RoRo vs Terminus (was Loop vs 2-headed train network)

[Grimakar]

I have a little problem with the Terminus style. I always plan my outposts with an extra shuttle station. The shuttle trains are always with one loc. That is why they need a RoRo station, otherwise they are stuck at stations and I cannot call them to where I want them. One solution could be, that I use 2 locs that are connected to each other in the opposing direction. Btw. never tried that, don't know if that works. But in that case the locs are slower and that really is an important thing to me with one loc, the quick travelling speed. Of course, I could only build RoRo at those shuttle stations and the rest goes with Terminus, but then other trains could be seduced to use that as stupid paths.

[golfmiketango]

I've wondered for a while now how hard it would be to implement a 'reverse to station' instruction.

Like so, if the train's next stop contained a 'reverse to station' instruction, the train would pathfind in reverse to reach that station.

It would then become possible to program the trains to automatically traverse a Wye formation to turn themselves around without needing a full loop.

This can be achieved with double headed trains simply by putting stations at the terminal ends of the Wye, oriented accordingly, and instructing the train to stop there. Potentially useful, for example, if you want to create a 1-12-4 train and have it "do the right thing".

[ratchetfreak]

I have a little problem with the Terminus style. I always plan my outposts with an extra shuttle station. The shuttle trains are always with one loc. That is why they need a RoRo station, otherwise they are stuck at stations and I cannot call them to where I want them. One solution could be, that I use 2 locs that are connected to each other in the opposing direction. Btw. never tried that, don't know if that works. But in that case the locs are slower and that really is an important thing to me with one loc, the quick travelling speed. Of course, I could only build RoRo at those shuttle stations and the rest goes with Terminus, but then other trains could be seduced to use that as stupid paths.

if you press R on the train you can rotate it right on the track. though that only works when you are there to rotate them.

[Grimakar]

if you press R on the train you can rotate it right on the track. though that only works when you are there to rotate them.

I know, but sometimes when I go somewhere else, maybe by car, tank or foot, it happens that I want to return by train. This happens quiet often, eg when fighting, clearing the woods or building something new and big or I am testing the rail signals of another train and want it to stay at an outpost for loading. Or in multiplayer, someone took my train, I need it back.
Many situations, where I just call the train, no matter where I am or where the train is.

[Dasani]

I don't play with mods that tinker with the game's functionality, I play the game as-is, so I can't comment on how certain mods might impact things. But i'm still not seeing the issue you're having with expandability. You just throw on more track to make an additional loop through an area that has more potential resource outposts. I tend to make these loops to cover at least 5-6 different locations and make that big loop another part of the main line. And again, if you're worried about travel time around the entire loop, you can always add additional rail to make shortcuts to bypass sections of it.

Personally, I don't see travel distance as a factor in any shape or form. 90% of the time, in the unmodded base game, your train is going to end up waiting to get filled up after it gets back to an outpost anyway. So if it takes the drills 5 minutes (just throwing out random numbers) to generate enough resources to completely fill a train, it doesn't matter if the train's travel time is 1 minute or 4 minutes, the bottleneck is the same. And in the event that you do generate more ore than that where travel time "might" become a factor, it still isn't because you can solve the same problem by slapping an additional train on the track. So your resource outpost generates iron (or whatever) so quickly that it's completely full up and ready to offload before the train gets back and it shuts down? Just add a second train. Again, travel distance doesn't matter because once it starts to become a factor, additional trains solve the issue. When you play on a map a really long time just for kicks, you end up with expansions in extremely long distances unless you're playing with mods or silly settings that give you enormous resource patches that are practically inexhaustible. The way this system is setup, you can throw an infinite number of trains on the track. All you have to do is make sure your offload track queue is long enough to hold them all (which is easy) and a slightly redesigned fork to lead into bases having a queue track waiting to get into the resource outpost long enough to hold the entire train while it waits without blocking the main line.

Travel time is never really a factor since more rails and more trains completely negate that. And if your latest expansion is so far away you believe the train wouldn't make it back before it stops producing, then you'd know that ahead of time before building the outpost, so you could easily plan in advance having the extra queue track length, etc.

In my latest game, the one after I made this video from, i've increased the size of all trains to 2-6-2 and made the unloading station appropriate to that as well. It's currently handling 27 trains and as usual, bottleneck is only how fast the resources are generated at the outposts. So even if those trains arrived quicker, wouldn't make a difference, and again if my outposts had magical mod equipment to generate resources 3-5 times faster, i'd just throw more trains at them. Wouldn't need to change the underlying design at all aside from the little "fork" before each expansion to serve as a queue track while they wait to be loaded up.


So far i've been able to expand to cover an enormous area of the map. Only requiring one big loop for each area cuts on the size the bases need to be and number of rails needed. I'm loving it.

