It is important to question the assertion that fully autonomous vehicles will be available in showrooms for household consumption by 2020. This seems to be implied in marketing language used by automakers and suppliers and Silicon Valley and is often passed on uncritically by journalists and commenters.

Fully autonomous, SAE Level 5 means that no driver is needed to operate the vehicle. The vehicle can be summoned empty, arrive quickly, and go where instructed without further human intervention. Such a vehicle can be sent off with a package, a child or a disabled passenger. It is generally understood that these vehicles would be safer, would spend less time parked, would be shared, electric, smaller, part of the internet of things and generally much better for everyone, our cities and the planet than are today’s motorcars. In other words, salvation from the 20^th century car.

The question is not whether such a vehicle can be built. The question is whether one can be built by 2020 for household consumption to operate effectively and safely on every road, street, driveway and parking lot that one can currently and reasonably use with one of today’s vehicles to commute, shop, visit, worship and play — i.e., Level 5.

Perhaps that is too demanding. So could a limited-range Level 5 be built by 2020 to operate as a public service within a constrained area or on planned routes? Could this be deployed as a commuter vehicle — a robotaxi or shuttle — to address daily commuters?

Yes. This has been demonstrated by CityMobil2 using robotic shuttles that operate along fixed routes using French and Dutch minibuses from EasyMile, Navya, Robosoft and 2GetThere. These EU robo-shuttles currently require pre-trained routes (from which the vehicle cannot stray), sensible weather and most importantly a remote screen-filled control room with a human monitor (“driver”) to ensure smooth operation. The CityMobil2 trial vehicles also had human stewards on board, a big red panic button and followed a wide pre-marked lane — like an over-sized bicycle-lane.

With fewer sensors and presumably far fewer lines of code than, say, the Google car, the technology and consumer-appetite gap between the CityMobile2 trial vehicles and an imagined 2020 Tesla Model 3 in full Level 5 kit is striking.

That’s a lot of gap to close by 2020.

Access anxiety

Level 5 vehicles have to meet with an infrastructure that will accept them. Without deploying isolated lanes or special roadside infrastructure —anathema to the Level 5 purist — we need to at least be sure that our infrastructure does not defeat these vehicles. But even making sure that lane markings are sufficiently readable for all permitted roads and circumstances (depending on methods used) would be a significant challenge.

For a showroom (er… household) vehicle to make sense, the target consumer would need to believe that her purchase could be used reliably in enough places, routes, and times to address a useful portion of her trip expectations. Car buyers exhibit a similar concern with EVs, which, while desirable in so many ways, suffer from range anxiety. Even if one asserts that EV range anxiety is, by now, irrational for most people, sales dramatically lag EV’s utility.

Fast-charging technologies and the Tesla Model 3’s range may well address range anxiety, but adding Level 5 autonomy, so that something like the Model 3 could be a self-arriving and self-stashing robotic car operable essentially or nearly everywhere reasonable, would produce access anxiety, and would do so until 2040 or 2050, or longer as expert Steven Shladover insists.

Until Level 5 can operate nearly everywhere, why would a household consumer move from SAE Level 3 to Level 5 and risk getting stranded or denied? The 2020 Level 3 would operate robotically at the majority of times and places, while its driver would get it to, or through, all the other places by simply taking over the driving task. This argument is supported in chapter 6 of Jurgen Nieuwenhuisjen’s fascinating thesis.

If one accepts that the Level 5 vehicle won’t experience showroom/household sales by 2020, then perhaps it would be ready for the public-service market by then. This is the robo-Uber solution Travis Kalanick proposed well over a year ago. Now Elon Musk agrees and Chris Urmson does as well—except using robo-vans for shuttle bus services.

Even when a Level 5 public-service vehicle is ready for these constrained Transit Leap applications, will cities and people be ready? The speed and agony with which many city councils struggled with Uber gives one pause to bet heavily on Level 5 for any measurable robo-taxi or robo-shuttle uptake by 2020.

Back to Level 3

Certainly, we can look forward to household SAE Level 3 AVs by 2020. And we are sensible to expect aggressive growth in this market at the expense of Level 0 to Level 2 vehicles (and perhaps some transit trips!). Level 3 sales mean revenue, which means competition, which means R+D, and which cycles back to a yet better Level 3. But better Level 3 is not Level 5. It is not robocab, and not the end of parking, driving or congestion. Coupled with driver distraction, Level 3 will be a mixed blessing.

The significant promise of full autonomy can only happen with Level 5, and the route to Level 5 through transit and constrained robofleet applications is far more understandable and livable than is the route through household Level 5s in showrooms by 2020.

* We could have used BMW, Google, Mercedes, Nissan, Volvo, or any of several other brands in the title, but recently Elon Musk has more of our fickle eyeballs because of Model 3.