4-lane RHD intersection with optimized left turn [images+blueprints]

Smart setups of railway stations, intelligent routing, solutions to complex train-routing problems.
Please provide - only if it makes sense of course - a blueprint of your creation.
Post Reply
AlienRaven
Inserter
Inserter
Posts: 45
Joined: Wed Apr 26, 2017 7:48 pm
Contact:

4-lane RHD intersection with optimized left turn [images+blueprints]

Post by AlienRaven »

Backstory
Usually, when new factorio major version comes out, I start a new map from scratch. Doing so I try to consider mistakes and poor designs I made on the previous map. With 0.17 releasing next week I was thinking what I can improve on my next playthrough and one of the things. that came to me mind was my railway setup.

Previously I always used two-lane setups and this time I want to go four-lane. I browsed through this forum and reddit, however most 4-lane setups I saw were either too big or too throughput limited. Since I didn't have much to do on a previous map, I decided to try and create a junction, that would satisfy my needs:
  • 1L3C trains
  • Singleheaded
  • Right hand drive
Edit Notes
I slightly updated the post thanks to theolderbeholder's comments. New pics and bp attached.
Idea
The idea comes from traffic optimization strategy, that forbids direct left turn at crossing, instead forcing you to turn right and make a U-turn, if you want to go left.
Inspiration
Result
This is the result of several iterations, I tried to find something similar here on forum or on reddit, but all designs I saw were focused on the ability to immidiately go in all three directions. This is why I decided to make this post and share the results.

This intersection is optimized for 1L3C trains, however, technically you can expand it to any train size by extending spacing between signals.
Preview
You can preview the blueprint in better detail here: http://fbpviewer.trakos.pl/b/y1zwK1Z4th ... bL0LuQVSgo.


Testing
I created a separate rail network with 4 U-turns added at the end of rails and 8 stations, 2 jn each end, one on inner circle and 1 on the outer. After that I populated to crossing with 35 trains, running in different patterns, taking turns and switching lanes. No deadlocks for and hour of running.
Testing process
Community
Since I never really used four-lane I'd like to hear your criticism and feedback, is there anything, that might be improved or taken into account? Thanks in advance.
Last edited by AlienRaven on Sun Feb 24, 2019 4:02 pm, edited 3 times in total.

User avatar
CDarklock
Filter Inserter
Filter Inserter
Posts: 342
Joined: Sun Dec 03, 2017 2:17 am
Contact:

Re: 4-lane RHD intersection with optimized left turn [images+blueprints]

Post by CDarklock »

AlienRaven wrote:
Sun Feb 24, 2019 12:35 pm
The idea comes from traffic optimization strategy, that forbids direct left turn at crossing, instead forcing you to turn right and make a U-turn, if you want to go left.
I'm a complete idiot about trains, so forgive me if I'm... well... an idiot.

How does this optimise anything? What problem is it supposed to solve, and how does it solve it?

I can see - to some extent - that focusing on "I want the train to be able to go any of these three directions" might be myopic and ignore better solutions. But how is this solution better?

AlienRaven
Inserter
Inserter
Posts: 45
Joined: Wed Apr 26, 2017 7:48 pm
Contact:

Re: 4-lane RHD intersection with optimized left turn [images+blueprints]

Post by AlienRaven »

CDarklock wrote:
Sun Feb 24, 2019 12:45 pm
How does this optimise anything? What problem is it supposed to solve, and how does it solve it?
Thanks for asking. One of the reasons I wanted to post it was to check if it even makes sence with the rest of the community :)

From what I can see in the majority of blueprints (for example here: viewtopic.php?f=194&t=46855) left turn on junktions in most cases causes the waiting time of trains, that want to go through the junktion grow, due to left turn is the thing, that creates most of rails intersections, and as a result rail segments (and signals required). This later forces train to stop or slow down when the other train goes left.

The idea behind this junktion is to let trains that go straight or right exit the junktion as soon as possible with minimum possible braking or stopping needed, which causes higher throughput, which means there will always be less trains in the junktion, that leads to decreasing the time its needed for a train to make a left turn as well.

Also a small bonus, compared to compact designs is that it allows U-turns :)

User avatar
CDarklock
Filter Inserter
Filter Inserter
Posts: 342
Joined: Sun Dec 03, 2017 2:17 am
Contact:

Re: 4-lane RHD intersection with optimized left turn [images+blueprints]

Post by CDarklock »

AlienRaven wrote:
Sun Feb 24, 2019 1:03 pm
CDarklock wrote:
Sun Feb 24, 2019 12:45 pm
How does this optimise anything? What problem is it supposed to solve, and how does it solve it?
Thanks for asking. One of the reasons I wanted to post it was to check if it even makes sence with the rest of the community :)
My vote isn't really relevant, since I've built three trains in my life, only two of them worked, and none of them had any intersections. I have enough trouble with straight rails. ;)
left turn on junktions in most cases causes the waiting time of trains
Thinking about this just as a problem in graph theory, I would say the slowdown is because a left turn necessarily crosses more tracks.

Take this with a pound of salt, because I don't really know what I'm talking about. It's just similar to other things I've done.

