• Welcome to SC4 Devotion Forum Archives.

NAM Traffic Simulator Development and Theory

Started by z, August 02, 2008, 05:07:50 PM

Previous topic - Next topic

0 Members and 2 Guests are viewing this topic.

jplumbley

Quote from: Pharaon-Kheops on November 11, 2009, 04:52:19 AM
Do any of you have financial interest concerning the use of his own created simulator??? Is there a price for being the creator of the one and only simulator used in the NAM???
The reality is that there are many game style and thus the need for several traffic simulator, each adapted for a different styles... There is room for the both of you in this community and you  not even have to apreciate each other... Just keep improving your work, each with his own fans and detractors! I even hope we'll soon have more different simulators...
But please, stop this ridiculous fight! None of your sim is better than the other, they only have different goals!

No there is no financial gain.  I have not once said there should only be one Simulator in the NAM, in fact I am fighting to keep that choice there and I agree with you that I hope to see more people like Lenny trying to understand the Simulator and make their own for the Community.  I learned that lesson last year, and I definitely tried not to fight with Steve again this year but I got dragged in by his comments about my work and my posts which generally were not directed to him at first.  I have noticed a lot of the same attitude and response to critism from Steve that I had last year towards him, and I took almost a year away to come back and realize that it doesnt matter who's Simulator is better.  My intent here was not to attack Simulator Z or get into a direct confrontation with Steve and up until yesterday I did fairly well at that....  My intent was to help Steve realize that there is room for all the Simulators something he has started to come around in by finally making a suggestion to keep both Simulator A, Z and the Maxis Default Simulator in the NAM while removing all the other choices.  And I also wanted Steve to stop touting his Simulator as the best in all situations, while attacking Simulator A as sub-par.

All I have been looking for is for him to see that there is no "perfect" Simulator, and for him to give the credit to Simulator A it deserves and stop bashing it as he has for the last year with impunity.  I admit he is making strides and has come around saying Simulator A is good, but still has made his attacking remarks about it and I thought we were getting close to a truce on that until yesterday.  On the "perfect" Simulator idea though, I do not believe we have made any headway on that one, unfortunately.

It will be a good thing for Steve to take some time away, as I did last year.  Clear his mind, reflect on some things and come back hopefully with a rejuvenated desire to continue his work....  I just hope he realizes how insignificant this is though.
"You learn something new everyday."

http://img517.imageshack.us/img517/169/nhpjplumbleykv3.gif
Bringing the new horizons closer to reality.

Berethor ♦ beskhu3epnm ♦ blade2k5 ♦ dmscopio jplumbley ♦ moganite ♦ M4346 ♦ Dedgren ♦ Ennedi Shadow Assassin ♦  Tarkus ♦ wouanagaine
Street Addon Mod - SAM

Blue Lightning

#341
The thing is... both of your simulators do a great job in tweaking the internal traffic engine of SC4. Sim A was with me when I was building my largest city ever (2.5 million sims). Sim Z brought me through my most ambitious region-wide transit project. Both situations loaded the simulator quite a bit but they both ended up running them next to perfectly. I had not experienced mass abandonment and redevelopment cycles, huge traffic bottlenecks (only localized ones, in the places you would expect them to be), or any sort of major issue. And this isn't because I adapt my building style to the simulators (I've been playing with the same basic ideas for a long while now, with some new things here and there, but the core is still the same).

Sim A and Z are both great simulators, I must say that. Both Jason and Steve, you two have done a great job in building these. Though the conflicts really don't seem to be doing any good to either.

I've seen projects at other sites (not related to SC4) where projects have gone down the drain because of people in the same team come up with different ways to, for example with a community game I followed a few years ago, build a physics engine for a game. Both ways were perfectly valid and would have worked, with having their own strengths and weaknesses, but neither could have been used and developed further since the two developers were constantly quarreling. The game ended up stalling in development.

Jason, Steve, I'd like to ask you two if this conflict is really helping you, your simulator, or the community. I can imagine that this is not helping the advancement very much, and this is getting in the way of people wishing to learn about traffic simulators, and those who are choosing. Now remember, I'm not placing blame on either one of you.

All I ask is that you two please, drop this and accept each other and his simulator the way they are. Cool off and continue making advancement in the area of traffic simulators. You two may be taking different paths, but they both lead to the same place. Its just a matter of style. Neither is wrong, but both are right.

Now Steve, I personally don't want someone like you, who has contributed so much, to leave. However, if you really feel the need to take a break, I can accept that. Sometimes we all need a break from our tireless efforts.
Also known as Wahrheit

Occasionally lurks.

