Traffic bug?!

[NorthStars]

I'm going to take a nice middle-of-the-road (pun intended) approach here...

1. I agree that the traffic/lane issue is not working quite the way it should, and I wouldn't mind if they were able to tweak it a bit.
2. HOWEVER, the more I work with my city, the less this issue bothers me in any significant way. If I work hard to lay down a really efficient road network, this issue doesn't seem to happen to me, or at least not much. When there's a problem with my network, it presents on screen in a weird way (long long lines in a single lane). But not if my network is done the way I think it should be.

The point is, I don't feel like I have to do "unnatural" things to fix this problem. It would bother me enormously if I had to do something weird to fix it, but honestly, at least in my experience so far, I just keep focusing on one single question... "If this were a real city, how would I fix the problem in real life?" Once I work that out, and I implement it, no more issues.

So, even though I think there might be room for some small improvements in the lane selection algorithm, I really hope they don't mess too much with that. What I really wanted from this game was a challenging traffic model that creates difficult situations that need to be resolved, but that are resolvable with effort. IMHO, that's exactly what the game gives me, so the little issues really don't bother me much.

 

[yuuzhantar]

Yuuzantar, you're idea is well thought out and for that I thank you. It's wonderful to see community members being constructively critical and trying to help the game be better.

The logic you present is solid, but the unfortunate matter is that what you are suggesting is so extraordinarily difficult to program. What seems like simple traffic logic to you, is very hard to put into mathematical formulas and call functions that don't use up all your computers memory. That's one big reason why vehicles have to choose their routes early on. Can you imagine the strain on your system if they were constantly calculating every other vehicle sharing a road with them, plus all the vehicles past the next intersection?

I'm not trying to shoot you down, you really have given great ideas that I've taken note of. But I ask that you trust that CO is doing the best they can to simulate traffic, which is an extremely difficult thing to do. We want to improve, you want improvements, but at the end of the day we need to accept that the traffic AI will never be perfect.

Hope that makes sense and doesn't come off too blunt! =)

In my idea, it's only the lane choosing which is made after an intersection

 

[arctvrvs]

Yuuzantar, you're idea is well thought out and for that I thank you. It's wonderful to see community members being constructively critical and trying to help the game be better.

The logic you present is solid, but the unfortunate matter is that what you are suggesting is so extraordinarily difficult to program. What seems like simple traffic logic to you, is very hard to put into mathematical formulas and call functions that don't use up all your computers memory. That's one big reason why vehicles have to choose their routes early on. Can you imagine the strain on your system if they were constantly calculating every other vehicle sharing a road with them, plus all the vehicles past the next intersection?

I'm not trying to shoot you down, you really have given great ideas that I've taken note of. But I ask that you trust that CO is doing the best they can to simulate traffic, which is an extremely difficult thing to do. We want to improve, you want improvements, but at the end of the day we need to accept that the traffic AI will never be perfect.

Hope that makes sense and doesn't come off too blunt! =)

Just to make sure I understand, does It mean that the path of any car is calculate either at the beginning normally or when you destroy a road ?
well for a computer performance point of view it makes sense (less calculus means less RAM taken)
I guess you already thought about it and my input is probably useless, but the possibilities I would explore would be:

• to take into account the traffic in the original path calculus: same as what google maps does, sometimes a national is better than a crowded highway (the macro Algorithm of Yuuzantar I guess)
  • Pros: a bit closer to reality (same as following a GPS path), should take care of some major traffic congestion
  • Cons: need to find a way to quantify the waste of time implied by the traffic congestion, still a few lines that are useless
  • Plus: maybe randomize between to paths that are close enough in time ?

• to teach them to use the other lines: As shown by Yuuzantar right now only one lane is used. What I would like to see is for the pathfinding to make the car cross the lines.

Let's say that you have ndirlanes lanes that allow you to go in the direction you want to go before the next intersection. d is the distance you're driving straight before making the direction change. Let's take a minimal distance for changing line equal to the walking distance (10 squares if I'm right), dwalk. The next direction you want to go to is dirNext(i)=(-1=right, 0=straight, 1=left) i being the step you're at
Let's numerate the lanes 1:ndirlanes, the lane 1 being the one on the right (movement direction) and the lane ndirlanes being the one on the left.