In a four lane system, you get this scenario with any turn:

- The train blocks the current track
- Every track it crosses must stop to clear the track for the turn

If all tracks are bidirectional, then any turn from any track blocks 1 to 4 tracks arbitrarily. If all tracks are unidirectional, however - split evenly in both directions - then one direction necessarily blocks three or four tracks while the other necessarily blocks one or two. It depends on which side your tracks in each direction are. If you have American tracks where trains drive on the right hand side, a left turn crosses more tracks. If you have British tracks where trains drive on the left hand side, a right turn crosses more tracks. So instead of "left turn," let's say "big turn" because depending on your lane structure it might not be to the left. This is just me being pedantic.

But it's clear that "big turn" always blocks more lanes than "little turn," and it's necessarily a slowdown.

So let's imagine that we are going to replace "big turn" with "little turn plus U-turn." We'll ignore the tracks in the perpendicular direction, for now. In order to make the initial "little turn," you are necessarily blocking one or two lanes. And in order to make the U-turn, you are necessarily blocking four lanes because you have to cross all of them.

But the result is dependent on where you start. If you are in the inner lane, then your big turn blocks three lanes and your little turn blocks two. If you are in the outer lane, your big turn blocks four lanes and your little turn blocks one.

So from the inner lane, little turn plus U-turn blocks six lanes, where big turn would block three. And from the outer lane, little turn plus U-turn blocks five lanes, where big turn would block four. This is a worse result in both cases.

And this assumes an atomic action, which we probably don't have. In all likelihood, little turn plus U-turn involves stopping to let the tracks clear before crossing the four tracks. That means a train which would NOT stop now HAS stopped, unless - and this is rather important, I think - making the big turn would not have stopped other trains anyway.

I expect a lot of this to be corrected, since - again - I do not really know what I am talking about.

theolderbeholder
Fast Inserter
Fast Inserter
Posts: 135
Joined: Wed Sep 20, 2017 5:45 pm
Contact:

Re: 4-lane RHD intersection with optimized left turn [images+blueprints]

Post by theolderbeholder »

Wouldn´t it make sense to add a bypass to the first 2 to 1 junction? You are kind of lacking an uninterrupted "through lane".

AlienRaven
Inserter
Inserter
Posts: 45
Joined: Wed Apr 26, 2017 7:48 pm
Contact:

Re: 4-lane RHD intersection with optimized left turn [images+blueprints]

Post by AlienRaven »

Good point about size of turns. I come from a country with right lane car driving, so its easier for me to discuss it as RHD fr trains and left turns :) A also see your logic there in blocked lanes counting.

I believe, when it comes to trains, there are also couple Factorio limitations to consider. Let me try to explain in the example of this crossing, designed for right hand driving, so I'll reference "left turns" if you allow :)

Here are two junctions:
Pic
Left square doesn't have left turn and essentially allows trains to pass simultaneously as soon as they came in parallel lanes due to chain signals.

Right one has left turn, however due to limitations of signal placing, it can only come as a single segment, meaning that it will block train from any direction, if one is in the junktion, even if technically they will not meet (one goes up and one goes down).

By freeing this square of turning lanes we deblock parallel lanes increasing throughput.

As for the right turn, the train that goes left will ofcourse block the lane, while there, but since there are no rail intersections, I believe it will leave the segment faster, unless of course there are 100 trains over intersection, occupying each next segment all the time.

One of the thoughts I had was to remove left turn completely. My base mostly comes in production squares, surrounded by rails, so instead of left turn the train would make 3 right turns. however i think I was able to squeeze the roundabput nicely to avoid this.

I also specifically made most of the segments being able to host 1L3C train to ensure that there's always space to move to until train reaches just straight rails at the intersection end.

Hope this makes sence :)

AlienRaven
Inserter
Inserter
Posts: 45
Joined: Wed Apr 26, 2017 7:48 pm
Contact:

Re: 4-lane RHD intersection with optimized left turn [images+blueprints]

Post by AlienRaven »

theolderbeholder wrote:
Sun Feb 24, 2019 1:49 pm
Wouldn´t it make sense to add a bypass to the first 2 to 1 junction? You are kind of lacking an uninterrupted "through lane".
I was actually thinking to make it two lanes without a connection with lane switching done earier on direct tracks. Good point, thanks for spotting.

User avatar
CDarklock
Filter Inserter
Filter Inserter
Posts: 342
Joined: Sun Dec 03, 2017 2:17 am
Contact:

Re: 4-lane RHD intersection with optimized left turn [images+blueprints]

Post by CDarklock »