RHW Project

jplumbley

Quote from: Blue Lightning on November 11, 2009, 12:53:57 PM
All I ask is that you two please, drop this and accept each other and his simulator the way they are. Cool off and continue making advancement in the area of traffic simulators. You two may be taking different paths, but they both lead to the same place. Its just a matter of style. Neither is wrong, but both are right.

This is exactly what I have been trying to say....  All I want is for Steve to stop putting down my work while promoting his and accept it.  That is all that has had to happen, but it has not yet happened.  I have already accepted his as good work, all you have to do is read my posts over the last few weeks to see it.
"You learn something new everyday."

http://img517.imageshack.us/img517/169/nhpjplumbleykv3.gif
Bringing the new horizons closer to reality.

Berethor ♦ beskhu3epnm ♦ blade2k5 ♦ dmscopio jplumbley ♦ moganite ♦ M4346 ♦ Dedgren ♦ Ennedi Shadow Assassin ♦  Tarkus ♦ wouanagaine
Street Addon Mod - SAM

sumwonyuno

#343
z,

I would be at a loss if you did permanently leave this site and all other Simcity fansites.  A break is welcomed if you need one.  You have contributed more to the game than I ever will.  There are numerous things about the game mechanics that I learned from your work.  I've been an advocate of Simulator Z ever since I found it a while (a year?) ago.  I did not mean my latest test findings to refute all of the testings that you've done.

Isn't the difference among the traffic simulator plugins just the collective changes of the exemplar values?  That makes it all seem trivial, as it's not about designing completely different algorithms, but just varied changes to the input of the same algorithm.  Given either Simulator A or Z and a city tile of my region as input, the game will still do its job.  Now, Simulator Z does a good job at showing how inadequate the transportation network of Capitalis is and highlighting the available and appropriate job a Sim can get to the fastest.  For any traffic simulator, what may be the merits and consequences of one city tile, may be the complete opposite of another.

I wouldn't want the NAM to just have one traffic simulator.  If that traffic simulator customization tool comes out, I do hope there's presets.  If the goal is for the benefit of the (diverse gameplay style) community, the outcome of a competition for the one "best" traffic simulator isn't going to be a benefit to everyone in the community because of diverse gamplay styles out there.  It doesn't matter who apologizes to who, explicitly or implicitly.  It just matters if there is a mutual understanding between you and jplumbey that any discussion that devolves to the point of personal politics isn't going to do anyone any good.  Both of you are good (in understanding the game) and both of you show that you both can bring much bounty to the table that is SC4Devotion.com.


The City & County of Honolulu, a Mayor Diary based on Honolulu, Hawai'i.

mark's memory address - I've created a blog!

RippleJet

#344
Jason and Steve, it's a pity you always have to fight.
It makes it even harder to find and understand the results of all your hard work.


Now, it's taken me till today to completely read and understand Steve's testing November 7th (and the follow-ups on that).
Steve, a huge kudos to you for actually taking your time to make these extensive tests.
At least I am convinced that the pathfinding heuristic does play a significant role.
Also the influence of the Max Commute Time on the computing time was quite interesting and well presented.

However, I wish you would stop starting your posts by promising the solution to all problems we've ever had.
If there's something we've learnt by now, it should be that there is no (and never will be a) single ultimately best version of a traffic simulator that suits everyone.


Jason, you need to give Steve credit for his testing as well.
At least he is wasting a lot more time than you doing them.

However, your posts following those tests is where the level of the discussion started to deteriorate.
Instead of commenting on the main results of the tests, you focused on nitpicking on jokes you didn't understand, relations to R$$$ demand and whatever...
You know, you don't have to live up to that Plum Fierce CML of yours...

Btw, I would love to see a Simulator A with a lower PH.
Shouldn't Steve's testing lead to you testing Sim A yourself with a lower PH?


Then over to sumwonyuno's excellent and valuable testing November 10th.
That perfectly demonstrated the side-effects of having an "unlimited" maximum commute and too high network capacities.
In your case you did not experience eternal commuters around corners, you got them going straight through the whole city tile.

At this point I have to confess that I have for some time had concerns regarding Simulator Z and its influence on eternal commuters,
and I'm glad sumwonyuno so clearly brought this out in the daylight.

Steve, I know you feel a bit cornered here and I hope you could take a deep breath before your next post.
I don't think there is a solution that in the same simulator would both solve the above problem and allow your highly marketed intercity commute.

z

A brief technical note here...

Quote from: RippleJet on November 11, 2009, 03:46:22 PM
However, I wish you would stop starting your posts by promising the solution to all problems we've ever had.
If there's something we've learnt by now, it should be that there is no (and never will be a) single ultimately best version of a traffic simulator that suits everyone.

