*[13.1]Train PF don't give penalty to circuit closed signal

[decay42]

I don't know if it's a bug or just wrong signaling, but I wanted to add it here to let you guys figure it out.

When a signal is blocked (turned red) by a circuit network condition and there is a different station with the same name available, a train still wants to enter the first station (which it does not try to, if a train is inside blocking the signal - then it will automatically pathfind the second station).

See my reddit post for further details: https://www.reddit.com/r/factorio/comme ... led_train/
If i didn't explain enough, please let me know.

Screenshots from the reddit post, in case you don't want to go there:
http://i.imgur.com/UQfP8PB.jpg
http://i.imgur.com/OpNNm7z.jpg

[Loewchen]

I was actually worried that something like this would happen.
The good news is the cause of this is documented, the bad news is the fix is now scheduled for .14

Could not find the original report, but its similar to this one: viewtopic.php?f=173&t=26409

[Twinsen]

For the moment, this is working as intended.

Setting the signal red will not add a penalty to the pathfinder, so the train will not try to find an alternative path.
I added a clarification to the tootip: Note: the train will very often stop at the signal and not try to find an alternative path even if one is available.

I think situations like this will always be problematic since after a certain distance, the train will not change it's path no matter what.

[decay42]

I just thought of a possible "solution" to this. Not exactly to this problem, but to the "task" intended by me originally.
Would it be easily possible to make train stations circuit network controlled (disabled/enabled) to achieve the wanted behavior?

So you wouldn't control the signal for the entrance to the station but the station itself, thus making the train chose another station right away.

[XKnight]

Train pathfinding algo ignores blocks that were locked using circuit network signal.
As a result, train prefer "shortest path + unknown waiting time" than "slightly longer path".

In my opinion, locked signal should have some extra value for pathfinding algo (maybe infinity value? or user-defined?) if it was locked by ciruit network.

[Nexela]

I agree, otherwise the only usefulness I can see for setting a signal to red is to make safer charcter railroad crossings.

Still lots of great train improvements in .13!

[MrGrim]

I can't figure out a way to request update notifications for a thread without replying so.. replying.

[Mazian]

I can't figure out a way to request update notifications for a thread without replying so.. replying.

Subscribe Topic link is at the very bottom of the page-- part of the theme. It's a yellow bar. Didn't stand out for me at first either.

[Twinsen]

That is planned in one way or another for 0.14

[Loewchen]

That is planned in one way or another for 0.14

I'll move this to Waiting for 0.14 then.

[ixnorp]

How costly is it in terms of processing to have a train re-evaluate its path or to add something like path finding weights for paths which will clearly never be available or the time to wait is unknown? Could trains re-evaluate their paths, say, once a second (every few seconds?) when stopped (even when moving?) to ensure they have the best path? Perhaps there could be a toggle or research item to enable advanced path finding only on certain trains which will be on complex networks and other trains would default to the cheaper default path finding which may not be as smart but is suitable for networks with only a few trains on them.

[Loewchen]

The PF does not see signals but blocks, so I could imagine that it considers the block as occupied and as the block is part of both routes both get the same penalty and the shortest route is considered fastest. More testing would be necessary to confirm that.

[XKnight]

The PF does not see signals but blocks, so I could imagine that it considers the block as occupied and as the block is part of both routes both get the same penalty and the shortest route is considered fastest. More testing would be necessary to confirm that.

Pic

[Loewchen]

Thank you for testing.
By placing a cargo wagon in the way instead of the circuit it seems that the PF does indeed not give penalties to via circuit network closed signals.

[DevilXD]

Umm, I found this topic in "waiting for 0.14" section: viewtopic.php?f=173&t=27513

[Loewchen]

Umm, I found this topic in "waiting for 0.14" section: viewtopic.php?f=173&t=27513

Dammit, I misunderstood the case in the other report completely, this is indeed the same "bug".

Merged both reports.

[Loewchen]

This is going to be fixed in 13.17:
viewtopic.php?f=30&t=31116

Reply here if it didn't.