RTMT Station Capacities

[z]

When Cogeo handed the RTMT project over to me last year, there were five levels of station capacities, each based on a different traffic simulator.  At the time, Cogeo urged me, for the sake of efficiency, to reduce those to no more than two, or maybe even one.  It seemed to me at the time that two levels were necessary as a minimum, and that's why ever since the release of the RTMT v3.51 patch, there have simply been Low and High Capacity levels of stations.

What has been learned in general about station capacity since then has caused me to re-evaluate that decision for the upcoming v3.70 release and beyond.  The way the game is set up, it appears to the player that a station's service quality gradually degrades once the station passes 100% of capacity.  This is visible both in the service quality icons in the station query, as well as in the way that the stations change color from green to red in the traffic congestion view, just like the networks.  However, extensive experiments have shown that the service quality does not change at all during this process.  A station that is at 300% of capacity and shows up as deep red on the congestion map processes passengers just as efficiently as a station at 5% of capacity.

There is a limit to how many passengers a station will process during a day, though, and it's a very hard limit.  Once this limit is reached, service doesn't start to decline; it simply stops.  It's as if the station just closed its doors for the rest of the day.  The limit seems to vary among different stations, but it can be as low as 400% of capacity, which is in fact a very common limit.  This means that many stations will work perfectly until they reach 400% of capacity, but then they stop working completely.  This is, in effect, a type of catastrophic failure.

The capacities of RTMT stations are set high enough so that such catastrophic failures should never happen; long before they do, congestion on the associated networks should limit traffic through the station.  But this means that having different capacity levels for stations really serves no purpose, since the stations never suffer the effects of congestion.  Therefore, I am proposing to eliminate the Low capacity level for stations, since its only function is to show lower numbers, and the lower capacities have no effect at all on game play.  Reducing the number of capacity levels to one will make future RTMT development and maintenance much simpler, and will also be one less thing that the player has to worry about, both during installation and during play.  Such a change would be backward compatible; stations previously plopped with the Low capacity would continue to function as before, and would retain their original capacity.  Comments on this proposal are welcome.

[catty]

The capacities of RTMT stations are set high enough so that such catastrophic failures should never happen; long before they do, congestion on the associated networks should limit traffic through the station.  But this means that having different capacity levels for stations really serves no purpose, since the stations never suffer the effects of congestion.  Therefore, I am proposing to eliminate the Low capacity level for stations ......

I use the low capacity stations, but my preferred choice when playing the game is always to use the high capacity stations

...and it will certainly make beta testing easier 

[Andreas]

Interesting findings. How can we determine that a station closed its doors, though? I have seen stations that ran at some 4000% in my game, and apart from the congestion, they appeared to work properly. I'd figure that a "hard limit" would mean a certain number, such as 65536 (which seems to be the limit for a neighborhood connection). But then you say it's "usually" 400% of the actual capacity of a station? Does the limit get resetted every in-game day, and how can we check this?

[z]

Interesting findings. How can we determine that a station closed its doors, though?

Through various tests, you can discover that a station's volume for a day never goes beyond a specific number.  It's often an integral multiple of the station's capacity, but not always.  The limit may vary by a few Sims from one traffic cycle to the other.

I have seen stations that ran at some 4000% in my game, and apart from the congestion, they appeared to work properly.

On the upper end, I've seen hard limits up to 10,000% of capacity, but that's rare.  More frequently, you can see numbers well above the hard limit for exactly one traffic simulator cycle, which typically lasts a number of months.  Then the simulator sees that it's gone over, and slams the cap on.

Does the limit get resetted every in-game day, and how can we check this?

Yes, it gets reset every day.  Some of the most definitive experiments were done in two consecutive posts by Tage and me in a private thread, but I think you can access them here, starting in the third paragraph of my post.  Tage's experiments were most extensive, and he came up with the 400% number, which I have since seen personally in other situations.  If you don't have access to that thread, I believe that Tage copied his post to another thread, which either he or I can find and point you to.