I tried to frame all of these tests in terms of testing parameters used in the simulator, not the simulators themselves.  When this didn't appear clear in the reactions to my first test, I specifically added it in the disclaimer to the second.

I agree completely that "there is no (and never will be a) single ultimately best version of a traffic simulator that suits everyone."  Even for users of Simulator Z, there would have to be too many versions in order to perfectly suit every situation.  I was starting to explore a whole different approach, based on the use of the NAM Tool or something like it, and elaborated by concepts introduced by Jonathan and Andreas, that would allow the proper simulator to be custom-built for users at installation time and based on their needs.  It would not be Simulator Z or Simulator A; it would be a whole different approach to simulators.  But to do that requires extensive understanding of what the relevant parameters in the simulator do.  Hence my tests.

QuoteThen over to sumwonyuno's excellent and valuable testing November 10th.
That perfectly demonstrated the side-effects of having an "unlimited" maximum commute and too high network capacities.
In your case you did not experience eternal commuters around corners, you got them going straight through the whole city tile.

At this point I have to confess that I have for some time had concerns regarding Simulator Z and its influence on eternal commuters,
and I'm glad sumwonyuno so clearly brought this out in the daylight.

The high commute time is one of the first possibilities that occurred to me as well.  That's why the second file I sent to sumwonyuno had a commute time slightly lower than that of Simulator A, with capacities very similar to that of the version of Simulator A that worked (Simulator A (Hard) & Simulator Z (Low)).  I was surprised to find that this did not help at all.  But this would seem to rule out high commute time and high capacities as the problem, as these did not exist in that test.

I would have liked to continue the testing to find out what the problem is, but by this point, things had gotten pretty much out of control.

sumwonyuno

#346
RippleJet, z clarified what I was going to say.  There is the issue I descibed in the second test of November 10th (the 9th where I live).  At face value, it seems like a bad thing for Simulator Z, but I wouldn't jump to that conclusion.  Whether I used Ultra, High, Medium, Low or Classic, the behavior is still there.  Network capacities and commute time were ruled out.  I would guess it's some other exemplar property value that's the cause of the issue.

I can't remember where, but I asked z sometime before about neighbor connection jobs.  If Sims were going to the closest (available and appropriate) job, then for a particular area of the city tile, all neighbor Sims that get sent to the same neighboring city tile should be taking the same path to there.  But that wasn't the case.  Sims were crossing the freeway and taking other neighbor connections.  Even for neighbor connections on adjacent tiles, Sims weren't choosing the closest.  z thought it was some randomness in the simulator.

So under Simulator Z (any version), for some reason, Sims weren't going to the south shore to work and westbound commuters from the eastern suburbs were skipping the Downtown city tile.  With Simulator A, this behavior did not exist.  The first test I did on November 9/10 was the secret subway.  I did get Sims to work on the southern shore with Simulator Z.  Even some westbound-freeway bus riders got off the freeway and transferred to the secret subway.

I made a statement on page 16 that z agreed upon:
QuoteIt almost seems as if commuters aren't trying find the closest (available and appropriate) job and then a path to it, but rather the available and appropriate job they can get to the fastest.

z then stated "Is it faster for them to get to the next tile than to get to the unused jobs?  That would explain everything."

It's established that Maxis used A* Manhattan pathfinding.  There is the assumption of a start point and a destination point.  A path must exist that includes those two points for a Sim to be employed at an appropriate and available job.  I've come to believe that the assumption that the destination is the closest (available and appropriate) job isn't necessarily correct.  If the closest (available and appropriate) job always holds true, then it wouldn't matter if I used Simulator A or Z.  The commercial jobs to the south of the freeway are always "closer" to the westbound commuters than the Maxis highway neighbor connection on the opposite city tile edge.

If closest job assumption is correct, how would that work?  Would the destination-determiner do closest by Euclidean distance, path distance or random?  Euclidean distance would probably have to do radii around a Sim's building.  But then, the Euclidean distance is ruled out because of those "ignored" jobs in the city tile.  Random would be a good candidate for destination-determiner, but it's quite common to see all commuters of a residential building work at the same employment building.  So what about path distance?  Speed of network wouldn't be a consideration, and we'd essentially be at shortest path pathfinding.  The traffic simulator would have to iterate through the possibilities of 1 to X tile lengths and pick whatever comes first, but that's extremely inefficient route to take.

Back to assertion I made, the available and appropriate job a commuter can get to the fastest answers the destination question and the pathfinding problem.  A* does find the fastest path from a start to a destination point.  Then, from what I gather, the traffic simulator would not have a predetermined destation point; the traffic simulator finds a destination as it's going through A* Manhattan.  Under that idea, for Simulator A, it's faster for the westbound freeway commuters to get off the freeway and take the jobs in the city tile.  For Simulator Z, it's faster for the same commuters to stay on the freeway.  With the secret subway test, it was faster for the Sims living in the high-rise residential area to take the secret subway to the southern shore than to crowd the neighbor connections of the city tile to the east.