To be more clear, when you arrive at you intersection, you should have a matrix:
0 1 1 1 -1 -1 -1 0 The direction of the road (-1 incoming traffic at the intersection, 1 exiting traffic, 0 well the limit)
0 1 0 0 0 0 1 0 indicates which lane allows to turn right (>0 indicates that the arrow will be towards the right) or that cars will turn right into
0 0 1 2 3 2 1 0 indicates which lane goes straight
0 0 0 1 1 0 0 0 indicates which lane goes left (here the left lane allows to go straight and left)

another example, the 6 lane one direction street that with all arrow straight and left:
0 1 1 1 1 1 1 0
0 0 0 0 0 0 0 0
0 1 2 3 4 5 6 0
0 1 2 3 4 5 6 0

We will call this matrix intersection.

Now for a simple street intersection:
0 1 1 1 1 1 1 0
0 0 0 0 0 0 0 0
0 1 2 3 4 5 6 0
0 1 2 3 4 5 6 0
[---------------------0 0 0 0
[---------------------6 0 0 -1
[---------------------5 0 0 -1
[---------------------4 0 0 -1
[---------------------3 0 0 -1
[---------------------2 0 0 -1
[---------------------1 0 0 -1
[---------------------0 0 0 0
0 1 2 3 4 5 6 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 -1 -1 -1 -1 -1 -1 0

Now the rules for turning are -if I'm not wrong- if you have two lanes to turn right, you have two lanes to turn right into.
The rule here is the same: you are on line 5 to turn left, you go on line 5 to turn left into.

So, the car just spawned on a street (let's say our 6 lane one direction one), on lane 4.
It wants to go left next ( dirNext(1)=1 ) then left again ( dirNext(2)=1 ) because pathfinder said so.

you have chosen left, so you go to intersection(4+1,dirNext+3) and get lane=4
ndirlanes=max(intersection(find(intersection(1,==1),dirNext+3)=6;

LaneChange(i)=min(int(d/dwalk),ndirlanes-lane); the number of possible lane change before the intersection.
ideally it would go from lane 4 to lane 6 before the first intersection. (ndirlanes-lanes=2)
But if we have only d=1.5*dwalk, then LaneChange=1;

Now we just have to change line.

So basically, instead of having just a direction vector to decide the path of the car, we add a change of line one.
for i intersection, we have dir(1,i)=dirNext(i)=-1, 0 ou 1 et dir(2,i)=LaneChange(i).
dirNext(i)*LaneChange(i) gives how you move through the lanes (+1 lane, no movement (going straight), -2 lanes,...)


Well that was a part of my two cents but at some point I should probably go to sleep ^^
Then again, I don't have access to your code and what I propose may be incompatible with it. But I think that by putting the lane change at the same time as pathfinding, we should not burden the computer too much.

Well what I did was most probably useless but it was a fun intellectual exercise
In the scenario where you may like what I propose but don't quite understand, feel free to ask.

@Azurespecter, like I said I am conscious of the fact that I don't have access to your code, your past tries and that you guys are much more experimented at traffic simulating that I will ever be. But if you have some feedback that would be nice.
I anyway have more proposals and I will post them anyway but my level of abstraction will probably depend if the devs find this kind of input useful or not

 

[rbrucemtl]

In quantum computing could traffic be perfect? How about CS D-Wave Edition?

 

[justachap]

Build highway around edge of city like a big circle, connect ramps to everything. No more jams XD

 

[Iche_Bins]

[...]
The logic you present is solid, but the unfortunate matter is that what you are suggesting is so extraordinarily difficult to program.
[...]

Well there was the nice and easy idea by morwynd
https://forum.paradoxplaza.com/foru...rformant-traffic-lane-distribution-ai.846028/

Has that been given any more thought?

 

[Lord Canterbury]

Well there was the nice and easy idea by morwynd
https://forum.paradoxplaza.com/foru...rformant-traffic-lane-distribution-ai.846028/

Has that been given any more thought?

I just hunted through the forums to find that thread and you beat me to it. It would be interesting to know if CO have looked into it further.

 

[King Wingit]

Yuuzantar, you're idea is well thought out and for that I thank you. It's wonderful to see community members being constructively critical and trying to help the game be better.

