Self Driving Cars: The Future is Here(!?)

I spent last week in San Francisco as part of the M&T program’s Immersive Week. During this week, the entire sophomore class, along with the faculty of the program, meet with investors and founders in SF and the Bay Area, and also attend other activities, which are SF specific such as firm visits and factory tours.

This time we also had the opportunity to experience one more activity: riding in the Cruise robotaxi. 

This was not part of the week’s plan, but one of our students had access to the beta program and invited a colleague and me to join him for a ride. 

And I have to say (and this may sound very glib and provincial)...I have seen the future!

Before this ride, I was completely skeptical of whether self-driving cars could make a meaningful difference in human mobility, at least for the next few years, since the few experiences I had (e.g., with Tesla’s autopilot) were underwhelming. But now, I believe we are very, very close to commercially-viable, self-driving cars.

But allow me to elaborate. Since last year in SF, several autonomous vehicles have been allowed to drive between 10:30 pm and 6:00 am. The specific car we used was manufactured by Cruise, which was acquired by GM in 2016. The car is a Level 4 autonomous vehicle, and while fully automated, this means that a person still maintains the ability to take control remotely, and the car has a steering wheel. However, we didn’t have access to the steering wheel area, pretty much like riding a taxi where there is a barrier between you and the driver. 

The experience was nothing short of amazing! 

The car was waiting for us at a designated area (you can’t pick them up at any point in the city), we set our designated destination (again, not every destination is possible), and for the next 45 minutes, we drove through a route that was not the shortest but rather the one that maximized the self-driving system’s learning. 

We drove fast (without exceeding the speed limit) both through wide and very narrow streets, stopped at lights, yielded to other cars, dealt with impromptu road construction, and even a few people crossing the street. After the first few seconds, during which everything felt very new and odd, the ride felt extremely safe and smooth. When I sit at the back of a car I usually get motion sickness, but to my surprise, I didn’t feel like that for a second. 

Once the car dropped us off, we took an Uber back to our hotel, and during the 5-minute Uber ride, I felt much less safe than during the entire 45 minutes with the Cruise. 

Over-Hyped

Am I at the top of the hype cycle?

Not exactly. 

In 2015, self-driving cars were indeed at the top of the Gartner hype cycle:

But since then, there have been many disappointments and realizations that the future is not all that close and that the problem of allowing cars to drive fully autonomously is much more complex than it seems. 

By 2020, self-driving cars were at the bottom of the ‘Trough of Disillusionment.’

So, it’s clear that we are now in the phase where small and more incremental improvements are being made while getting us closer and closer to fully functioning Level 5 cars. 

I am not an expert in this area, so I’d like to refrain from talking about the feasibility and the timeline for a fully autonomous, Level 5 car, but I would like to talk about their operational implications. 

Operational Implications

As I have mentioned before, I am a huge fan of self-driving cars and the fact that they can completely transform last-mile logistics and logistics in general. The collaboration between Nuro and Dominos is just one example of the possibilities. 

It’s also clear that self-driving cars can have immense implications for human mobility. 

I personally hate driving. To alleviate the boredom from driving, I listen to podcasts, but I would gladly do many other things while driving if I didn't have to pay attention to the road. 

So clearly, there are going to be significant benefits in productivity if self-driving cars become the main form of transportation. 

But at what cost? 

Neda Mirzaeian is among those who try to understand the impact of autonomous vehicles on different aspects, from traffic to land use, over time traveled. In one of her papers, Neda and her co-authors look at the impact of autonomous vehicles on morning commute patterns. 

The model they study is one in where everyone uses a self-owned autonomous vehicle (AV). People can choose a time to leave home, as well as where they want to park: downtown or the outskirts of the city. One of the main advantages of AVs is that you don’t have to park near your workplace. The car can just drop you off and pick you up at a set time. In fact, some cities are already engaging with designing parking specifically for self-driving cars. In their model, Neda and her colleagues study the equilibrium that emerges given parking fees and congestion resulting from the choices people make. 

The model is theoretical but rich enough that the authors can calibrate its parameters using data from Pittsburgh. Their analysis shows that all AV commuters choose to park in the outskirts of the city, which increases both vehicle hours and vehicle miles traveled as compared to the case with all human-driven vehicles. Given the change, the authors recommend repurposing downtown parking spots to commercial and residential areas once autonomous vehicles are adopted at mass. 

One of the important findings of the model is the need for drop-off capacity for these autonomous vehicles. In particular, the authors show that converting downtown parking spaces to curbside drop-off spots for AVs can reduce the cost incurred by the people (congestion and change of schedule penalties) by up to 70%. 

There are also many other suggestions on how to improve the new equilibrium. For example, by adjusting parking fees and imposing congestion tolls until the infrastructure catches up with the new technology.

The point is that the technology is already here. Now it’s a matter of moving toward making it socially acceptable, technologically safe, and economical. So let’s use the next few years to understand how to build the appropriate infrastructure and economic incentives to ensure we don’t just end up creating more damage and confusion when the technology is fully adopted.