Relatively speaking, Capitalis only has a few high-capacity east-west thoroughfares in a general east-west linear city layout.  North-south roadways don't have the capacity to handle both north-south commuters and east-west commuters that turn left/right to get to jobs.  So, traffic simulator is going to have to work with the limited number of alternate routes for a dense and linear urban region.

Anyone, feel free to add or criticize.

[EDIT]

I dug through the pages of this thread and found pictures of my region:



With Simulator Z Ultra beta 2 Road Volume:


Even though the above picture was from another test, the city-tile-skip issue is still there.

There is the single Maxis highway through the middle of these 3 city tiles.  Almost all of that red is westbound commuters from the suburbs to the east.  The one red north-south roadway is from the high-rise residential commuters from the Downtown (center) city tile that have decided to work in Waikiki.  The commuters aren't getting off the freeway, even though the surface streets leading away from freeway aren't congested with Ultra.

[EDIT 2]

I understand the congestion views do not have the same volume values, so I kept them in the pictures.

Simulator Z Classic - Car Volume

Simulator A Hard - Car Volume


The thing to notice is the area to the southwest of the crater.  With Simulator Z, the westbound freeway has about 40,000 cars.  Only about 3000 cars get off the freeway, and the remaining 37,000 cars continue on to the city tile to the west.  With Simulator A, the westbound lanes of the freeway have about 30,000 cars and almost all of those cars exit off the freeway.

Here are the bus volume views.  I don't know what conclusions to get from these results.

Simulator Z Classic - Bus Volume

Simulator A Hard - Bus Volume


The City & County of Honolulu, a Mayor Diary based on Honolulu, Hawai'i.

mark's memory address - I've created a blog!

jplumbley

Quote from: RippleJet on November 11, 2009, 03:46:22 PM
However, your posts following those tests is where the level of the discussion started to deteriorate.
Instead of commenting on the main results of the tests, you focused on nitpicking on jokes you didn't understand, relations to R$$$ demand and whatever...

Needless, classless jokes especially when the person knows the history is outright wrong.  And yes I will call him on it every time he does, and any other person who does.  I would never crack a joke in a situation like this about the person who I have had problems in the past with.  I believe he knew exactly what he was doing was wrong, if he didnt... well, I cant say what I think then.

Nitpicking, isnt that how you are supposed to analyze a test you have done?  Look for all the things that could have caused the results and determine whether they could or could not have affected your outcome...  I was trying to contribute and it didnt seem to me Steve had thought about those things and was looking for results that would prove him right and jumping to conclusions.

I feel the discussion deteriorated prior to my responses on his tests, but that is just my opinion and if you want me to point out where I can easily do so.

Quote from: RippleJet on November 11, 2009, 03:46:22 PM
Btw, I would love to see a Simulator A with a lower PH.
Shouldn't Steve's testing lead to you testing Sim A yourself with a lower PH?

Im sorry if I havent made one fast enough for you.  I have not had the issues Steve has reported, how would you like me to test them in the first place if I do not experience them in my cities.  None of my regions, let alone cities reach the population he does in one city tile, where the problem is most likely to occur.  I have no ability to test it if I wanted to, unless I built a multi-million Sim city which I have never reached because that is not the way I play.

Based off of Steve's testing I have been asking questions analyzing his tests with my own questions (focusing on the so-called nitpicking topics) trying to understand what is happening since what Steve shows is not always the whole picture.  Not saying he is hiding anything, but maybe something he notices but neglects to add in his posts because he simply forgot or thought it was obvious or something he simply missed (we all do it).  I have definitely been thinking about making a Simulator A with the .003 PH, but I wanted more answers first.

But, since I am too lazy and cannot reproduce his experiences with my own cities, then I have doubts that must be answered.

Quote from: RippleJet on November 11, 2009, 03:46:22 PMYou know, you don't have to live up to that Plum Fierce CML of yours...

I didnt give myself that CML...  I had a idea why it was given to me, now I know.

"You learn something new everyday."

http://img517.imageshack.us/img517/169/nhpjplumbleykv3.gif
Bringing the new horizons closer to reality.

Berethor ♦ beskhu3epnm ♦ blade2k5 ♦ dmscopio jplumbley ♦ moganite ♦ M4346 ♦ Dedgren ♦ Ennedi Shadow Assassin ♦  Tarkus ♦ wouanagaine
Street Addon Mod - SAM

ldog

#348
Well, as people might have noticed I dropped out of this conversation right after things started really going downhill.
While I still had things to say, I did not feel they would do anything but add fuel to the fire.
And apparently that fire already had more than enough fuel.

