Some toughts about LTN

[ZoeF]

As I have been trying to get the LTN mod to work for the past few days I am wondering if their is a design flaw in it. If someone needs to study a book or even watch 20 hours of youtube videos to get the basics down there is something fundementaly wrong with the mod. Wich should make the game easyer and more managable not harder.

Some small toughts:
- The combinator is a nice idee but take some time to maybe make a custom click menu on the station itself.
- LTN Helper is a nice concept for just that.
- Now you need to set stack limits - ammount of items - .... wich makes it cluttered and to be frank most "usecases" are pick a full load and drop a full load.
- It happens alot that a train leaves a station with a small ammount of items stil in the train. Allow some setting of altering the way trains work. Now it is either incative or time passed when a train leaves a station. Why not allow it to be empty?
- All the logic needed to make it working properly is taking away from the enjoyability of the mod. I for one am happy to have saved my game before i started meddeling with this. So I can atleast continue playing without having to constantly go and look why anotger train is in the depot with a load of goods.

Just my two cents?

If anyone is able to assist me to setup a small game with this mod in a working order I will be happy to recieve the help.
Discord ZoeF#3764

[jodokus31]

LTN is very powerful and does a very good Job with the possibilites the modding API offers, but it's also a bit complicated and not very beginner friendly. That is on purpose to keep it efficient.

If you cant wrap your head around it, better leave it alone.

[Optera]

Running full trains from A to B is simpler than ever with Factorio 1.1.
LTN can do that and then some. It all depends on setting the control signals correctly.

Mostly empty trains sound like too low thresholds.
Even the most basic setup should set thresholds per station:
- Provide threshold should be number of trains * train capacity or buffer capacity whichever is lower.
- Request threshold should be train capacity.

There's a lot more control signals you can find in the manual, some showcased in the Demo as well.

[ZoeF]

@Jodulko: easy answer, if you don't understand don't use it?

Normal procedure of anything created (games or real life appliances). Easy to start but hard to master. Any game or tool that is supper complex does not get used by the average person. Imagen a smartphone that would require you to setup the bios to even start using it ... .

Also their is no simple example of the LTN system anywhere. The example files are so complex with 3 arit and some combinators and then some. Make something easy to understand and people will start using it. Make something hard to understand and people will want it but not use it.

[jodokus31]

You should consider, that this mod does not need to gain money on a mainstream market. Also, it's quite popular and actually used by a lot of people, who propably have some technical background.
It's not soo complicated, if you have some experience with circuit network and combinators.

[Optera]

1) My Station designs need 1 or 2 combinators not 3
2) If you consider 3 combinators complex neither LTN, TSM or Smarter Trains are for you. Stick with vanilla and manual schedules.

[ZoeF]

Last reply I will give on this matter.

1) If you need just one combinator to make the system work properly, there is a initial flaw in the system.
2) 3 combinators is not complex per se. But it is if you are trying to figure out why a system is not functioning. You forget you have to think about the new concept of requester and provider and negative values and length of trains and load balancing and .... .
3) The overal getting started with LTN is to complex to get it running.
4) I don't like arguing with people who make it personal.

The one thing that i find missing the most is the inabbility to make a train wait at a requester station until it is fully unloaded. And you can say that it is a mistake on my end for not deciding the exact ammount correctly or what not. But that could be overcome by just having one option (that is in vanilla factorio)

@optera: one of your first designs in your topic for blueprints is one to clean out residual items in a depot. So even you noticed a flaw in the system and needed a fix for it. A simple -> wait in station until empty should fix this and would make your first blueprint not necesary.

[Optera]

@optera: one of your first designs in your topic for blueprints is one to clean out residual items in a depot. So even you noticed a flaw in the system and needed a fix for it. A simple -> wait in station until empty should fix this and would make your first blueprint not necesary.

That design is a 2nd safety net recommended for new players worried they might mess up their loading and unloading system configuration.
I myself never had the need to use it.

[ptx0]

did you even look at the map settings for LTN to see that it has options to tweak the experience?

you can disable the 'finish loading' so that it fully empties train.

tons of people enjoy and use LTN, so "people won't use it" is patently (and hilariously) false.

[jodokus31]

Let's be clear: LTN is nearly perfect, sometimes even too much, bc once setup, you dont have much to think about.
Im not sure how it could get more beginner friendly, i fear you have to bite the bullet and watch those 20h

[Adrenalex]

Running full trains from A to B is simpler than ever with Factorio 1.1.
LTN can do that and then some. It all depends on setting the control signals correctly.

Mostly empty trains sound like too low thresholds.
Even the most basic setup should set thresholds per station:
- Provide threshold should be number of trains * train capacity or buffer capacity whichever is lower.
- Request threshold should be train capacity.

There's a lot more control signals you can find in the manual, some showcased in the Demo as well.