Here's a sample we happened to come across during our testing using the low-volume version of an SFBT station:



What's interesting is that this is no megalopolis; the population here is only around 150,000.  I know that SFBT has high-volume versions of these stations available, but I think you can see how someone could think that low-volume would be sufficient for a city of this size.  (EDIT:  This is another argument for getting rid of the low-volume versions of RTMT stations.)  Here's the RTMT high-volume station in contrast:



In this case, we're nowhere near the 618% limit for this type of station, so the station works at full efficiency, processing almost twice as many passengers.

Discovery of this effect is what led me to relabel "Service quality" to "Reserve capacity"; this still isn't completely correct, but it's the closest thing I could think of.

[Andreas]

Unfortunately, your link leads to nowhere, so please check it again. I don't know if I have access to that thread, but if I do, I didn't follow it closely, as the whole traffic simulator stuff is a bit over my head.

[z]

I made a mistake in copying the link - sorry about that.  It's fixed now.

...as the whole traffic simulator stuff is a bit over my head.

I have to confess that Tage's experiments were comprehensible to me only after long study, so you might have some difficulty there.  The ones I ran in that thread were a bit simpler; I can PM you a copy of the post if you can't access it, or I could even reproduce it here if people want.  Tage used more realistic numbers for capacities, but my experiments showed the exact same effect.

[Andreas]

Ok, looks like I can access that section (not sure if I could before or not, though) - I'll have a closer look once I got some more free time.

[ldog]

Steve,

I can't access that thread either

At any rate, at this point in time I agree with you. My own experiments have shown that station capacity does not seem to have much of an effect at all, at least not until the saturation point.
It also still isn't clear whether at that point the simulator continues to try to send sims through the station or just knows to use a different route; but my guess is that the effect on the pathfinder can't be good. As I think I said, my experiments in trying to use the stations to control use (no congestion) were not even remotely promising.

I think one set of capacitys, high enough for whatever traffic sim is sufficient. It is bad enough you need to maintain different TSEC versions for each let alone caps as well.
So not just for the RTMT but for all station batters it would make sense to have a single set of guidelines to make everyones life easier.

Lenny

[RippleJet]

I can't access that thread either

That post was later reposted in a public board, here:
http://sc4devotion.com/forums/index.php?topic=5962.msg193105#msg193105

[ldog]

That post was later reposted in a public board, here:
http://sc4devotion.com/forums/index.php?topic=5962.msg193105#msg193105

Error
The topic or board you are looking for appears to be either missing or off limits to you.

[z]

That link points to the original post on the private board.  The public version can be found here.

[ldog]

That link points to the original post on the private board.  The public version can be found here.

Oh...well why didn't you just say it was that thread
Everyone knows the "Why are there no Lots in the NAM?" thread. 

[RippleJet]

That link points to the original post on the private board.  The public version can be found here.

I'm having too many IE windows open at the same time...
It's amazing it doesn't happen more often that I copy the wrong link...

[z]

Oh...well why didn't you just say it was that thread
Everyone knows the "Why are there no Lots in the NAM?" thread. 

Except me, even though I made plenty of posts there; I couldn't for the life of me remember the name of that thread.  The old brain cells are dying off at an ever-increasing rate, as a number of people have noticed...    I'm lucky to have remembered the board the thing was on.

I'm having too many IE windows open at the same time...

IE?  You're still using IE?  I thought that was illegal. 

[ldog]

Except me, even though I made plenty of posts there; I couldn't for the life of me remember the name of that thread.  The old brain cells are dying off at an ever-increasing rate, as a number of people have noticed...    I'm lucky to have remembered the board the thing was on.

Drink more booze! It kills off the weaker brain cells quicker and makes room for new stronger ones to grow. Works for me 

IE?  You're still using IE?  I thought that was illegal. 

Yeah!
Shame on you Tage! 