Tage, I think you summed it up pretty well (that video was priceless  ;D ), although you were maybe a little bit harder on Jason than he deserved, but I agree with the points made.
Pharaeon also makes some good points.

All of us involved were getting defensive, which inevitably leads to the counter-attacks that usually turn into full fledged wars.

While I have made my apologies to Steve in pm for my part in the whole thing, I have still felt that some sort of public apology should be made.

Intentions may be good, but it is very hard to see intentions across the forum. We only see words and their effects. While I still respectfully disagree on several points, I think there has been too much emphasis on disagree and not enough on respect. I think maybe that might be true for some others around here as well, but as we are all presumably adults here I will leave everyone else to their own discretion.

This bickering doesn't do any of us any good. It is a waste of time we could all put to better use.
It certainly isn't worth losing someone who brings a lot to the table over.

Anyway, that's enough candy-ass heartfelt touchy feely diplomatic stuff for one day; I'm gonna go back to being a jerk now before someone comes and takes my "man card" away, mmkay  :P

RippleJet

Quote from: z on November 11, 2009, 04:33:50 PM
The high commute time is one of the first possibilities that occurred to me as well.  That's why the second file I sent to sumwonyuno had a commute time slightly lower than that of Simulator A, with capacities very similar to that of the version of Simulator A that worked (Simulator A (Hard) & Simulator Z (Low)).  I was surprised to find that this did not help at all.  But this would seem to rule out high commute time and high capacities as the problem, as these did not exist in that test.

I would have liked to continue the testing to find out what the problem is, but by this point, things had gotten pretty much out of control.

Steve, I would very much like to see a continuation regarding this discussion.
Let me also ensure that things now are under control in here.

z

Quote from: RippleJet on November 15, 2009, 01:01:38 PM
Steve, I would very much like to see a continuation regarding this discussion.
Let me also ensure that things now are under control in here.

I will be happy to continue this discussion.  I've been using the break during the last few days to build my new computer, though, and right now, everything's in pieces all over the place.  (This is my laptop that I'm writing from, which doesn't have any of my notes or files.)  It will probably be around the end of the week before I get everything back together enough to do anything useful, but you'll hear from me as soon as I do.

In the mean time, I'd like to offer my apologies to the community for my part in the problems that have arisen.  I think ldog made some very good points in his post, and I wanted to say that I particularly agree with what he said about intentions and respect.  I have no question that everyone involved had good intentions in that we all want to make this a better game for the community.  I also believe strongly in the absolute importance of respect for everyone, but I know I have a style that can become argumentative in a way that is not at all helpful, and I think it's pretty obvious how that trait contributed to the current situation.  In addition, I've noticed how my approach in general can be unskillful at times, and create unnecessary contention.  Again, I apologize.  And more than that, I want to assure the community that in addition to working to create high quality content for the game, I will be working just as much to support harmony in this community, which of course is essential if we are to flourish.  I look forward to continuing to work with the community, and specifically to contribute to the whole range of qualities that make this such a great site.

Pharaon-Kheops

Well, that's a very good thing, Z... It would really be a sin to see you and JPlumbey unable to accomodate each other. The community needs the both of you precisely because you are heading toward different directions. As he said, his simulator is created for tile by tile region play, while yours is trying to find a way to play regions as metro areas. Personnaly, I'd rather play in the second category, but I definitely can't deny that sim A did solve many of my traffic problems at the time it did appear...

There is one and only one way to succed in your respective work guys.... this way is called Cooperation!  ;)
"to be is to do" - Kant
"to do is to be" - Sartre
"to be or not to be" - Shakespeare
"to be do be do" - Sinatra

b22rian

Quote from: ldog on November 15, 2009, 12:31:04 PM



Intentions may be good, but it is very hard to see intentions across the forum. We only see words and their effects. While I still respectfully disagree on several points, I think there has been too much emphasis on disagree and not enough on respect. I think maybe that might be true for some others around here as well, but as we are all presumably adults here I will leave everyone else to their own discretion.



I know this is the traffic thread.. and we will get back to that shortly...
but these words are about as well spoken and mean so much as anything ive seen in these forums..
Im very proud of you Lenny for this, if you dont mind me saying that to you :)

Brian

z

Quote from: RippleJet on November 15, 2009, 01:01:38 PM
Steve, I would very much like to see a continuation regarding this discussion.

I had sumwonyuno do a few experiments with his region after he made his last post.  The most surprising thing came out of the first test, which showed that there's a whole side of the traffic simulator that no one knew existed.  These experiments verified some of the things that sumwonyuno said, while disproving others.

