Why is AI Pathfinding so broken and terrible?

[xenon5]

I really love the concept behind CiM and I love the series. That said, I'm completely dismayed by how poor the pathfinding is in CiM2. Although it wasn't perfect, I feel like it was better in the original CiM. Without fail, as soon as you start having connections with even a remote amount of complexity, the citizens are unable to complete basic pathfinding tasks. I build out a big network, then follow a few cims to discover that building the network was pretty pointless since the citizens have no idea how to navigate it.

Here's an example. When you have two services that connect more than once, citizens will often get off the first line, transfer to the second line, then transfer back to the line they were originally on. It doesn't matter if the line transferred to is a local bus that makes a bunch of stops before connecting to the next stop on the original line. here's an illustration:



Sometimes they'll even connect to 2 or 3 additional lines before getting back onto the line they were originally on. From watching the cims I have no idea how CO programmed the pathfinding algorithm. Often it honestly seems like the cims are taking random paths until they reach their destinations.

Is there any hope that these issues will be fixed? I want to play this game, but every time I do, I get let down as soon as my network gets big enough that there's more than one possible path between two points on the network. And the cims tend to pick one of the worst routes possible.

 

[unmerged(334932)]

You are not the first to have state these facts There are threads already posted about this, and the Devs said that the system works fine. I believe that the Devs are in denial, and can't be bothered to re-write code for the entire game. It is true that in CIM their path finding worked perfectly, but in CIM2 i can only guess that someone messed up.

 

[kensternation]

You are not the first to have state these facts There are threads already posted about this, and the Devs said that the system works fine. I believe that the Devs are in denial, and can't be bothered to re-write code for the entire game. It is true that in CIM their path finding worked perfectly, but in CIM2 i can only guess that someone messed up.

+1,000,000

They are absolutely in denial. They claim pathfinding has to do with average wait time. Instead of the more reliable following the schedule. Of course, it still does not explain why a CIM would get off a vehicle get on another line take it one stop and transfer back to the original line.

Having said that... No Xenon5 there is no hope. CO has categorically stated that pathfinding works perfectly and there is no plan to 'fix' it since it is not broken.

 

[Norfolk_Chris]

It's a pity that CO did spend the last four months resolving this problem (It does exist even if CO say that it doesn't; we've all seen it!), instead of wasting their time a stupid monorail DLC which is horrible and adds nothing useful to the game.

 

[jasiekzar]

Try tweaking the quality of service of the vehicles in the ruleset editor so that all of them are 100%. I did it in my cities, right now people don't leave metro to take tram at the same route or tram to take bus at the same route, their choice is METRO>TRAM>BUS.

 

[ggjono]

It's a pity that CO did spend the last four months resolving this problem (It does exist even if CO say that it doesn't; we've all seen it!), instead of wasting their time a stupid monorail DLC which is horrible and adds nothing useful to the game.

1+ i would have actually paid to have an upgrade of the path finding ....

 

[douglasrac]

Try tweaking the quality of service of the vehicles in the ruleset editor so that all of them are 100%. I did it in my cities, right now people don't leave metro to take tram at the same route or tram to take bus at the same route, their choice is METRO>TRAM>BUS.

This is a very interesting tip. I didn't think about it. Definitely will try it.

 

[unmerged(326364)]

Try tweaking the quality of service of the vehicles in the ruleset editor so that all of them are 100%. I did it in my cities, right now people don't leave metro to take tram at the same route or tram to take bus at the same route, their choice is METRO>TRAM>BUS.

This may help to prevent from changing vehicle types but not the strange pathfinding. I have the issue in a larg tram network where all rolling stock is all the same vehicle-type.

 

[kerts93]

Try tweaking the quality of service of the vehicles in the ruleset editor so that all of them are 100%. I did it in my cities, right now people don't leave metro to take tram at the same route or tram to take bus at the same route, their choice is METRO>TRAM>BUS.

I'll try that also

 

[unmerged(733534)]

1+ i would have actually paid to have an upgrade of the path finding ....

Agree with this!

