Will Self-Driving Cars Need Traffic Lights? Maybe Not, Says MIT

Imagine no traffic lights. It’s an idea so out-there even John Lennon couldn’t conjure it up in his heaven-less, country-less, possession-less, and religion-less fantasy land. And yet, according to an MIT study, “Revisiting street intersections using slot-based systems” (Tachet, R. et al., 2016), traffic lights could be made redundant if self-driving cars wind up dominating our streets. MIT News sums it up perfectly, as you might expect they would:

The study is based on mathematical modeling. The researchers examined a scenario in which high-tech vehicles use sensors to remain at a safe distance from each other as they move through a four-way intersection. By removing the waits caused by traffic lights, these so-called Slot-based Intersections (SIs) speed up traffic flow.

“An intersection is a difficult place, because you have two flows competing for the same piece of real estate,” says Carlo Ratti, a professor of the practice and director of the SENSEable City Lab in MIT’s Department of Urban Studies and Planning, and a co-author of the study. But a system with sophisticated technology and no traffic lights, he adds, “moves control from the [traffic] flow level to the vehicle level. Doing that, you can create a system that is much more efficient, because then you can make sure the vehicles get to the intersection exactly when they have a slot.”

As you can see from the nifty computer simulation below, sensor-equipped cars will be able to zip through intersections just by synchronizing with the rest of the traffic. The result is a freely flowing intersection which no car has to wait to go through.

This system is purely theoretical, of course, and the study’s authors acknowledge that there are some shortcomings, starting with the fact that this research only studies what happens at one isolated intersection. It remains to be seen how the slot-in system can work in a dense, urban area in which intersections appear very frequently.

The bigger, unacknowledged problem with SIs, as you may have noticed from the simulation, is that they make no allowances for non-car traffic.

Throw pedestrians, cyclists, and self-driving taxis (which will still presumably need to stop to pick up and drop off fares) into the mix, and the dynamics of the SI system may break down considerably.

The SI system is based on something called the “slower is faster” effect, which, according to Gershenson and Helberg (“When slower is faster,” 2015), “occurs when a system performs worse when its components try to bet better. Thus, a moderate individual efficiency actually leads to a better systemic performance.”

Think of a freeway on which one lane is closed for construction. With human-driven cars, what inevitably happens is that some cars will anticipate the lane closure and merge early, while others will seek to head-off the backup by staying in the merging lane until the last possible second. By speeding up to merge late, those driver are actually creating the backup they’re attempting to avoid. However, if everyone heeded the merge signs, slowed down slightly, and merged early, there would be no bottleneck and no massive backup to avoid.

The slower-is-faster effect is the traffic management version a maxim well known to students of John Stuart Mill and fans of Star Trek II: “The needs of the many outweigh the needs of the few.”

That phrase actually encapsulates the entire problem of traffic, at least the non-self-driving kind: how do you get people to understand that their selfish interests may actually be undermined by their selfish behavior? Come to think of it, that’s pretty much the problem with humanity in general.


Originally published at The Daily Transporter.