[ldog]

Some more thoughts on the matter:

On the upper end, I've seen hard limits up to 10,000% of capacity, but that's rare.  More frequently, you can see numbers well above the hard limit for exactly one traffic simulator cycle, which typically lasts a number of months.  Then the simulator sees that it's gone over, and slams the cap on.

My recent tests with network capacity and CvS support this theory as well. When I raised the network caps by a large amount but did not change stations at all the useage shot skyhigh, but then every other run of the simulator resulted in the traffic volume shooting up or down. As if it let everyone and their brother use the station for 1 pass. Then the beginning of the next it realized that this was not possible so it sent them elsewhere. Then when it ran again it let the station get slammed again. (Sorry if I am repeating myself; I may or may not have gone into this detail elsewhere...you aren't the only one losing brain cells around here  )

I think Tage pretty well established what the minimum station capacity should be (in relation to network capacity) in the referenced post
( http://sc4devotion.com/forums/index.php?topic=6726.msg213985#msg213985 in case anyone forgot) although one flaw I would like to point out is the Ave/Highway toll booth.
Because while we know network capacity is per tile, station capacity is per station. So the ave and maxis highway tollbooths (and of course Ave RTMT) capacity should be doubled.
Other than that I think that really establishes the proper capacity but having a capacity that is too high is more desirable than one that is too low.

I use tollbooths on all of my city connections; freight has to get out so even the 0.2 TSEC is not enough deterrent to them and the revenue helps 
Now since I added a neighbor to Gridlock my tollbooths are deep red yet the underlying highways are green, and the caps being used were only 25% higher than defaults (5000).

So to make your life easy I would say use that formula applied to the highest capacity traffic simulator. Which would be Z ultra (I think). If you are doing a different version for A,B and Z because of different proper TSEC for each then you could still cut it down to only 1 version for each, and just use the highest capacity version of each sim.

[RippleJet]

IE?  You're still using IE?  I thought that was illegal. 

I cannot comment on that, since I would be doing that in an illegal window...

although one flaw I would like to point out is the Ave/Highway toll booth.
Because while we know network capacity is per tile, station capacity is per station. So the ave and maxis highway tollbooths (and of course Ave RTMT) capacity should be doubled.

Good thinking, Lenny!

[z]

Because while we know network capacity is per tile, station capacity is per station. So the ave and maxis highway tollbooths (and of course Ave RTMT) capacity should be doubled.

RTMT Avenue stations are already set to account for the whole avenue width.

Other than that I think that really establishes the proper capacity but having a capacity that is too high is more desirable than one that is too low.

That's definitely implied by these findings.  Now all we have to do is change all the stations... 

So to make your life easy I would say use that formula applied to the highest capacity traffic simulator. Which would be Z ultra (I think).

By a strange coincidence, that's exactly how the High capacity RTMT stations were set.

There are various formulas out there for how station capacities should be set, but I found through extensive testing that linear formulas don't work well, especially when multiple networks are involved.  Instead, the ideal formula would have a logarithmic component to it.  But since the station capacities don't have to be exact, I simply picked ones that worked in a wide variety of situations, with the capacities bearing a reasonable relationship to each other.

As I mentioned elsewhere, all networks actually have infinite capacities.  To match that, all stations would have to have infinite capacities.  But infinite capacity is not really a good thing for the simulation.  Instead, by using the current capacities for RTMT stations, we actually provide an upper limit on capacities for the networks that they serve.  This seems quite reasonable, and makes the game behave more realistically.  So from this point of view, RTMT stations are actually good for the traffic simulator.

My recent tests with network capacity and CvS support this theory as well. When I raised the network caps by a large amount but did not change stations at all the useage shot skyhigh, but then every other run of the simulator resulted in the traffic volume shooting up or down. As if it let everyone and their brother use the station for 1 pass. Then the beginning of the next it realized that this was not possible so it sent them elsewhere. Then when it ran again it let the station get slammed again.

