New data suggests that the days of using variable speed limits to manage traffic could be over
I think you probably know by now that I’m not a fan of the hype the floats around driverless cars. Don’t get me wrong, I love the tech behind them, and I am genuinely excited about the impact they could have on tomorrow’s cities. That’s why I wrote about research from MIT that showed that driverless cars might not need traffic lights. And why I highlighted work from Duke University that explored how driverless cars could best communicate with pedestrians.
But I am a natural-born skeptic. And as I pointed out in this article, many of the remaining questions this young industry faces have nothing to do with the tech at all. And in my book, I expressed concern that the benefits of driverless vehicles wouldn’t be felt until every car on the road is autonomous.
Well, on that point, I may just have been proven wrong!
Before we can understand why, we need to talk about why traffic jams form (and this bit features an excerpt from Science and the City)
Could the addition of just a few autonomous vehicles make us all better drivers? (AP Photo/Yorgos Karahalis)
We’ve all been stuck in a horrible city-center traffic jam, and found ourselves wondering how hard it can actually be to manage traffic. Well, it turns out, it’s very difficult indeed, and it requires a lot of mathematics.
Most of the time, traffic congestion occurs when vehicular density exceeds a critical threshold – in other words, when there are too many cars on not enough road. Or occasionally, there might be an accident or roadworks are to blame. But you’ve probably also witnessed a so-called ‘phantom traffic jam’ too, where for no discernible reason, traffic builds up and then eases. A number of years ago, a group of Japanese physicists rented a closed circular track to investigate what would happen to traffic flow in the absence of a bottleneck. In the now famous experiment, twenty-two volunteers in different cars were instructed to get up to 30kph (just under 20mph) and maintain that speed at a safe distance from the car in front. Very quickly, the system broke down, with some cars at a standstill while others sped up.
The reason given for this is surprisingly simple – drivers have trouble maintaining a constant speed. One person drives slightly too fast, so then puts a foot on the brake to correct for it. The person behind them then over-compensates for the sudden braking, and so on to the car behind. This effect continues to ripple back to other cars, growing all the time, until traffic eventually grinds to a halt. On a busy motorway, once one person brakes too hard, it can cause a start-stop ‘shockwave’ that travels backwards. From above, you’d see a road filled alternately with tightly packed cars and sections of busy-but-moving traffic. According to Tom Vanderbilt, author of the (excellent) book Traffic, ‘You’re not driving into a traffic jam, a traffic jam is driving into you.’
In the book, I interviewed Benjamin Seibold, Associate Professor of Mathematics at Temple University, about ‘jamology’ – the use of math to better understand traffic jams. Interestingly, he told me that these waves (which they call jamitons) occur even when everyone is driving perfectly. He said, ‘we’re inclined to blame individual drivers, but the models show that even if no-one does anything wrong, these waves can still arise.’
Benjamin and I kept in touch after our initial chat, and a couple of weeks ago, he sent me a copy of his latest paper – available for free here – written with colleagues at the University of Illinois, University Grenoble Alpes, University of Arizona, Yale, Penn State, and Rutgers University. In it, they’ve recreated the closed circular track experiment….but with one difference. They added an autonomous car to the mix, to see what (if any) impact it would have on traffic flow.
In the (sped-up) video above, 20 cars take to the track. And while there are humans sitting in all 20 driver’s seats, the black arrow highlights the one vehicle that can also drive autonomously. At the start of the experiment, all of the cars are in the hands of human drivers. Unsurprisingly, after 79 seconds, jamitons (traffic waves) begin to appear. After 126 seconds, autonomous control takes over the grey SUV, and its target speed is set to 6.5 m/s (23.4 km/h). If you look at the velocity profiles at the bottom of the video, this is represented by the red trace – and it shows that controlling just one car rapidly dampens the entire wave, though doesn’t remove it entirely. Increasing the speed of the autonomous car further to 7 m/s (at 222 sec), dampens the traffic wave even further. At a speed of 7.5 m/s (27 km/h), optimal dampening is achieved.
What this means in reality is that the presence of just one autonomous car can reduce congestion for all drivers – at least in the absence of any bottleneck. The researchers calculate that at 7.5 m/s, the standard deviation of the velocity (in effect, how different everyone’s speed is) is reduced by 80.8%, compared to when humans control all of the cars. This also means adding one autonomous car reduces excessive braking events from 8.58 per vehicle per km, to just 0.12/vehicle/km, and reduces fuel consumption by more than 40%. In addition, because the autonomous car increases the average velocity of traffic on the road, the throughput of the road also goes up by ~14%.**
The possible implications of this work are significant. The authors say that “(traffic) flow control will be possible via a few mobile actuators (less than 5%) long before a majority of vehicles have autonomous capabilities.” They also describe it as a “paradigm shift in traffic management” – and having read their paper, I’m inclined to agree. Though this set-up is really only indicative of a stretch of single-lane roadway, it could also potentially apply to multi-lane freeways – where lane-changing provides another trigger for producing jamitons. But, as Benjamin told me, it’s not all down to the autonomous car – the human drivers also play a crucial role in traffic flow dynamics, “The proper design of autonomous vehicles requires a profound understanding of the reaction of humans to them,” he said, “and traffic experiments play a crucial role in understanding this interplay of human and robotic agents.”
The team’s next target is to extend their experiment to denser traffic, on multiple lanes. I’m already looking forward to seeing their results!
** At higher speeds, it gets a bit more complicated. 347 seconds into the experiment, the autonomous car speed is set to 8 m/s, making it faster, on average, than the traffic flow. This induces a traffic wave, which brings with it increased fuel consumption. As the speed is reduced again, wave dampening reappears.
Leave a Reply