It might be possible one day to hail a driverless taxi in Charlotte. Automated cab company Cruise tested its vehicles in the city last week. It’s already launched services in San Francisco, Phoenix and Austin. And more automakers are testing self-driving cars that would be sold directly to consumers.
Advocates of the new technology point to improved safety and quicker trips. But a new study finds automated vehicles can actually slow down traffic. Ali Hajbabaie is the author of that study and an associate professor in the Department of Civil, Construction, and Environmental Engineering at NC State. He joins us now.
Marshall Terry: We'll get into your research and what you found in just a moment. But first, can you explain generally how these self-driving cars work?
Ali Hajbabaie: There are a few components in self-driving cars that essentially allow them to drive on our streets.
One thing is perception. So, these cars are equipped with sensors so that they can see and perceive the environment — other cars, lane marking, traffic control signs, pedestrians and so forth.
And then there are algorithms inside the cars that are going to help the car come up with passes to go through the road while they are seeing the car, so that algorithm is gonna come up with several trajectories, that's what we call them, and the car is going to select the best [trajectory] to proceed through an intersection, through an arterial street and other facilities.
Terry: Now getting into your research, you mentioned vehicles that are connected and those that are not connected. What's the distinction there, and what effect do they have on traffic?
Hajbabaie: So, connectivity is technology that allows cars to communicate with other cars, with traffic control devices and also with other system users, such as pedestrians.
Now, if a car is able to communicate with the computer that controls the traffic light, the car can have a better idea about what traffic lights are going to be shown up in the near future and based on that, the car can plan its arrival to the intersection so that it does not have to stop or it can go through the intersection with minor disruptions to the speed. And as a result of that, the travel time could be reduced [and] the number of stops could be reduced, so the efficiency is going to go up and at the same time, the emissions are going to be reduced.
So, connected vehicles, we have found that they can increase the capacity at signalized intersections so that we can have smoother travel through the intersections. Now, if we have a vehicle that is autonomous but not connected, the story is going to be different.
Terry: And why is that?
Hajbabaie: Well, autonomous vehicles now are programmed to be more cautious than human drivers so that we can increase safety or at least maintain the same level of safety.
So, the effect of that is that when they get to an intersection, they are going to reduce their speed a little bit sooner than us. When they see a green light, they have to be more cautious in speeding up. They have to maintain larger gaps with the vehicle in front of them.
As a result of that, what we have seen in our computer-based simulations is that they reduce the capacity and what that does is that we are going to see longer lines of vehicle at signalized intersections. We're going to have longer travel time and that's something that we hope we don't see, but it's happening as of now.
Terry: Now one interesting finding in your paper was that the biggest reduction in road capacity and increase in traffic actually happened with a fleet of 100% autonomous vehicles. That seems counterintuitive.
Hajbabaie: Yeah. If all of the vehicles — 100% of the fleet, only autonomous, do not talk to each other — you have a fleet of cars that are super cautious. And as a result of that, we are going to have up to 20% reduction in the capacity. Now, that is the way that we have programmed autonomous cars as of now and as the technology matures, this programming can change and they may end up maintaining shorter distances between them and as a result of that, the capacity can go up.
But now, if you have a 100% of connected and autonomous vehicles, these are cars that are driven by computers and they can communicate with each other and with traffic light controllers, then what we have observed is that the capacity could be increased by 80%, which is a huge increase in the capacity. It will reduce our travel time significantly and also the emissions at intersections.
Terry: Now even if self-driving cars don't always make streets faster, could they still make them safer?
Hajbabaie: Well, that's the objective — and if you compare the safety to human-driven vehicles, the technology suggests that they would be able to make our streets safer.
Terry: Just because they're more cautious than we are?
Hajbabaie: They are more cautious than we are. They can do thousands and thousands of calculations in finding the best way to go forward and they are much faster than us in reacting to something that is going to happen.
As a driver, for example, we are trained not to be that much worried about the car behind us, but an autonomous vehicle can observe the cars also that are approaching from [the] back. As a result of more observability and also being able to react much faster than us, they hopefully would improve safety in our streets.
Terry: So what’s the takeaway from all of this for consumers and also for city leaders in places, like Charlotte, where this technology could soon roll out?
Hajbabaie: I think one main takeaway is the effects of connectivity on traffic operations here.
When we have connected vehicles, we have already observed that the capacity is improving, travel time is reducing, emissions are reducing — and as a result of that, cities [and] state [transportation departments], if they provide hardware to signalize intersections so that they can communicate with cars that are equipped with wireless systems and share information, I think that would significantly help our traffic in signalized intersections and perhaps in other highway facilities as well.