The logic you present is solid, but the unfortunate matter is that what you are suggesting is so extraordinarily difficult to program. What seems like simple traffic logic to you, is very hard to put into mathematical formulas and call functions that don't use up all your computers memory. That's one big reason why vehicles have to choose their routes early on. Can you imagine the strain on your system if they were constantly calculating every other vehicle sharing a road with them, plus all the vehicles past the next intersection?

I'm not trying to shoot you down, you really have given great ideas that I've taken note of. But I ask that you trust that CO is doing the best they can to simulate traffic, which is an extremely difficult thing to do. We want to improve, you want improvements, but at the end of the day we need to accept that the traffic AI will never be perfect.

Hope that makes sense and doesn't come off too blunt! =)

We've heard several times that it's difficult to deal with lane choice vs congestion due to the load it would place on the cpu. Every time I see this, I can't help but wonder how much cpu is being used to simulate water physics. I'd happily trade a little bit of water flow for traffic flow.

 

[hammer256]

So what I've seen was that a car entering the highway from the right side, and has its route planned to exit on the right side (so entry and exit are on the same side of the highway, no need to cross lanes), still would merge onto the center lane from the right lane. Once close to its destination, it would go back to the right lane to exit. Can we change that behavior, so that the car stays in the right lane, or change the merging algorithm a bit, so the car doesn't come to a dead stop unnecessarily when merging on the highway?

 

[bluespottedhors]

Yuuzantar, you're idea is well thought out and for that I thank you. It's wonderful to see community members being constructively critical and trying to help the game be better.

The logic you present is solid, but the unfortunate matter is that what you are suggesting is so extraordinarily difficult to program. What seems like simple traffic logic to you, is very hard to put into mathematical formulas and call functions that don't use up all your computers memory. That's one big reason why vehicles have to choose their routes early on. Can you imagine the strain on your system if they were constantly calculating every other vehicle sharing a road with them, plus all the vehicles past the next intersection?

I'm not trying to shoot you down, you really have given great ideas that I've taken note of. But I ask that you trust that CO is doing the best they can to simulate traffic, which is an extremely difficult thing to do. We want to improve, you want improvements, but at the end of the day we need to accept that the traffic AI will never be perfect.

Hope that makes sense and doesn't come off too blunt! =)

Thank you for your response to Yuuzantar. I think the current traffic system is super. There are unusual driving habits but in the real world I think it's much worse and people seem to do everything against the rule of traffic laws. I don't know how traffic is calculated but in the real world people usual don't make last minute changes...unless there is a serious traffic jam and there is a way around it. So the calculation should be based on fastest roads and quickest time because normally higher speed road guarantee a shorter travel time...unless their is an accident or traffic jam. : )

 

[Myquandro]

What if the average congestion on a road is used to calculate the best route. The average congestion could be calculated and used to give higher chances to chose an alternate route. If about 10% of the cars would choose an alternative route when there is a more then average congestion rate on a certain road these traffic jams would be less frequent.

It's like in RL when someone knows that there is a considerable chance of traffic jams they might take another route to their destination, not everyone will do that, but some might.

This takes a bit extra calculations for the average congestion, but you don't have to recalculate paths along the way.

 

[MaxxPower]

So what I've seen was that a car entering the highway from the right side, and has its route planned to exit on the right side (so entry and exit are on the same side of the highway, no need to cross lanes), still would merge onto the center lane from the right lane. Once close to its destination, it would go back to the right lane to exit. Can we change that behavior, so that the car stays in the right lane, or change the merging algorithm a bit, so the car doesn't come to a dead stop unnecessarily when merging on the highway?

They only do this if there is an off-ramp between the point where they get on and where they get off the highway.
I am not exactly sure why they do this. Maybe to avoid getting stuck in a jam from an overloaded off-ramp.
Merging On-ramps don't cause this behavior, and other road types don't either.
Maybe it is a bug?

 

[Azurespecter]

It would be interesting to know if CO have looked into it further.

Sorry, I should have responded here sooner. I think the only appropriate answer to give here is "Yes" haha.

I don't have any announcements, but CO does review your ideas. I've collected quite a few great ones, including suggestions given by Morwynd, ItalianGuy, Yuuzhantar, arctvrvs, BrazlianJoe... and a bunch of others that I can't recall right now.

