Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Traffic Simulator

traffic-simulation.de

Post reply on HN
Print view View original

21–30 of 143 posts

Re: Traffic Simulator

#21

Post by jeffhuys »

pc86 wrote: post #19
vbtemp wrote: post #16

Things like this are cool. What I would be curious to see though, rather than setting global parameters (i.e., politeness), would be to have each agent have its own politeness score (as well as other attributes). Models like the one linked to here work as if each agent is a particle, rather than autonomous decision-maker adapting to their circumstances locked in long-term, iterated games with other anonymous drivers.…

This would be a neat reinforcement learning project. > Some places the drivers are so polite to the point it really makes things dangerous I've known people like this, who will approach a 4-way intersection and stop even when they don't have a stop sign and the perpendicular street does.

Participating in traffic all comes down to being as predictable as possible, I think.

Re: Traffic Simulator

#22

Post by kharak »

searedsteak wrote: post #14

Looks like maxing out acceleration solves all other problems. I guess we should require a minimum power-to-weight ratio for cars from now on.

I've made the same observation. But I don't believe human drivers could reasonably take advantage of high acceleration, at least not as well as the simulated cars here. This is a great showcase of the usefulness of autonomous cars, though. Take humans out of the equation and traffic will flow.

Re: Traffic Simulator

#23

Post by bones6 »

On the ring map, with default settings, just stopping traffic with the traffic light for a second produces the dreaded rubber-band effect propagating forever around the circle. Fun simulator!

Re: Traffic Simulator

#24

Post by amcoastal »

bhupy wrote: post #20
Cthulhu_ wrote: post #18

Earlier quoted context omitted.

In practice that would mean a lot more head/tail accidents, I've seen that happen plenty of times in traffic jams from over-eager drivers, sudden stops (either people trying to go too fast or not paying attention), etc.

Easier to do with self driving cars, and can probably be attained with Level 3 autonomy.

Yeah, and there would be no reason to increase acceleration -- increasing the acceleration only aids in this project because it shortens the time when the car behind the stopped car in front of it can get up to speed. If the cars are all talking to each other -- they can all start accelerating at the exact same time and rate as soon as the light turns green etc.

Re: Traffic Simulator

#25

Post by andrewstuart2 »

kharak wrote: post #22
searedsteak wrote: post #14

Looks like maxing out acceleration solves all other problems. I guess we should require a minimum power-to-weight ratio for cars from now on.

I've made the same observation. But I don't believe human drivers could reasonably take advantage of high acceleration, at least not as well as the simulated cars here. This is a great showcase of the usefulness of autonomous cars, though. Take humans out of the equation and traffic will flow.

The other thing that really seems to make the biggest difference (in combination with better acceleration) is following distance, which is another thing that presumably gets less important with autonomous cars because you don't really have to account for a relatively slow reaction time or poor braking.

The observation on acceleration does make me wonder whether teaching proper merging etiquette better (reach freeway speeds before you're at the merge junction) would make a significant difference if properly followed.

Re: Traffic Simulator

#26

Post by andrewstuart2 »

amcoastal wrote: post #24
bhupy wrote: post #20

Earlier quoted context omitted.

Easier to do with self driving cars, and can probably be attained with Level 3 autonomy.

Yeah, and there would be no reason to increase acceleration -- increasing the acceleration only aids in this project because it shortens the time when the car behind the stopped car in front of it can get up to speed. If the cars are all talking to each other -- they can all start accelerating at the exact same time and rate as soon as the light turns green etc.

There's also the frequent problem of split attention when you're accelerating like that. Often you may be trying to change lanes, and checking a mirror for an opening. Even without that kind of better communication, computers are just better at tracking multiple data points at once.

Re: Traffic Simulator

#27

Post by vbtemp »

pc86 wrote: post #19
vbtemp wrote: post #16

Things like this are cool. What I would be curious to see though, rather than setting global parameters (i.e., politeness), would be to have each agent have its own politeness score (as well as other attributes). Models like the one linked to here work as if each agent is a particle, rather than autonomous decision-maker adapting to their circumstances locked in long-term, iterated games with other anonymous drivers.…

This would be a neat reinforcement learning project. > Some places the drivers are so polite to the point it really makes things dangerous I've known people like this, who will approach a 4-way intersection and stop even when they don't have a stop sign and the perpendicular street does.

Places where I've seen it, people who have the right of way take that to mean they get to play "traffic cop". E.g., at a 4-way stop the "extra polite" person who gets there first, instead of going through the intersection first, waits and waves the person who does not have the right of way to go through first, which makes that person (and everyone else) confused, and then causes a lot of slowdown, frustration, etc. Of course only happens in certain locales, where it's part of the local driving culture.

Re: Traffic Simulator

#28

Post by beerandt »

jeffhuys wrote: post #21
pc86 wrote: post #19

Earlier quoted context omitted.

This would be a neat reinforcement learning project. > Some places the drivers are so polite to the point it really makes things dangerous I've known people like this, who will approach a 4-way intersection and stop even when they don't have a stop sign and the perpendicular street does.

Participating in traffic all comes down to being as predictable as possible, I think.

Yes, but in a way that's obvious to other drivers.

Driving decisively is more important to overall safety than driving defensively.

Re: Traffic Simulator

#29

Post by bhupy »

amcoastal wrote: post #24
bhupy wrote: post #20

Earlier quoted context omitted.

Easier to do with self driving cars, and can probably be attained with Level 3 autonomy.

Yeah, and there would be no reason to increase acceleration -- increasing the acceleration only aids in this project because it shortens the time when the car behind the stopped car in front of it can get up to speed. If the cars are all talking to each other -- they can all start accelerating at the exact same time and rate as soon as the light turns green etc.

That's true only when every car is a self driving car. In the interim, if only a subset of cars is running Level 3 autonomy, then including a quick acceleration into the adaptive cruise control is a good way to minimize these phantom traffic jams.

Re: Traffic Simulator

#30

Post by t0astbread »

throwaway0a5e wrote: post #12
snarfy wrote: post #8

Earlier quoted context omitted.

You can almost always fix the traffic by maxing out acceleration. Sadly, this vindicates my real world aggressive driving behavior.

Why is it sad if it turns out a subset of people are already behaving in a way that analysis indicates promotes better flow?

Because aggressive driving is dangerous? Perhaps aggressive driving also contributes negatively to the overall flow in the real world when not everyone drives equally aggressive (i.e. through extra lane switches caused). But I don't know much about traffic analysis so idk.
Post reply on HN