First, I'll start with a description of what I now understand of how the traffic simulator works, followed by the evidence I have.  The traffic simulator contains the following two sequential parts:


  • The first part, the destination finder, finds a Sim (actually, a Sim's residence) and matches him or her with an appropriate job.
  • The second part, the pathfinder, attempts to find the fastest path to that job.

Until now, virtually all of the discussion and research with the traffic simulator had to do with the second part; this is where A* pathfinding, network capacities and speeds, etc. all come in.  Little was known about the first part.  Some people, like sumwonyuno, guessed that it took place simultaneously with the second part - that Sims could change their destination in the middle of a journey.  But on one hand, this is very difficult to fit in with the way that A* works.  More importantly, there is fairly conclusive evidence that these parts are indeed sequential.

So how does the traffic simulator select an appropriate destination?  There are standard factors such as matching up wealth types and looking at desirability.  Beyond that, it was assumed by many (including myself) that the simulator would look for the closest appropriate destination.  Since Maxis has stated that efficiency in the simulator was paramount, a simple method would need to be used.  The Euclidean (straight-line) distance presents itself as an obvious choice here.  It requires relatively few calculations, and will give a reasonably good answer.  The following test that I did about a year ago also gives strong evidence that that's what's happening:

In my Chicago region, I had a very bad eternal commuter loop.  Sims would enter the Downtown city at one corner, travel just a dozen squares or so, and exit into the next city.  Meanwhile, in the middle of my large Downtown tile, there were all these large office buildings standing vacant.  So I tried to get the Sims to go to them instead.  I cut all the connections between the incoming city and Downtown except for the one heavily used subway line, and then made that turn toward the middle of the tile.  I removed stations along the way, so that the Sims had to stay on the subway.  But when they finally emerged, they'd turn around and find some way to get back to the original crossing point into the third city.  I extended this "express subway" until it was at the edge of the central business district, but it made no difference; the Sims still turned around and headed out of the city.

At the time, I was quite perplexed by this behavior.  Now it makes sense.  When the simulator wanted to find a destination for the Sims entering the city, it just looked at the shortest Euclidean distance, and decided that the next city over made sense.  It didn't know or care that the Sims would have to do a long detour bringing them much closer to the empty skyscrapers than the third city boundary.  That part of the path wouldn't be known until the second part of the simulator ran, and had to create the long, tortuous path.  At that point, the destination was already fixed and could not be changed; the second part of the pathfinder could only determine the best way to get there.  And the big skyscrapers remained empty.

You can also see that the destination finder and the pathfinder may run at different intervals.  The pathfinder runs every four months or so and recomputes all the paths.  The destination finder may run as often, but only needs to run when there are Sims needing new jobs.

None of this presents a major problem.  The eternal commuter loop is well understood, and there are various ways around it; in my particular case, moving the allowed connection to the center of the tile edge made the central downtown area an attractive destination.

So all seemed well and good, and everything seemed reasonably well understood, until sumwonyuno's city came along.  In his downtown tile, Sims would drive by perfectly good jobs on the expressway and continue to the next tile when Simulator Z was run, but they'd get off the expressway and go to the jobs when Simulator A was run.

At first, this made no sense at all, and some quick experiments showed that the problem was not caused by differences in simulator capacities or maximum commute times.  This situation also seemed to completely disprove the theory that Sims would automatically go to the closest appropriate jobs.  Clearly, in Simulator Z, they weren't.  There was clearly something in the two simulators that was causing the different behavior.  It was if there were a parameter called Nearest Destination Attractiveness that was set much higher in Simulator A than in Simulator Z.

Well, that's exactly what turned out to be the case.  And when I had sumwonyuno triple the value of Nearest Destination Attractiveness in Simulator Z so that it matched the setting in Simulator A, large numbers of Sims started getting off the highway and going to the local jobs.

Well, that was easy.  Case solved, right?

Actually, no.  :(  In fact, this whole situation now looked like the worst mess possible.  :'(

Why, you might ask?  Well, when I asked sumwonyuno to triple the value of Nearest Destination Attractiveness, what I actually said was this:

QuoteCould you please take Simulator Z (Low) and change the pathfinding heuristic property from .003 to .009, and then see how it works in downtown Capitalis?

Those of you familiar with the traffic simulator may be aware of the Pathfinding Heuristic property, but see no property named Nearest Destination Attractiveness.  Others may see a property named Nearest Destination Attractiveness, and have assumed (as I did) that that was just another name for the Pathfinding Heuristic.