Yes, this is exactly what I found in my tests, which were done at the same time as Tage's tests.  Since he's copied his post in a public thread, I might as well copy mine here, as the example is very relevant to your observations.  It also explains why this is happening:



There is a very real congestion that occurs in standard TE lots; the reason no one noticed it is that it works completely differently from network congestion, and has nothing to do with the CvS curve.  If a station maxes out at a certain level, whether it be 5x or 10x of capacity, then that means that no more Sims can enter it until others leave.  It's like a completely jammed freeway, where cars can't get on until others get off.  So at these station caps, total congestion occurs in the station.  It appears that since Maxis had no other way of simulating station congestion, they set these caps to at least provide an upper limit on station usage.

As for network-top TE lots, it's not that they have no volume cap, it's just that their volume cap is much higher than standard TE lots.  I have found through numerous experiments that this volume cap is 100x the nominal station capacity.  Why it's so much higher, I can't say, but it clearly has something to do with the through traffic running through the lot.  I have more details on these findings below.  But first, I'll show some pictures from a TE-lot-as roadblock experiment.

The following are pictures of a different area of the same city I've been displaying.  I take a standard RTMT bus and subway station, through which heavy traffic is flowing, and replace it with an RTMT subway station whose capacity has been reduced to one.  (I used a plain subway station for simplicity's sake; later, I reran the experiment with a bus and subway station identical to the one being replaced, and got identical results as the first experiment.)  First, the "before" pictures; here's one with the car volume view of the whole area:



And here's one of the exact same area with the subway view:



Now I demolish one of the combination bus and subway stations.  In a later experiment, instead of demolishing the station, I demolish a road tile next to it, so that I can control for the effects of disrupting traffic.  In both experiments, the subway volume goes to zero for a few squares on either side of the demolished station, as is normal, but nothing else changes.  Here is the picture from the first experiment, where the station was demolished:



Now I plop the modified subway station and run the game long enough for the traffic simulator to run.  Here's the new car view; car volume on the road has dropped by about 97%.  In the control experiment at this stage, car volume is beginning to recover, and is back to about 50% of the original volume.



And a year after the station was demolished and replaced, here's the new subway view, which is even more impressive.  Not only has subway traffic below that road dropped to virtually nothing, but the orange line that used to go down that entire road is now a light blue, indicating very little subway traffic.  When the traffic simulator saw that it couldn't use the subway below the road as a major thoroughfare due to the overly congested station, it rerouted traffic away from that whole subway line.



So I think that these pictures clearly demonstrate that there is a congestion effect for network-top TE lots.  But things continue to get interesting after this.  If I let the game run another ten months, the orange subway line comes back, just as in the first subway picture.  The road traffic also comes back, although at half its original volume, and it never rises above this volume.  Meanwhile, in the control experiment, the road traffic has risen all the back to its original volume.

The return of traffic here caught me by surprise, until I realized what was happening.  With no road or subway traffic, as shown in the final pair of pictures, station traffic dropped off until it was 61, as you can see in the final picture.  This is below the volume cap for network-top stations, so the station was no longer congested.  So the simulator once again started routing traffic through it on both road and subway lines.  The road traffic never got above half its original value, because that would have led to station congestion, which would have shut down the road and subway completely again.

In all the experiments I did, the highest station usage I got for network top stations with capacity of one was 100.  Not only is this a suspiciously round number, but it led to slightly different effects than what I described above.  First of all, I know that the 100 was an accurate count, because the only traffic around the station was bus traffic on only one side of the station, and it consisted of exactly 100 Sims.  There was no road traffic going through the station.  This situation was completely stable and non-changing for years.  It's easy to see why.  The station had stabilized at a level that was exactly its congestion limit.  So it was always completely congested, which kept road traffic from returning, yet there were still enough Sims using it to keep it at that congestion level.

The much higher numbers I reported seeing in these stations (e.g., 180,000% of capacity) occurred for only a single traffic simulator cycle, and can be accounted for by the one cycle delay that results from the way the finite state machine works.