I am running into an issue that I'm not quite sure how to solve with LTN. I've used LTN for some 100 odd hours in the factory and the size of the factory is significantly larger than most. LTN seems to prioritize sending trains to the largest provided resource station rather than the closest. It also seems to prioritize the train closest to the requester station rather than the provider station. So I end up in these situations where I have a train travelling 10 minutes across the map to get iron rather than the station a minute away. However, there's also a factory near the "10 minute" station because everything is modular. So trains basically just get sent all over the place rather than the nearest available resource.

I have each factory module compartmentalized into its own network ID so I don't have problems with intermediate products being shipped all over the map. However, I'd still like for every provided raw resource station such as copper, iron, coal, etc to be available for any requester station. I'd just like them to be more efficient for which station they go to. I can't use the provider priority because then all the factory modules will start using that station whether it's close or not because of how large the factory is.

I know Vanilla will send a train to the closest station of the same name but I don't want to go back to Vanilla because LTN is such an elegant solution for train efficiency. I just can't figure out how to use it a little smarter.

Any ideas?

[Optera]

LTN seems to prioritize sending trains to the largest provided resource station rather than the closest.
It also seems to prioritize the train closest to the requester station rather than the provider station.

That's 50% correct.
When a Request Threshold is passed LTN will look for Providers on the same surface, in the same NetworkID, with matching Train length.
Out of those it picks the one with highest priority then highest item count.

After finding a provider it looks for trains, again on same surface and in same Network.
Out of those it picks the smallest train capable of hauling the whole request, or largest train if no train capable of hauling all at once is found.
Only if multiple trains are available will it pick the one closest to the Provider.

So I end up in these situations where I have a train travelling 10 minutes across the map to get iron rather than the station a minute away. However, there's also a factory near the "10 minute" station because everything is modular. So trains basically just get sent all over the place rather than the nearest available resource.

I have each factory module compartmentalized into its own network ID so I don't have problems with intermediate products being shipped all over the map. However, I'd still like for every provided raw resource station such as copper, iron, coal, etc to be available for any requester station. I'd just like them to be more efficient for which station they go to. I can't use the provider priority because then all the factory modules will start using that station whether it's close or not because of how large the factory is.

I know Vanilla will send a train to the closest station of the same name but I don't want to go back to Vanilla because LTN is such an elegant solution for train efficiency. I just can't figure out how to use it a little smarter.

Any ideas?

Splitting each product into each own network removes reuse of trains, one of the best UPS saving LTN has to offer.

Instead build depots for your trains in all cardinal directions on the outskirts with separate Network ID for each direction.
Set all mines to matching IDs and Furnaces to allow all Networks.
e.g. Depot and Mines: N=1, E=2, S=4, W=8; Furnaces 15.

The whole inner network could be 16 so large ore trains don't traffic through the busier inner factory. Alternatively the same can also be achieved with different train length limits.

Fluid and item trains can coexist in the same network and depots, so those 5 Networks generally are all even a 10k spm megabase needs.

[Adrenalex]

Ah, that's a very neat solution. I appreciate the time it took you to respond and your detailed explanation!

[ptx0]

Splitting each product into each own network removes reuse of trains, one of the best UPS saving LTN has to offer.

reusing trains is one of the reasons that LTN is so rough on UPS, why do you say it is the other way around?

instead of having a train that's full of materials and sitting idle, you have extra trips to/from a depot - on-demand loading requires more optimisation than lazy loading, where already-full trains sit at a stacker full of materials.

[Optera]

Traveling trains have only little higher ups impact than trains sitting idle.

As for LTNs impact.
Over 90% is spent on reading circuit signals, the actual loop for selecting trains is only going through cached values. The difference between going through 10 or 100 trains should not even be noticeable.

[ptx0]

Traveling trains have only little higher ups impact than trains sitting idle.

this has been tested and shown to be false by mulark:
- https://mulark.github.io/tests/test-000 ... 00103.html
- https://mulark.github.io/tests/test-000 ... 00104.html

here you can see that collision checks on trains only occur while moving: https://mulark.github.io/tests/test-000 ... 00026.html

an accelerating train uses much more UPS than a constantly moving train, but nuclear fuel has the shortest acceleration time, shortest trip time, lowest overall UPS consumption: https://mulark.github.io/tests/test-000 ... 00106.html

different wait conditions have different performance penalty, and LTN's item count is heavy: https://mulark.github.io/tests/test-000 ... 00019.html (and revisited in .17 https://mulark.github.io/tests/test-000 ... 00047.html)

so I eagerly await your data, because everything here is showing that:

- the extra trip to/from the depot is a UPS sink (on top of its on_tick loop) due to the extra time the train is out on the rail, the increased number of acceleration events, the inefficiency of how trains are selected from depot, etc
- the common use case of LTN for a large number of on-demand small trains is killing base performance
- large number of trains are fine so long as they are idle at a normal signal, or a station

[Optera]

Seems like that changed since I ran my tests.
Back when I ran my tests, stationary trains took a lot of entity update.

[NotRexButCaesar]

Stick with vanilla and manual schedules.

Or use vanilla and 3 combinators for automatic load balanced schedules.