Self-driving cars experiment demonstrates dramatic improvements in traffic flow

[@Garbosack]

In fact, this can be predicted by (or validates) an amazingly simple, yet accurate (as proven by the experiment) mathematical model. Roughly speaking, we are in the case of an infinite, 1-lane road (legitimate, as crossroads and multi-lanes can only decrease flow). There, the flow F is given as function of velocity as follows: F(v)= v / (L + t*v + K* v^2). Here: - L is the average length of a vehicle, - t is the average human reaction time (time between seeing the front vehicle braking and the driver effectively braking), so that t*v is the reaction length, the length over which the driver is reacting, - K is a factor depending on the vehicle - more precisely, proportional to the vehicle's weight - so that K*v^2 corresponds to the security length, i.e. the length needed to dissipate the kinetic energy when passing from a higher velocity to a smaller one. The function F has then a unique maximum, which is exactly for V=sqrt(L/K). An implementation for L=4 and K=0.0635 (typically, this data is given by the vehicle's manufacturer) gives that the optimal velocity is v=7.94 m/s, so quite close to the experimental value obtained here !

[@Burner.Account..]

It's much more complicated, the problem is not in the physical car, it's the human driving it. Over the past 60 years, lots of mathematical models were developed to model human driving behavior. I'm working on a sim of this exact experiment atm, and trust me, the maths looks simple, and then when you put 10 or 20 of modeled "cars" together you see the greatness of "string stability" (or the lack of). Spent lots of nights screaming at a computer like a maniac. Would suggest reading this paper for more details, then look up "Intelligent Driving Model" by Martin Treiber, his sims are almost porn-worthy. www.sciencedirect.com/science/article/pii/S136984780000005X

[@Magicsugus37]

Pértinent ! Merci

[@nananismsm]

Thank you very much. Traffic jams ends worldwide.

[@deandreatistan9564]

Extrêmement bien expliqué et véridicte

[@benjaminhelfer811]

D'accord faisons comme ça

[@SeanLumly]

That seems similar to what I do on the highway! Rather than stop-and-go, I try to move forward at a constant rate to meet the overall rate of movement. In the end, the people behind me enjoy slow, but steady, travel vs. stop and go that is characteristic of jams. I have been doing this for over 20 years! It is so wonderful to have a intuitive supposition validated by actual data. Of course, with lane changes, this becomes far less sustainable. But with more than one steady lead vehicle, I would wager the flow of traffic would improve markedly even on multi-lane roads.

[@SeanLumly]

Wow! That ~8 m/s has been discovered as a performance threshold is very, very interesting. I would love to see the curve that represents 'smoothness' with density, and other variables. But this experiment is also powerful because it hints at something that can be done today to ease traffic congestion: deploying a set of pacer vehicles along a busy motorway to reduce stop-and-go traffic under many common scenarios. Anyway, really wonderful research!

[@jefersonvilaede]

Indeed, it IS wonderful to have an intuitive supposition validated by actual factual data. But also, one can learn that we just need to trust ourselves more, our intuition.

[@dw734]

I think the cause of the stop and go traffic is from people trying to tailgate the person in front of them so people in other lanes don't attempt to cut into their lane and get in front of them.

[@TheGamingAlong]

I did this and the car from the other lane fills up the gap lol

[@tedmonson2000]

Yes, increasing following distance tends to lead to other vehicles switching into that gap, but if you increase your following distance further - you then have enough of a buffer that you can have a more gentle slow down even when a few cars switch into the lane. Some of those doing so will legitimately need to be able to get over to other lanes, others are just trying to speed up for a few feet to ride the next persons tailpipe. I follow the same principles as Sean when in traffic - it works, and it smooths out the flow of traffic behind me significantly.

[@grantorino2325]

*BEEEEEEEEEEEEEEEEEEEEEEEEPPPPPPPPPPPPPPP!!!!!!!!!!!!!!!!!* Paging Mr. Grey, paging Mr. C.G.P. Grey. White, courtesy phone, please.

[@brainfloss9710]

Yeah, this works if all the human drivers are just bumper following dingbats. But what happens when you throw one of the commenter's behind the wheel of one of those? How would they compare with the AI? How would the results change? Just thinking out loud here.

[@illygah]

Cool1!

[@mechanoidwarhead5530]

lol, University of Arizona. When is the thesis on the autonomous car's gender being released?