[mrvn]

if you press R on the train you can rotate it right on the track. though that only works when you are there to rotate them.

I know, but sometimes when I go somewhere else, maybe by car, tank or foot, it happens that I want to return by train. This happens quiet often, eg when fighting, clearing the woods or building something new and big or I am testing the rail signals of another train and want it to stay at an outpost for loading. Or in multiplayer, someone took my train, I need it back.
Many situations, where I just call the train, no matter where I am or where the train is.

The solution is simple. If you call a train then you are right there to turn the loc around. And when you arrive somewhere with a terminal station then turn the loc around before you leave. In multiplayer games teach everyone else the same rule.

Note: You can make a home station for each player names after the player and have a train per player that has the home station in it's path. That way each palyer can easily find their own train in the train dialog. Plus you can send trains away so they don't block the station you leave the train.

[mrvn]

The way this system is setup, you can throw an infinite number of trains on the track. All you have to do is make sure your offload track queue is long enough to hold them all (which is easy) and a slightly redesigned fork to lead into bases having a queue track waiting to get into the resource outpost long enough to hold the entire train while it waits without blocking the main line.

Except that won't work with single track terminus stations. The first train will enter the station and then second train then blocks the main line. Adding a second train to an outpost means adding a waiting bay and then you have 90% of a RoRo station.

What is simpler is adding a second (third) car to the train. That also speeds up loading and unloading.

On the other hand dealing with double ended trains of different lengths adds some challenge to the refuling process. If you have terminus stations then most likely most trains will go between exactly 2 stations. But you can add a 3rd station that is a terminus and only has one refuling chest. Then every loop the train does it reverses direction and the other loc gets refueled.

[Dasani]

If your main line ever gets blocked, it's because you aren't doing it right.

http://i.imgur.com/t0ITMT2.jpg

This is the standard single train-per-outpost method I use. Most outposts only need a single train.

But if for, some crazy reason, you needed more than one, it's a very simple and cheap solution.

http://i.imgur.com/I8cdB3P.jpg

This is the same outpost expanded to allow 3 trains non-stop. One train already there loading up, and 2 more waiting in queue to load up. Mainline is never blocked. Only cost 20 extra rails and a couple of rail signals.

Main two advantages of this style over the other is it takes MUCH less space and it is 100% impossible for trains to pathfind down the wrong track by accident because they are all dead ends except for the mainline. You will never run into a situation where a train decides to go down another track to "save time" and ends up getting stuck before another train that isn't moving anytime soon.

Edit: Was going to post the images in closed spoilers but for some reason it wasn't working, so I just provided the links instead.

[ribsngibs]

I use drive-through stations:

OpenTTD Drive-Through

Drive_thru.png

Together with a standardized modular factory grid, somewhat like this sketch:

Grid

Factory Layout.png

A Grid module consists of 9 Roboports (orange squares) placed at max distance from one another.

Interesting base layout - I'd love to see screenshots to see how you're putting the meat of the base (the belts/bots/assemblers or whatever) in between the stations. My big train base also has a big grid, but instead of just having drive-through stations that are just offset from the main lines like yours, I have probably much larger grid squares and put full stacker stations inside them. And the rail is 4 lane (2x2).

Train Grid looks like

Album looks like this

Zoomable map here


To the original question - I like terminus stations because I think the stations look better that way, no other real reason, and my trains don't care about order (all wagons carry identical cargo). I do have roro's for my resupply/personal access stations because the order of the those wagons matters and personal access trains are just a single locomotive. The terminus stations have 4 waiting bays, plus a 5th track for the outgoing train, so the roro resupply/PAX lane is next to the last incoming waiting bay and loops around to join the outgoing lane, so the roro loop for the resupply train ends up not actually increasing the footprint of the station that much. (You can see the loop in the zoomable map quite clearly).

[MeduSalem]

Oh, then I am sure you know Brasilia. I am really a fan of planned cities, esp that one. I even tried to rebuild the idea of the street network of Brasilia in Cities: Skylines. Actually in Factorio this would be a bus system.

Yeah I know Brasilia... Basically an Urban Planer's and Architect's dream. Though nobody who's a bit into humanism would ever do a project in that size ever again because it turned out quite a disaster in some socio-political and economical aspects. Maybe in dictorial governed states... or states like the Emirates or something where they are treating their workers like slaves all while the crème de la crème of the Rich and Bold surround the yearly Formula 1 event. As an responsible Engineer I'd refuse to accept taking on megalomaniac projects like that which are carried out on the backs of poor people.

I never played Cities:Skylines though... I stopped playing city-building games when everyone transitioned away from grid-based planning and axonometrical projections to full 3D perspectives. Most of the games ever since lost their "touch" somehow that Sprite graphic based games used to have.

