Autonomous Vehicles: The Right Tool for the Job?

To people new to the game of golf, one club looks pretty much like another. They are all long and shiny with a grip on one end and a bulbous thing on the other.

People who play golf, though, know that each club has a very particular purpose. Some are designed to maximize distance, others accuracy. Some are designed to hit the ball with a lot of loft, others to hit shots that are low and flat. Some are for use in difficult terrain, others are useless anywhere other than the smoothest parts of the course.

Moreover, context matters. A club designed to hit the ball a long way is useless for a short hole but may be the most important club to use on a longer one. Individual golfers also each have their own strategies and playing styles, choosing to use one club in a situation where another player might use another. Playing a golf hole effectively, much less an entire course, requires complex thinking about which club to use in which situation.

The same is true in transportation. What looks on the surface to be one task – getting people from point A to point B – is actually far more complex, depending on the location of points A and B, the needs of the individual, and the context in which the trip takes place. You would no sooner putt your way up a golf fairway than you would walk from New York to San Francisco. But you would also no sooner use a driver on the putting green than you would fly a 747 to the corner store for a bottle of milk.

And using the wrong tool in the wrong situation can have major consequences. In the mid-20th century, we used the wrong tool – urban freeways – to move people in and around our cities. The scars we left have yet to fully heal.

This week, a group of tech executives in the Pacific Northwest issued a report calling for the region’s leaders to go full-speed ahead on autonomous vehicles (AVs) – including by allocating space on highways for the exclusive use of AVs.

In pushing their argument, the group compared the cost of additional highway capacity for AVs to the cost of building high-speed rail from Seattle to Vancouver, claiming that, for the cost of building high-speed rail, every resident of both cities could be given a Tesla, or the region could buy out Delta Airlines.

Here, though, is the problem: neither an “autopilot” Tesla nor Delta Airlines can yet do the specific thing that high-speed rail is designed to do, which is to move large numbers of people efficiently, quickly and with minimal use of space between the centers of two compact and vibrant cities.

To date, testing of AVs has suggested that they are capable of doing two things: 1) traveling on well-marked, limited-access highways, and 2) traveling at slow speeds in well-understood urban environments. AVs, in other words, can currently serve as a driver or a putter, but cannot – as yet – provide us with the equivalent of a full golf bag.

Eventually, we may get to full “Level 5” autonomy in which autonomous vehicles can be used for any and all trips on our streets and highways. But that is not going to happen overnight. And while the prospect of a single iPhone-like vehicle that can do everything might be appealing to tech guys, it is not necessarily the ideal way to organize a transportation system. As Jarrett Walker has repeatedly argued, individual autonomous vehicles, even if small, do not solve the problem of urban geography in the same way that high-capacity modes of transit – like subways, bus rapid transit and high-speed rail – do. If we value cities, and I think we should, we need to avoid the temptation to use the new, shiny club in our bag in places where it can do more harm than good.

The new U.S. Department of Transportation policy guidance on autonomous vehicles (PDF), issued yesterday, recognizes that context matters, and calls upon manufacturers to document how their vehicles will function in various “operational design domains” – city streets, highways, etc. Even within these broad categories, however, conditions and needs can vary greatly, as Sarah Jo Peterson illustrated in this description of her specific operational design domain … er, neighborhood.

So, whether you like high-speed rail or not, or think it is a good investment or not, to posit autonomous vehicles as a substitute for it without recognizing that they are not currently capable of solving, nor particularly well-suited to solve, the specific challenge rail is intended to address is to miss the point.

One other thing: while tech companies and automakers may dream of a single vehicle that does it all, I think the promise of autonomous vehicles – especially if married to shared mobility networks – is different. Autonomy and sharing can enable the proliferation of a wider variety of high-quality transportation tools, each of them attuned more closely to the needs of a particular individual making a particular trip. From almost any perspective, our current reliance on privately-owned vehicles results in an absurd amount of waste – for example, using a two-ton SUV to move a 16-inch pizza a mile in congested traffic. Not only can shared networks enable people to choose the right tool for the job, but if we integrate sharing and autonomy into our cities intelligently, we can allow for the safe use of additional modes – like walking and bicycling – that are currently crowded off our streets by our over-reliance on the 20th century’s intended one-size-fits-all solution to our transportation challenges: the personal car.

Understanding what it is that our transportation system investments and policies are trying to achieve is essential to making sure we invest in the right tools for the job. If the state of the current debate over autonomous vehicles is any judge, we have a long way to go.