The Fuse

Interview: What to Expect from MIT’s Slot-Based Intersections for Autonomous Vehicles

by Matt Piotrowski | April 13, 2016

Carlo Ratti is a professor at the Massachusetts Institute of Technology (MIT) and the head of MIT Senseable City Lab, a research group that studies how new technologies are changing urban environments. His latest project dealt with how autonomous vehicles will reshape traffic at intersections. “Slot-based intersections” would operate through coordinated computer control and would not need traffic lights or stop signs.

Ratti took the time to speak to The Fuse by phone. Below is a transcript of the interview.

Could you tell us about MIT Senseable City Lab, how long you have been there, and what your current projects are?

We started around 10 years ago and we helped with the development of the smart city concept. We’ve been working on how digital technologies are changing the way we understand, design, and ultimately live in cities.

Much of your work has involved how new technologies are changing the ways urban environments operate. Autonomous vehicles have huge potential to improve mobility and alter the way we live now. How do you see this new technology being integrated into today’s cities? 

The integration is not too difficult. Unlike 20th century technology, digital in most cases does not affect the physical structure of our cities. It’s about new layers of the built environment. What is happening with intelligent mobility in cities is one example of that. Technology is enabling new ways of moving—such as car-sharing, ride-sharing (such as Uber Pool) or autonomous driving. All of this is changing the mobility landscape and urban life, without the need to change the physical structure of the city.

Do you think that autonomous vehicles will be deployed first in cities that have done the groundwork and are prepared for it? Do you see this as a city-by-city deployment?

I don’t see this as a city-by-city deployment. The interesting thing about autonomous vehicles is that they can mingle with standard vehicles—you do not need special zones to initiate deployment. You need a basic infrastructure in terms of cartography, GIS (geographic information system), etc. But once you have the digital map of the city, so to say, you can deploy self-driving pretty much anywhere.

In your most recent research, you show how autonomous vehicles will reshape traffic at intersections, and that we won’t even need traffic lights or stop signs. What is the penetration level of connected and AV technology that would make this system worthwhile? 

I guess that the deployment of autonomous vehicles will start very soon, and spread uniformly throughout our cities. Slot-based intersections (as we call them) might take a bit longer because you need the entire system to have a certain level of intelligence. We might have some localized testbeds in small communities where you have autonomous vehicles alone. But otherwise, it will take longer before we will see widespread adoption.

Could you have these intersections without a 100 percent autonomous system?

You want to have control of how cars approach and cross the intersection. You need an intelligent system that will allocate slots—and you want to make sure that cars respect slots. You don’t need 100 percent autonomy, but you want to have enough intelligence to avoid collisions.

One of the biggest questions surrounding this study is about pedestrians and bikers. Those who are on foot or on bikes wouldn’t be able to travel through city streets in the way they do today. How would they be able to access city streets at a time when autonomous vehicles dominate? 

Today bicycles or pedestrians also cross an intersection when there is a slot—you have a green slot and you cross the street. It wouldn’t be any different than that, a slot-based intersection is compatible with pedestrians and bikes. Also, if you have a large, busy intersection you could separate flows on two different levels—as we do today. But in general, the system is totally compatible with all traffic flows.

What time savings does this system offer relative to traditional traffic controls whose cycles are optimized based on road conditions? 

It depends on what the traffic conditions are. You cannot just make a general point, we can say that transitioning could dramatically improve intersection performance, with traffic volume queues vanishing and travel delays cut to almost zero. Also, stop and go would be largely avoided, which would have the effect of reducing pollutants and greenhouse gases caused by acceleration and deceleration cycles.

How common is streamlined signal time, and how can we ensure that it is properly implemented? In order for this system to work, wouldn’t signals need to be streamlined throughout cities? 

The interesting point is that traffic intersections are particularly complex spaces, because you have two flows of traffic competing for the same piece of real estate. A slot-based system moves the focus from the traffic flow level to the vehicle level. Ultimately, it’s a much more efficient system, because vehicles will get to an intersection exactly when there is a slot available to them.

If you solve the intersection issue, you solve most of the problems at the network level. In other terms, after you deal with the intersection, scaling up is pretty easily. There will be no more green lights to synchronize—just using slots to optimize flows.

One major hindrance to the penetration of autonomous vehicles is the poor infrastructure we see today, with faded lane markers, damaged signs and lights, and uneven streets. How daunting is the task of rebuilding infrastructure in order to accommodate the new vehicle technology? 

When we are talking about signage and so on, the interesting thing today is that you can think about digital signage. Your car knows what the signage and the rules should be at a given point, without physical signage, and it would know exactly how to behave—even if there isn’t a physical manifestation in the form of a sign.

For a lot of people, this all seems very futuristic. How soon do you see widespread adoption of intersections cropping up the way they are presented in your research?

As we were saying, slot-based intersections will take some time to be fully deployed, because you need a certain level of intelligence in every car. What you can foresee, and what we’re seeing already under certain conditions, is small testing enclaves where there are just autonomous vehicles. Think about university or company campuses, and places like that. I believe that’s where we will start testing slot-based intersections very quickly.