I would be pleased, if you posted some screenshots of your designs in Factorio.

Interesting base layout - I'd love to see screenshots to see how you're putting the meat of the base (the belts/bots/assemblers or whatever) in between the stations.

Sorry guys that I didn't look into the thread for a while. Also I'm sorry to say that I didn't make any big screen shots of my "old" big base and I don't have that anymore because it somehow got corrupted during a failed Cloud Sync during power out.

So I started a new one a while ago based on some of the improvements I mentioned earlier in the thread, though it's not nearly as big yet and doesn't really put the train system to it's use yet because I'm still transitioning away from my "starting mess", so all that is currently done by train is basically ore delivery right now:

Map

Map.png (209.07 KiB) Viewed 6985 times

Train Station, Smelting and Storage

Station, Smelter, Storage.jpg (414.43 KiB) Viewed 6985 times

I'm probably going to scrap the map again anyways once 0.15 is out... because I already dislike certain things I did there. Also don't wonder, I accidently started to play on peaceful mode... and then didn't feel like starting over after I was through the initial setup of automating research. Also one of the reasons I'd eventually start over again because peaceful mode is boring on long-term... But I can't change it without triggering the cheat warning.

The base is entirely Steam Power driven. That's why there are no Solar fields.

[Optera]

I'm probably going to scrap the map again anyways once 0.15 is out... because I already dislike certain things I did there. Also don't wonder, I accidently started to play on peaceful mode... and then didn't feel like starting over after I was through the initial setup of automating research. Also one of the reasons I'd eventually start over again because peaceful mode is boring on long-term... But I can't change it without triggering the cheat warning.

The base is entirely Steam Power driven. That's why there are no Solar fields.

You can use mods like Space Book and Creative Mode to switch between peaceful and aggro mode without cheat trigger.

[MeduSalem]

You can use mods like Space Book and Creative Mode to switch between peaceful and aggro mode without cheat trigger.

Oh I didn't know that... I thought the cheat trigger was hardcoded and can be overcome as the mods themselves somehow disable achievements as they basically count as a cheat.

Well let's see if its worth it... because maybe I have to restart anyways with 0.15 because of the many changes that come with it... especially Terrain generation stuff once more as I find the sudden changes in Terrain ugly if a patch fiddles with the algorithm.

[Deadly-Bagel]

Mods disable Steam achievements and use a separate set of 'modded achievements' that don't count to your vanilla achievements. Only really a deal if you're playing through Steam.

[Optera]

Mods disable Steam achievements and use a separate set of 'modded achievements' that don't count to your vanilla achievements. Only really a deal if you're playing through Steam.

Mods do while they are active. As soon as you deactivate them on a modded save achievements trigger normally.

I built a heavily modded base in peaceful mode, switched biters to aggro with space book, removed all mods, loaded the save into steam and got nearly all achievements instantly.

[MeduSalem]

So I decided that I'll probably scrap my current map anyways in about 2-3 months when 0.15 will be released due to the amount of changes... so I'll give it a fresh start then...

... that said I have started to "play around" a little bit since I'm past launching the rocket anways.

I have looked into some suggestions from the thread here and I have found a pretty interesting "evolution" to my crossings, though it's efficiency is up for debate:

Here's an updated map view:

Map

2016.12.13 Map.jpg (221.14 KiB) Viewed 6795 times

Here's is what the crossings look like now:

Crossing and Railway Station

2016.12.13 Crossing.jpg (798.84 KiB) Viewed 6795 times

Yeah I get that the rails are quite far apart now... but now I can make a U-turn directly before crossings/after stations without having to go into the crossing!

Probably there could be some more chain signals to "free" the blocks sooner improving some efficiency of the crossing.

[mrvn]

Is that really more efficient that a simple turnaround? I imagine the shear size will cause long blocking times for the crossing whenever a train needs to cross anothers path.

PS: I wish one could tell the route planer to not go 3/4 around a turn around. Or more generally tell each entrance what exits it may take.

[MeduSalem]

Is that really more efficient that a simple turnaround? I imagine the shear size will cause long blocking times for the crossing whenever a train needs to cross anothers path.

Well I could squeeze it a litte bit more tighter by moving the U-turns 3-4 tiles closer to the intersection, moving the left-turns (diagonals) more central. Also with more chain signals (directly before and after two tracks intersect) I could also squeeze out a few more % efficiency because then a train will occupy a block only as long as absolutely necessary, entering it as late and leaving as soon as possible.

I don't know if it is more efficient in general though because with more routes the possibility for detours also rises, but from what I noticed is that traffic jams are almost avoidable due to how a train entering/leaving a station doesn't really cause any traffic jam further up the track. But I didn't ramp up the traffic much yet (though I could for testing purposes).