When I say "we're open to suggestions" it's not just a generic appeasement statement, we really are looking for improvements.

 

[AustinPowersFas]

Cars changing lanes in the middle of a road section only causes more jams!
If it changes into a lane where theres already congestion, it will only back up every car behind him.
This could work only IF cars were able to adjust paths DURING their trip to adjust their lane choice. This is unfortunately not possible as that would completely cripple ur pcs cpu.

In CiM2, cars choose the lane at entering a road section. If the first exit was their exit, they would choose the most right lane. If it wasnt, they would choose another. There were ways to make ur roads more efficient by increasing/decreasing number of lanes. Unfortunately we do not have that freedom (yet).

I get your idea, but i think we should have a system that we can direct traffic flow easily.

 

[Tropod]

Why can't a split level (or randomised) approach be used . In simple terms;
-Maintain existing; ie path choosen at start of journey. However;
-Randomise path routing for a portion/% of traffic, such that this % of traffic will recalculate their path part way through journey. This randomised amount should be smaller % so calculations don't break PCs.

In real world majority of commuters do plan their journey from the very start &/or use a particular route out of habbit. And yes people real world do change their journey mid way through but this typically only occurs if/when commute path choosen is an (or constant) issue, but even then some times people have no choice; like being stuck on public transport or stuck in highway jam. Having some sort of split system or way to alternate between 2 different methods for some portion of population would help this aspect of the game greatly. I don't believe some sort of single approach method is right or accurate.

 

[yuuzhantar]

Sorry, I should have responded here sooner. I think the only appropriate answer to give here is "Yes" haha.

I don't have any announcements, but CO does review your ideas. I've collected quite a few great ones, including suggestions given by Morwynd, ItalianGuy, Yuuzhantar, arctvrvs, BrazlianJoe... and a bunch of others that I can't recall right now.

When I say "we're open to suggestions" it's not just a generic appeasement statement, we really are looking for improvements.

Could you give us a feedback when the devs will have reviewed that?

 

[setapca]

Spoiler: Please explain

[/SPOILER

 

[Lord Canterbury]

@setapca
The first video looks to be a bug of some kind. I haven't seen that one happening in vanilla. Maybe a mod? Or maybe something strange with that bridge.
The second video is just a bad road layout. There are cars coming in on the left trying to exit on the right, and cars coming in on the right trying to exit of the left, and they are being forced to try to change over at the same point on a short stretch of road. You would be better with across-over intersection instead of a roundabout.

 

[Henry - South Africa]

Lately I have noticed that allot of times traffic in my city is causes by what looks like to be a bug in the game. My biggest traffic jams happen on the highway, cars will stop in the middle of the highway to change lanes. Treat the right lane as an off lane only, even though its not according to the arrows on the lane. So when traffic is getting on the highway they will stop until they can get on the middle lane causing a traffic jam.
I would like to know if this is happening to anyone else?

15 oCT'17

https://www.facebook.com/video.php?v=10154854089451669

Is it my video card (GeForce gt 740) or is there something wrong with SC? I play the game without any mods Cars on the highway are " see through"...Some cars you see others not. Was wondering why cars stop in the middle of the highway. Now I know..phantom cars... Any Idea why here?

 

[Tyrantissar]

The cars should try to change lanes if they notice one of them is too busy, no car should be hell bent on going into one particular lane. Another thing that bothers me is the game trying to take the shortest possible route all the time. Some cars should consider taking less busy lanes as the quickest possible route. The Ai should consider time as a factor instead of the shortest possible route. If the timing is better elsewhere because the traffic lanes in one particular area is too congested, than they will take the less busy streets. If they are going left or right, they should position themselves in such a way that turning that direction is easier.

I notice the Ai does not take into consideration how many cars are in front of them. If the could calculate the number of cars in the way to their destination Change into less busy lanes, maybe avoid busy streets etc.), I wonder if they would take the most optimal route. Add a cooldown or make them check every few seconds to it so that they don't always change lanes like crazy just because one way was shorter for a millisecond. Can anybody mod this in or something?

Anyways, there are some bugs present with traffic. I seen ambulances park sideways and there is about 7 of them sideways in another lane not moving very well.