Self-driving vehicles represent one of the most fascinating fields of technology development today. More than just a convenience, there is the potential to radically alter how we work and live, and to alter the essential layout of cities. Their development involves a coalition of many disciplines, a marriage of the auto industry’s epicenters with Silicon Valley, and is attracting hundreds of billions of dollars in annual investment, globally. Exciting as all this is, I think the viability a full-fledged self-driving car in the day-to-day real world is further off than many believe.
My ‘dose of reality’ is driven by two main arguments. First, I think that we’ve underestimated the technology involved in making true autonomous driving safe and reliable, especially in cities. Second, there are some significant infrastructure investments that are required to make our streets ‘self-driving ready’. This requires significant public sector attention and investment which, except for a select few places, is not happening yet.
Self-driving cars have already logged lots of miles and have amassed an impressive safety record, with a couple of notable and unfortunate exceptions. But most of this has occurred on test tracks, along the wide-open highways of the desert southwest, and in Truman Show-esque residential neighborhoods.
For a sense of the real world that the self-driving car has to conquer, I encourage you to visit my home town of Boston. Like most of the world, it’s not Phoenix or Singapore. It has narrow streets, an un- grid-like/not intuitive layout, and a climate where about half of the year’s days feature wet, snowy, or icy roads. Sightlines are bad, lane lines are faded, and pedestrians and cyclists compete for a limited amount of real-estate. In other words, a fairly typical older city. How would a self-driving car do here?
To get even more micro, I’ll take you to a very specific type of intersection that has all of the characteristics deigned to trump a self-driving vehicle. In this situation, the car has to make a left turn from a faded turn lane with no traffic light, and then cross a trolley track, two lanes of oncoming traffic, and a busy crosswalk. So we have poor lane markings, terrible sight lines, and pedestrians moving at an uneven pace and doing unpredictable things, before even getting into the wild cards of weather, glare, and so on. My heart rate quickens every single time I have to make this turn. I would want to see a car successfully self-perform this left turn a Gladwellian 10,000 times before I’d sign that waiver.
I’m sure each of you can provide countless examples of situations that would prompt the question of “can a self-driving car really handle that”? It shows the complexity and the sheer number of variables involved in pulling this off. Think of all the minor decisions and adjustments you make when driving, particularly in a congested area. Rapid advancements in AI will help.
This is not to diminish the mammoth progress that has been made on the road to the self-driving vehicle. The technology is getting there for self-driving cars to be viable in many situations and contexts, within the next five years. It’s that last 10-20% of spots and situations that will prove particularly vexing.
If we believe the self-driving car could be a game-changer over the next 20 years, I think we need to be doing a lot more thinking about the infrastructure required to support its development. We all get excited about how the potential benefits self-driving/autonomous vehicles will usher in, such as changes to the entire model of car ownership, less congested roads, the disappearance of parking lots, etc. This exciting vision assumes a world where the self-driving car is already mainstream. But I think it’s a bit naïve with regard to the type of investment that is needed to help make this happen. This is going to require huge public sector involvement and dollars in many fields and categories. As examples: improvements to roads to accommodate self-driving cars (lane markings, etc.); deployment of sensors and all sorts of ‘smart city infrastructure’; a better ‘visual’ infrastructure; a new regulatory apparatus, and so on. And of course, we will need advanced mobile broadband networks, combination of 5G with vehicle-centric capabilities envisioned by evolving standards such as V2X, to help make this happen.
This will be a really exciting field, with all sorts of job opportunities. There’s the possibility of a game-changing evolution of our physical infrastructure not seen in 100+ years. But worldwide transportation budgets are still mainly business-as-usual, with sporadic hot pockets of cities hoping to be at the bleeding edge.
Getting to this next phase of the self-driving car will require a combination of pragmatism, technology development, meaningful infrastructure investment, and a unique model of public-private cooperation.