Hacker Newsnew | past | comments | ask | show | jobs | submitlogin
The physics of traffic (here.com)
36 points by petrel on Jan 20, 2016 | hide | past | favorite | 8 comments


I absolutely love driving and think about it a lot. When I was in high school we watched some video about the future highlighting self piloting cars and how they could drive bumper to bumper. That doesn't make any sense though, you need to leave space for lateral movements.

The problem as I see it is that most people think about driving as competitive. They don't want you to get in front of them. This is odd, because someone who is passing one person will often carry on passing and be out of sight in a minute.

This is where the idea of driving collaboratively comes in. If you let someone merge, make their exit, pass you, you're enabling flow.

Instead, people drive competitively. They get as close as possible to someone's bumper during a merge to ensure that only one (or potentially zero) cars get in front of them. This forces people to stop or slow considerably before someone can get in front of them. The same with passing, if someone is blocking the fast lane, the passing car needs to slow considerably (affecting cars behind them) and change lanes (affecting cars in another lane). Everyone thinks it's about them, and some will even speed up when you try to pass as if you are insulting the speed that they are driving. It's madness.

If I ever have a ton of money I'm going to take out a national ad campaign encouraging people to drive collaboratively.


"If you let someone merge, make their exit, pass you, you're enabling flow."

yes and no.

If you let someone merge in heavy traffic, and (s)he has to break hard right after merging, the flow stops. This is what happens - indeed - in competitive driving. In my city people seem to be so full of adrenaline, more before than after work, that it happens all the time and already in the morning at 6:30.



Certainly an interesting topic and one which if exploited could improve travel for many. My question, though is whether it's possible to improve the situation without the recourse to technology? i.e. if you looked ahead on the road and saw a small jam (but one at a distance which was say 400 meters ahead), but which clears quickly, if a certain way of modulating your speed would help? Or is the only way to recognise it late, and brake as hard as possible (as I see so often!)?


What you're speaking of if pretty much analogous to driving at speed toward a light that just turned red the better part of a mile in the distance because you know (from experience, e.g. your commute) that in the time it takes to reach the light the light will change to green.

You're just making the location variable as well. This is a very common way of doing things for anyone that drives something that has horrible acceleration and a good view of traffic ahead (a typical forward control box truck full of stuff). When a temporary obstruction is noted up ahead, such as someone slowing for a right turn into a parking lot or it's easier to change lanes to go around the obstruction than it is to slow and speed up again. When changing lanes won't work, such as a bunch of cars that have yet to come up to speed after moving on a green light then slowing down slowly so that they're caught up to gradually is a viable alternative. Most people underestimate the amount they need to slow and wind up tapping the brakes at the end so fears of this causing unnecessary slowdowns are unwarranted IMO

Similarly, when encountering something like merging traffic on the right it's easier to maintain speed and move left to give the clusterfawk caused by the merging traffic space to run its course.

When driving a loaded truck the cost to slowing down is very high, these things have 0-60 times in the teens when unloaded and when traveling in traffic this is a pretty good incentive to avoid slowing down. The good view of traffic ahead allows drivers to adopt a strategy for driving at a consistent speed.


Isn't a risk of encouraging 'pre slowdown' that you might spread the slowdown to conjoining travel axes via lessened throughput per time clogging up intersections or other interface points?

Speaking only as an amateur of traffic manager type games, I see the problem as correctly identifying the repercussions of localised disturbances and I feel that each actor preemptively trying to attenuate jams would merely distribute the throughput problem to the adjoining network. Though the jams themselves might be resorbed, doesn't it stand to reason that (assuming the problem does distribute/diffuse itself) a localised jam may be better than a systemic slowdown? (eg, effect on emergency services, trade...)


We have variable speed limits here in the UK. During rush hour the speed will change (enforceable using cameras and fines) they also monitor the roads for these shock waves and alter the speed limit to reduce them.

We also have variable lanes, or tidal lanes. So a 3 lane road will have 2 lanes into and one lane out of the city in the morning and in the evening one lane in and 2 lanes out. Makes the most use of road space.

I imagine with more autonomous cars lanes and speed limits will become a thing of the past. Computers will just pack cars onto a road in the most efficient way possible.


"...and brake as hard as possible " I think braking hard is the beginning of all evil. Others behind will also have to brake hard and a whole bunch of cars comes to a standstill. Keep moving is key, which means: anticipate what will happen in front and react accordingly. This would also fit to the gas theory: as long as there is no shockwave (braking), things still flow. If things get denser in front of the car, move slower, as in a gas, but try not to stop.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: