We’ve been occasionally exploring examples of what could be the killer application for self-driving vehicles: autonomous freight deliveries, both long-haul and local, as well as some special use cases. Some, like UAV delivery of blood and medical supplies in Kenya, have taken off and are becoming both profitable and potentially life-saving. Others, like driverless long-haul trucking, made an initial splash but appear to have gone quiet since then. This is to be expected, as the marketplace picks winners and losers in a neverending quest to maximize return on investment. But the whole field seems to have gotten a bit sleepy lately, with no big news of note for quite a while.
That changed last week with Amazon’s announcement of Scout, their autonomous delivery vehicle. Announced first on Amazon’s blog and later picked up by the popular and tech press who repeated the Amazon material almost verbatim, Scout appears at first glance to be a serious attempt by Amazon to own the “last mile” of delivery – the local routes that are currently plied by the likes of UPS, FedEx, and various postal services. Or is it?
Too Much Deadheading
The Scout project is apparently well-supported at the executive level within Amazon, being headed up by a vice president, and is currently hiring a bunch of developers and engineers to work on it in the company’s Seattle research labs. They’ve even gone so far as to partner with Snohomish County, which encompasses communities north of Seattle and is home to many Amazonians, to test the autonomous delivery robot on its streets.
For as serious and glitzy as the effort appears, though, Scout seems a little underwhelming. The video accompanying its introduction is typical corporate fare, and reminds us very much of the announcement of Amazon’s Prime Air drone delivery service in terms of production values. In the Scout video, we see a rather unassuming six-wheel electric vehicle navigating suspiciously deserted sidewalks and streets on its way to a customer’s home. The vehicle itself looks like what everyone is describing it as – a cooler on wheels. Upon arrival in front of the destination, Scout parks on the sidewalk and waits for the recipient to approach, whereupon it pops its top to reveal the Amazon goodness inside. The vehicle presumably then returns to some sort of distribution facility to pick up the next package.
Without reading too much into the video, this scheme seems to have a few problems. Putting aside the difficulties all delivery services, whether human-guided or autonomous, have to deal with in terms of theft and vandalism of vehicles and their contents, the Scout model doesn’t seem to scale well. If, as seems apparent from the size of the thing, Scout can only carry a package or packages destined for a single address, it will spend about half of its time deadheading back to its depot. Deadheading is the bane of any delivery company because it does nothing but reposition equipment without turning a profit. Unless the geographic area covered is especially small and densely populated with potential customers, a model with that much deadheading is probably not going to be profitable.
Interestingly, Scout bears a strong resemblance to another autonomous delivery service, Starship Technologies, that has been under test in what seems like the perfect environment for such a service: college campuses. After four years of testing at various sites around the world and logging on the order of 100,000 km (62,000 miles) of autonomous operation, the company recently announced a rollout of 25 delivery vehicles to George Mason University in Maryland. There, hungry students and staff will be able to whistle up a food delivery using a smartphone app, for a charge of only $1.99. Unlike Scout, Starship’s vehicles operate completely autonomously – Amazon says that the six Scout test vehicles will be attended by human chaperones, at least for the time being.
It seems strange that as big a player as Amazon is so late to the last-mile delivery automation game. Starship Technologies has a half-decade head start on Amazon, and seems to have reduced to practice most of what Amazon still seems to be just toying with. Starship’s vehicles are said to cost only about $2,000, and at two bucks a pop for deliveries, it won’t take long before the things start making money for a company. Amazon seems only to be just dipping its toe in the water by comparison.
But I think the real problem with both of these services goes back to how poorly they scale. Both Starship and Amazon state that their services rely on local distribution hubs; given the 3 km (2 mi) range of the vehicles, that means a lot of hubs will have to be built to cover even a modest metropolitan area. Granted, the urban setting is about the only environment that has the population density for these to make sense economically, but still, all these services are doing is pushing the delivery problem down a level, since these hubs will probably not be able to be as big as a major distribution center and will have to be resupplied frequently themselves. And that will likely be done with trucks (with human drivers, at least for now) that can carry a lot of stuff cheaply and quickly.
Don’t get me wrong, I’m a big booster of autonomous freight. I still think it’s where we’ll see the first practical uses of self-driving vehicles, mainly because of the savings it represents. In a world where Walmart is offering its drivers $90,000 a year, it doesn’t take long before an investment in autonomous trucks to pay off. But the local delivery loop seems a harder nut to crack, and while Starship seems to have the right market in mind and the technology to capitalize on it, Amazon’s effort seems to be a non-starter, at least in its current form.