But it's not.  If you look at its name, it doesn't sound at all like what the PH does.  In fact, it does what I just described above.  Unfortunately, it's not a separate parameter.  Nearest Destination Attractiveness and Pathfinding Heuristic are two variables overloaded into the same parameter.  :'(  So when you lower the PH, which should always be safe, you also lower the Nearest Destination Attractiveness, which can cause the problem that sumwonyuno experienced.

Why did Maxis do this?  The basic logic is not hard to understand, given that the original game operates with a PH of .09.  With a PH that high, on a built-up large tile the Sims are lucky to find their way out the front door, much less find a job.  So Maxis apparently said, "OK, we'll make it so that the higher the PH, the closer to home the Sims will look for a job.  That way, they'll have a decent chance of finding one."  There's nothing wrong with that strategy, except that they overloaded these two very different variables into the same parameter.  Overloading variables is generally dicey at best, but this is one case where you don't want to do it.  Unfortunately, we're stuck with the result.

So what to do now?  I had sumwonyuno perform a few other tests, which also gave very strange results, but which indicated that there's even more going on here that we don't know about.  There's some hope that by adjusting other parameters in the simulator, the effect that sumwonyuno is seeing can be counteracted.  But it's going to take a fair amount of testing to see.  For the sake of efficiency, I've asked sumwonyuno to mail me a copy of his region and its plugins, and he's agreed.  I'll let everybody know what I find out.

In the mean time, what does this mean for other Simulator Z users, as well as for the proper setting of the PH?  For now, it appears to mean very little, if anything, for most people.  The situation in sumwonyuno's city is extremely anomalous; I've never seen or heard of anything like it before.  If other people have a situation similar to this, please let me know.  Otherwise, I wouldn't worry about it for now.  In general, the benefits of a lower PH still vastly outweigh this one anomaly, and I will be explaining new benefits that I've discovered in a post that will be coming soon.




At this point, I would like to mention that due to my RL situation, I am going to have to be cutting back a fair amount on my participation here.  I'll still be participating regularly, but it will be a little less frequently.  So if I don't respond to some posts or PMs immediately, that's why; responses will be coming, but they may just be a little slow.

Shadow Assassin

QuoteIf other people have a situation similar to this, please let me know.

I use Simulator Z medium, and I've noticed some strange anomalies occurring in my cities with regards to pathfinding. I have no idea why this behaviour is occurring...

1. I've got a farm that sends freight right through the centre of town, completely ignoring a connection to an adjacent city tile which is both shorter and quicker. For some reason, this farm chooses to use the more heavily trafficked route to get its freight out to the next city tile. Both possible routes are road. This is an old picture, however even though the city is now significantly larger, that farm still uses that route. The connection that the farm is completely avoiding is used normally otherwise by other freight coming from other farms in the area. Why would freight go such a long distance across the tile when there is a far superior option available?



2. In the adjacent city tile, I have sims getting on a train from the border, getting off at a train station right in the centre of the town in that city tile (which is some distance away, about 30 tiles as the crow flies)... then they all get on a bus then head back to the same city tile they were heading from, except through a different road connection that is some distance away. Both connections are actually close to the centre of the map. Also, the pathfinding tool is a little screwy on that section of rail: it sometimes reads zero people using that track, even though the station records usage. Is this a case of the 'eternal commuter', or is it something more sinister? This problem is very erratic, sometimes it happens, sometimes it doesn't.

3. Ferries. I have a case of ferries disappearing into the next tile where there is no safe connection (as far as visuals go and as pathing goes) between the border of the adjacent tile and a ferry station. In other words, there is a point where the ferry has to go over land to reach its destination. Also, it isn't recorded on the ferry volume data view in the adjacent city tile. So basically... commuters are disappearing, and I have no idea why.

4. I have had cases where the simulator will actually rather use a street (going the longer way) even though the street which formerly had all traffic using it was upgraded to a road. These problems were quickly fixed by severing the offending street into two.

The towns in question have a very complex street network, however curiously there is no abandonment at all (apart from farms due to air pollution, too much traffic).

Other than these anomalies, Simulator Z has worked quite well.

If you need more pictures, please ask.
New Horizons Productions
Berethor ♦ beskhu3epnm ♦ blade2k5 ♦ dedgren ♦ dmscopio ♦ Ennedi
emilin ♦ Heblem ♦ jplumbley ♦ moganite ♦ M4346 ♦ papab2000
Shadow Assassin ♦ Tarkus ♦ wouanagaine
See my uploads on the LEX!