Similar case like this

 

[quirinus]

Try tweaking the quality of service of the vehicles in the ruleset editor so that all of them are 100%. I did it in my cities, right now people don't leave metro to take tram at the same route or tram to take bus at the same route, their choice is METRO>TRAM>BUS.

i'll try this too.
if that works you are the hero of this forum

 

[jasiekzar]

i'll try this too.
if that works you are the hero of this forum

I hope it helps at least a bit. I use it along with careful placement of stops so that people change only at stops I want them to.

 

[ThrilledCim]

Pathfinding generally works well I find. The cims are able to find the best route that gets them from the starting point to their target BUT the limitation that I find is that cims are not able to travel back relative to their direction of travel to transfer onto another faster transit line even if it's right there. They would rather walk or take another transit line to get to their target. Somehow they are only able to look forward in their pathfinding task to find the best route but not able to look backward. For example, let's say the cim's direction of travel is north, if the target is in between two metro stops, the cim will get off at the metro stop south of the target and rather walk to the target instead of staying on the metro for one more stop to the one north of the target even though the stop north of the target could be right next to the target because the cim can't walk backwards as the target is behind that closer metro stop. For transfers, it's the same thing, the cim will only able to transfer onto the next transit line if the stop is in the direction of their travel and not the opposite direction even if it's farther. They will walk toward the next transfer point further up even if another transfer point is right there behind them the other way and it will take them where they want to go, but no, they will ignore it.

So in light of this, maybe if you place the 2nd metro stop before the tram stop in the direction of the cim's travel then the cim might be tempted to stay on the metro to the next metro stop. Just my guess.

 

[jasiekzar]

This doesn't explain why CIM would leave metro to take a few stops with tram and then go back to the same metro line.

 

[ThrilledCim]

This doesn't explain why CIM would leave metro to take a few stops with tram and then go back to the same metro line.

It could be that the 2nd metro stop is after the 2nd tram transfer stop so the cim got off the tram stop and transferred back onto the metro and continued on because the metro stop is after the tram stop. Try putting the 2nd metro stop before the 2nd tram stop to see if that makes a difference.

 

[The_13th_Ronin]

This needs to be acknowledged by CO.

Cities in Motion 2 is a good game but in a game where the entire objective is to move people from A to B the path-finding is a big deal.

 

[kensternation]

This needs to be acknowledged by CO.

Cities in Motion 2 is a good game but in a game where the entire objective is to move people from A to B the path-finding is a big deal.

Unfortunately according to CO pathfinding works perfectly.

 

[Lord Vetinari]

I did a bit of tests and I think I discovered the IMHO two major flaws of the pathfinding system.

1 - This is really big: the system doesn't seem to care about walking and waiting times. While it is really good at estimating the time it takes for the various vehicles to make each leg of a trip (which is good), it seems to assume that when a CIM changes line, he's going to be instantly on board of the new vehicle the moment he gets off the first one, which of course isn't true. The time it takes to walk from one stop to the other should be taken into account (or at least approximately calculate the distance between the two stops and use that value to choose the best transfer point) and, if making the AI aware of the actual timetables is too much of a problem, which I can understand, at least apply some form of fixed time penalty to each transfer, so that Cims are less likely to change line unless it's actually very beneficial or mandatory to reach their destination.

2 - The pathfinding seems to take in account comfort as a factor, which may be interesting for intercity transport, but makes little sense in a urban transport simulation.
It should be a parameter for customer satisfaction, not for pathfinding. Or at least, the very last parameter considered to discriminate between lines after everything else.

 

[Vadi]

I would disagree with 2 on you as that is a very subjective topic and I'd happen to lean the other way. I'll be fine in general on taking longer to get somewhere in comfort than shorter in discomfort is the trip times are up to ~25% longer.

 

[Lord Vetinari]

Well, it depends on what kind of commute we are talking about. If you have to spend 1 hour or more on a vehicle, I agree with you, comfort becomes important. If you have to make a short trip and/or a long trip made of multiple relatively short bits on different lines, I don't think you'd go out of your way to take the lines served by the most comfortable vehicles.
I guess though that having the AI figuring that out is pretty tricky, that's why I said that it shouldn't be a factor.