", but there are so many very obvious things that can go wrong, and so may ways users need to change what they do, that I think debates about air traffic control are a long way down the list of show-stopper issues that need (literally) bullet-proof solutions.

"don't want your stuff stolen."

I would like to see a startup for what amounts to background checked, bonded, doorman service. Someone in my neighborhood I could trust to take deliveries, stuff like that. Outsourcing trust, I guess.

I could see drone delivery company "tipping" some old retired guy to solve their "last tenth of a mile" problem as he walks down the block to my house.

Alternately make all uber (or insert other) driver watch some training videos and tell them to be at xyz address at abc time and they verify safety and walk the last 50 feet and hand deliver, for a modest tip of course.

Presumably if I watched the training videos and didn't have badminton nets and kids pool and ham radio antennae in my backyard the drone could deliver directly to me and save me the cost.

There are a couple ways to address these issues:

For one, you could always have a human operator coordinate the landing and delivery process. They could coordinate delivery using a phone call.

Secondly, you could have a drone deliver goods to a distribution truck (possibly with a special landing-pad marker on top,) and a human could then deliver the package manually.

No advanced technology needed for either of these, it's just communication + drones.

Also, it seems that energy consumption per pound delivered must be much higher for flying than for rolling over paved roads.

Drone might be able to fly more directly, and with less waiting.

You could have said the same thing about horse drawn carriages in comparison to walking. Need to keep the horse healthy, need to keep the carriage working. Walking: need to be alive. At some point, we make sacrifices for convenience's sake.

I imagine it's because airspace (at least up to 400/500ft) can be relatively empty compared to road traffic in large cities which can be very congested (especially at peak times). So it could lead to reduced delivery times as well as potentially being cheaper than cars. This depends on exactly what the drones are capable of delivering (not your grand piano, of course).