Robotaxis Will Be The Last Automated Vehicles To Hit The Road

  • Nick Jaynes has worked for more than a decade in automotive media industry. In that time, he's done it all—from public relations for Chevrolet to new-car reviews for Mashable. Nick now lives in Portland, Oregon and spends his weekends traversing off-road trails in his 100 Series Toyota Land Cruiser.

Share on facebook
Share on twitter
Share on reddit

Despite the splashy headlines Elon Musk generated when he predicted that Tesla would have over a million robotaxis on the road in 2020, the reality is that fully automated cars moving freely at full speed on the roadways is a long way off.

In fact, by 2025 it would be surprising if there were more than 2,000 Level 4 autonomous robotaxis with absolutely no driver onboard, on the roads, around the globe, let alone a million.

That’s easy to say, though. So, I wanted to take the time and go through exactly the reason why the rise of computer-driven cars will be slower than some folks like Mr. Musk would want you to believe.

Rise Of The Machines

The first automated driving vehicles to really become prevalent will not be personally owned self-driving cars. They will be slow-moving shuttles, like those at airports, corporate campuses and universities. Although the majority of Americans are currently afraid of self-driving cars, fear-factor has nothing to do with the rise of automated shuttles.

Sure, a self-driving box rolling around an airport will be a lot less intimidating than a steering wheel-less sedan doing 70 mph on the highway. And, yes, these shuttles will be a great low-stress barrier to entry to automated driving for the public. The real reason they’ll be the first self-driving vehicles is ease of implementation.

There are very few variables engineers have to consider with a low-speed shuttle. They essentially operate like a trackless train; they stick to the same route over and over. Heavily mapping and geofencing a small footprint route is infinitely easier than mapping Interstate 5 to the point that it’s safe to let self-driving cars loose on it.

Volkswagen tests highly-automated driving in Hamburg. | Photo: Volkswagen

 

Plus, replacing human drivers with a robot driver, in the case of a shuttle, is cost effective. This shuttle can run 24 hours a day. It doesn’t need a lunch break. It doesn’t take sick days or vacations. It doesn’t need health insurance or workers’ comp. In this way, an automated shuttle makes a lot of sense. There is a lot less of a business case for replacing your driving duties with a robot — at least not in the short term.

Next will be limited implementation of self-driving, over-the-road trucks, like those being developed by Daimler Trucks. This is another place where there is a business case for getting humans out of the driver seats. Again, for the reasons I just cited about replacing shuttle driver.

Third will be luxury vehicles. These will likely not be Level 4 automated driving vehicles, even by 2025. Luxury will be the car market where Level 3 automated driving rises from.

Let’s not forget that even components that compose a Level 3 automated driving system — including radar, LIDAR, laser scanners, digital cameras, hugely powerful onboard computers, and redundant safety systems — are incredibly expensive. So, in order to even recoup even some of the investment, automakers won’t be able to fit them to any car with a sticker price below $200,000 for a while.

These ultra-luxury cars will be able to handle all freeway driving duties from on ramp to off ramp. In the case of a problem, they’ll need to request the driver take over. If they can’t or the driver is unresponsive, the car will bring itself to a stop in its lane, activate the emergency flashers, and alert authorities. Even that sort of automated driving implementation will be a huge accomplishment, both from a technological but also a liability standpoint.

The last segment of the mobility market that automation will reach is actually the one that has been garnering the most attention lately: Ride-sharing or ride-hailing robotaxis.

The problem with programming robotaxis

Like I mentioned in the intro, as I see it, there probably won’t be more than 2,000 Level 4 autonomous robotaxis in the world on public roads by 2025. And those will be relegated to a few major cities. Three factors will hold back the widespread rollout of robotaxis: technical complexity, cost (in relation to return on investment), and regulation.

I am going to save regulation for another post. So let’s look at technical complexity and cost.

We know that self-driving tech is very expensive. The virtual laundry list of redundant sensors I listed above is just for cars designed to drive autonomously on high-definition-mapped freeways. This, while incredibly challenging, is nothing compared to urban driving. Add the complexity of dealing and communicating with pedestrians and other drivers, which autonomous city cars will need to do (through lights or otherwise), into the automated driving equation and the prospect of designing a car that can achieve all of this safely gets mind-bogglingly complicated … and expensive.

Volvo’s Autonomous Concept. | Photo: Volvo

 

It can be done, yes. Engineers will get to the point of making a car that can communicate with other vehicles, city infrastructure like traffic lights, and also communicate with pedestrians with things like external indicators (more intricate than turn signals). But what’s the benefit to a company like Uber to invest in these? Let’s forget the technical complexity of designing a self-driving car that can handle, say, a Beijing or New York City from a cost consideration.

Right now, with companies like Uber, the drivers own their cars — not the company. Uber is just the middleman. Stepping in and buying and running these cars would be a huge financial (and liability) undertaking. For what? The profits? At current Uber rates, these cars would have to operate 24 hours a day seven days a week for decades for the company to break even on the investment of designing and building them.

Then, riders will be subject to not-so-smooth, overly cautious robo-drivers that probably get routinely messed with by jerky pedestrians keen to disrupt a robotaxi. Think I’m being pedantic? Remember HitchBOT, the hitchhiking robot? It got beheaded in Philadelphia. So, yeah, I think people are going to take pleasure in intentionally getting in the way of robotaxis.

Safety first

This all said, I’m not down on automated driving ridesharing or ride-hailing robotaxis. I do really believe that in the long run automated driving technology will lower roadway death rates and improve efficiency thereby lowering the carbon footprint of transportation. That is, if automated driving is implemented safely.

We need to be very skeptical when people like Elon Musk make claims about unleashing a huge fleet of robotaxis on the road. That’s because the barriers I discuss above are those facing companies concerned with public safety and taking on undue liability. I totally believe that Tesla could enable a great number of its cars to drive themselves next year. That doesn’t mean they’ll be safe. That, however, is a discussion best saved for another post.


About the Author

  • Nick Jaynes has worked for more than a decade in automotive media industry. In that time, he's done it all—from public relations for Chevrolet to new-car reviews for Mashable. Nick now lives in Portland, Oregon and spends his weekends traversing off-road trails in his 100 Series Toyota Land Cruiser.

Close Menu