At least it is a lot more efficient than a classic roundabout-setup. Comparison:

1. Classic Roundabout (taking Mehve's picture from another thread):
   
   • 4 simultaneous right-turning trains.
   • 2 simultaneous straight-through trains.
   • 1 straight-through train AND 2 right-turning trains.
   • 1 U-turning train AND 2 right-turning trains.
   • 1 right turning train AND 1 left-turning.
   An average of 2.8 trains interacting with the intersection.
   
   Roundabout

   

   HxmpXT7.jpg (47.67 KiB) Viewed 6750 times

2. Classic 4-way intersection:
   
   • 4 simultaneous right-turning trains.
   • 2 simultaneous straight-through trains.
   • 1 straight-through train AND 2 right-turning trains.
   • 2 right-turning trains AND 2 left-turning trains.
   U-turning not possible. But an average of 3.25 trains to interact with the intersection.
   
   4-way Intersection

   

   Crossing.jpg (166.88 KiB) Viewed 6750 times

3. My new Crossing:
   
   • 4 simultaneous U-turning trains.
   • 4 simultaneous right-turning trains..
   • 2 simultaneous straight-through trains AND 2 U-turning trains.
   • 1 straight-through train AND 2 right-turning trains AND 1 U-turning train.
   • 2 right-turning trains AND 2 left-turning trains.
   • 2 U-turning and 1 left-turning AND 1 right-turning train.
   So it's possible for an average of 4 trains to interact with the intersection.
   
   Roundabout

   

   2016.12.13 Crossing A.jpg (416.72 KiB) Viewed 6750 times

Hopefully I didn't forget any possible combination.

PS: I wish one could tell the route planer to not go 3/4 around a turn around. Or more generally tell each entrance what exits it may take.

I think that the route planer has a lot of potential for additional improvements... maybe via additional signals or better circuit network features, who knows.

[ChurchOrganist]

Is that really more efficient that a simple turnaround?

Watch MadZuri explain why loops are a bad thing for efficiency.....
https://youtu.be/1IWgP33H3mc?t=2m30s

[MeduSalem]

Watch MadZuri explain why loops are a bad thing for efficiency.....
https://youtu.be/1IWgP33H3mc?t=2m30s

The A* algorithm is already fast for what it does (though I'd say it is not be the best path-finding algorithm for each path-finding problem), so the amount of processing time wasted to do a path-search is negligible with the few trains that are actively going somewhere (even if people have a 100 trains not all are trying to go everywhere at once).

In theory the A* algorithm would still choose the shortest (same path) in the roundabout/loop system as it would in a intersection-only system, because the shortest route has the lowest cost.

But what may happen is that the algorithm will choose a path that is not the shortest depending on how many blocks are occupied by other trains on the rail network (= increasing the cost of the shortest route)... which then may cause the train to choose an alternative route to prevent having to slow down and wait until blocks become free again. So you eventually have to live with trains taking detours to distribute the load all over the train network.

So in the end intersection-only systems may force the train to always go the direct route, but since they have no alternatives to choose from they cause MAJOR traffic jams around intersections that become a bottleneck if many trains want to go down the same path with no way around that. So there will be some trains waiting for one another at intersections, maybe even deadlocking the entire system. And not even multiple parallel tracks for each direction will help at this point.

In the end both systems fail when it comes to high throughput, there is no inherently better system. It's a matter of "alleviating" the side-effects or which problems you find it easier to live with.



Personally I'd prefer loop-based over intersection-only because I have looked a little bit into processor technology and there is a reason why many-core designs use (hyper-)loop-based interconnections between cores... because it would be a nightmare to route 128 cores over a single crossbar.



Maybe some of the problems would be better solve-able if there was at least some kind of tunnel/bridge (even if they have a long slope) so that you don't have to build at-grade-intersections.

[mrvn]

So in the end intersection-only systems may force the train to always go the direct route, but since they have no alternatives to choose from they cause MAJOR traffic jams around intersections that become a bottleneck if many trains want to go down the same path with no way around that. So there will be some trains waiting for one another at intersections, maybe even deadlocking the entire system. And not even multiple parallel tracks for each direction will help at this point.

Having intersections doesn't mean you don't have alternative ways. Just consider a simple grid with intersections like the one posted at the start. There are many alternatives ways only dependent on the number of grid cells.

One thing I'm thinking about is actually if those detours are harmful. Does a detour always include paths across a turn-around or intersection that blocks a lot of other tracks? Like going 3/4 around a turn around. That's why I mentions I want to limited the ways a turnaround can be used. Block the 3/4 way so it doesn't choose bad paths.