RippleJet

Interesting reading, Steve! :)


Quote from: Shadow Assassin on November 25, 2009, 12:03:17 AM
2. In the adjacent city tile, I have sims getting on a train from the border, getting off at a train station right in the centre of the town in that city tile (which is some distance away, about 30 tiles as the crow flies)... then they all get on a bus then head back to the same city tile they were heading from, except through a different road connection that is some distance away. Both connections are actually close to the centre of the map. Also, the pathfinding tool is a little screwy on that section of rail: it sometimes reads zero people using that track, even though the station records usage. Is this a case of the 'eternal commuter', or is it something more sinister?

Yes, it very much sounds like being the same as an eternal commute loop.
What happens to those sims when they re-enter the city they orginated from?


Quote from: Shadow Assassin on November 25, 2009, 12:03:17 AM
3. Ferries. I have a case of ferries disappearing into the next tile where there is no safe connection (as far as visuals go and as pathing goes) between the border of the adjacent tile and a ferry station. In other words, there is a point where the ferry has to go over land to reach its destination. Also, it isn't recorded on the ferry volume data view in the adjacent city tile. So basically... commuters are disappearing, and I have no idea why.

Obviously it's enough if the first city knows there's water and a ferry port in the neighbouring city.
Since the path in that city isn't known to the first city, it cannot take into account the fact the the ferry would have to cross land...

Shadow Assassin

Quote
Obviously it's enough if the first city knows there's water and a ferry port in the neighbouring city.
Since the path in that city isn't known to the first city, it cannot take into account the fact the the ferry would have to cross land...

That's what I was thinking... but I wasn't sure. It's a bit weird to me, since I've never used ferries before.

Quote
Yes, it very much sounds like being the same as an eternal commute loop.
What happens to those sims when they re-enter the city they orginated from?

They don't reappear back in the city tile. It's really bizarre.
New Horizons Productions
Berethor ♦ beskhu3epnm ♦ blade2k5 ♦ dedgren ♦ dmscopio ♦ Ennedi
emilin ♦ Heblem ♦ jplumbley ♦ moganite ♦ M4346 ♦ papab2000
Shadow Assassin ♦ Tarkus ♦ wouanagaine
See my uploads on the LEX!

z

Quote from: Shadow Assassin on November 25, 2009, 12:03:17 AM
1. I've got a farm that sends freight right through the centre of town, completely ignoring a connection to an adjacent city tile which is both shorter and quicker.

This looks suspiciously like the type of problem that sumwonyuno has been experiencing, in that it could be a case of Nearest Destination Attractiveness being set too low.  Could you please try changing this parameter (or Pathfinding Heuristic, as it is more often named) from .003 to .009, run your city for a couple of years, and see if the route changes?  If not, could you exit your city without saving, run Simulator A on it for a couple of years, and see if that changes the route?

Quote2. In the adjacent city tile, I have sims getting on a train from the border, getting off at a train station right in the centre of the town in that city tile (which is some distance away, about 30 tiles as the crow flies)... then they all get on a bus then head back to the same city tile they were heading from, except through a different road connection that is some distance away. Both connections are actually close to the centre of the map. Also, the pathfinding tool is a little screwy on that section of rail: it sometimes reads zero people using that track, even though the station records usage. Is this a case of the 'eternal commuter', or is it something more sinister? This problem is very erratic, sometimes it happens, sometimes it doesn't.

Having read Tage's reply and your response, pictures would be helpful for this problem.  Also, following the steps I outlined for #1 would be very helpful here too.

Quote
3. Ferries. I have a case of ferries disappearing into the next tile where there is no safe connection (as far as visuals go and as pathing goes) between the border of the adjacent tile and a ferry station. In other words, there is a point where the ferry has to go over land to reach its destination. Also, it isn't recorded on the ferry volume data view in the adjacent city tile. So basically... commuters are disappearing, and I have no idea why.

I see that Tage has answered this one, and I agree with his answer.

Quote
4. I have had cases where the simulator will actually rather use a street (going the longer way) even though the street which formerly had all traffic using it was upgraded to a road. These problems were quickly fixed by severing the offending street into two.

How long did you let the game run after upgraded the street to a road?  It can sometimes take between five to eight years for a major change to be fully propagated, even though traffic will start appearing on the road relatively quickly.


ldog

Very interesting. It does clarify some of the nagging suspicions that I've had for some time.

If you care to send the files to me as well I will be happy to assist in nailing this down.

sumwonyuno

#359
ldog,

Do you mean my files?

I'll be going to the post office in an hour or two to mail z the DVD he asked for.  Within the next few days I'll be redoing my plugins folder in preparation for my plans for my mayor diary (creating custom content and rebuilding the region).  I'll have backups of the region and plugins folder as it is right now.

[EDIT]:  All right, I'll be sending both of your copies sometime this afternoon.


The City & County of Honolulu, a Mayor Diary based on Honolulu, Hawai'i.

mark's memory address - I've created a blog!