AlienRaven wrote:
Sun Feb 24, 2019 2:10 pm
Good point about size of turns. I come from a country with right lane car driving, so its easier for me to discuss it as RHD fr trains and left turns :) A also see your logic there in blocked lanes counting.
Since Factorio wants your signals on the right side of the track, using right-lane structures also makes it easier to see which signals are on which track. I was just being weird and theoretical while stepping into directional weighted graph territory. Which I mostly understand from "sector warp" pathing in space games, where the idea of left and right and orthogonality are meaningless anyway.
I believe, when it comes to trains, there are also couple Factorio limitations to consider. Let me try to explain in the example of this crossing, designed for right hand driving, so I'll reference "left turns" if you allow :)
I'm not the boss of you, as long as I know what you mean it's all good ;)
Right one has left turn, however due to limitations of signal placing
This is where I'm likely to get confused, I don't entirely understand Factorio's signal placement rules. I've been playing around with it a little, watching how it colours the track segments, but I haven't sussed everything out yet. This "chain signal" thing seems deceptively complex.
it can only come as a single segment, meaning that it will block train from any direction, if one is in the junktion, even if technically they will not meet (one goes up and one goes down).
Well, that's because of the size, isn't it? I mean you're cramming the track so close together, you don't have room for multiple trains. Like think about highways: nobody would make this in a highway interchange. They'd make a cloverleaf, because maximal throughput in this flow diagram is garbage. The first thing you'd do to try and resolve it is split off a "right turn only" lane in each direction. Then you'd make a loop after the intersection to connect in the other direction.

And that loop is really your "right turn plus U-turn," when you think about it. So it seems like it's a false economy to try and make your right turn also head into your U-turn, when real-world analogues separate them.
By freeing this square of turning lanes we deblock parallel lanes increasing throughput.
Yeah, that. :D
One of the thoughts I had was to remove left turn completely. My base mostly comes in production squares, surrounded by rails, so instead of left turn the train would make 3 right turns.
Reminds me of living in DC, technically the beltway is just a big circle so if you don't care how far you drive it doesn't matter which direction you go on it. ;)

But that's kind of how the cloverleaf works, too. Three rights make a left.
I also specifically made most of the segments being able to host 1L3C train to ensure that there's always space to move to until train reaches just straight rails at the intersection end.
So when you start your U-turn, doesn't that mean you're likely to stop in it while the other trains pass?

I mean, the design you're showing appears (to my admittedly-inexpert eye) to be something that will work in the sense that it doesn't break. I'm just not sure it achieves the goal of improving things. And digging into that is not about being critical of the idea, but more about understanding your thought process as someone who knows a lot more about this train stuff than I do.

I mean, I want to build big expansive train setups, I just don't know how. ;)

AlienRaven
Inserter
Inserter
Posts: 45
Joined: Wed Apr 26, 2017 7:48 pm
Contact:

Re: 4-lane RHD intersection with optimized left turn [images+blueprints]

Post by AlienRaven »

CDarklock wrote:
Sun Feb 24, 2019 3:09 pm
This is where I'm likely to get confused, I don't entirely understand Factorio's signal placement rules. I've been playing around with it a little, watching how it colours the track segments, but I haven't sussed everything out yet. This "chain signal" thing seems deceptively complex.
Check out this deck: https://drive.google.com/open?id=0B22HA ... Fc5S0wzYjA Its old, but gold in terms of explanation and things haven't changed since then. It actually helped me a lot in understanding of how rails work in this game. You might also want to check in-game mini-tutorials (there's an icon in upper right game menu, called tutorails (with a graduation hat as an icon)). Launching one of those will pause your game and it also gives good portion of information about how it works with hands on experience. After completion it will return you back to your game.
CDarklock wrote:
Sun Feb 24, 2019 3:09 pm
Well, that's because of the size, isn't it? I mean you're cramming the track so close together, you don't have room for multiple trains. Like think about highways: nobody would make this in a highway interchange.
If trains will go say from north to east and from south to west they will be able to pass through. Its about space for signals. If one could squeeze in signals and break this into multiple segments it would help. Otherwise game considers this a one segment and it wont let other trains enter while its occupied.
CDarklock wrote:
Sun Feb 24, 2019 3:09 pm
They'd make a cloverleaf, because maximal throughput in this flow diagram is garbage. The first thing you'd do to try and resolve it is split off a "right turn only" lane in each direction. Then you'd make a loop after the intersection to connect in the other direction.

And that loop is really your "right turn plus U-turn," when you think about it. So it seems like it's a false economy to try and make your right turn also head into your U-turn, when real-world analogues separate them.
Cloverleaf might work, but here it will have same problems, as in real life - connecting traffic on exit\entry from the leafs. Here's a topic on the forum with various intersection designs, even tested for throughput: viewtopic.php?f=194&t=46855 In this case cloverleaf can get jammed by trains, as it only fits 1 train per leaf and if more are travelling the throughput will suffer.
Cloverleaf junction
CDarklock wrote:
Sun Feb 24, 2019 3:09 pm
Reminds me of living in DC, technically the beltway is just a big circle so if you don't care how far you drive it doesn't matter which direction you go on it. ;)
Oh, you are from States. So, maybe you heard how they turn left in Michigan :)
CDarklock wrote:
Sun Feb 24, 2019 3:09 pm
So when you start your U-turn, doesn't that mean you're likely to stop in it while the other trains pass?
Yes, thats possible. However in this case there's some saved space for 3 trains to sit there in roundabout, so direct ones should go through fast. Im also currently looking on this intersection between roundabout and direct path, surely something can be done there to improve, I just dont realize it yet.

User avatar
CDarklock
Filter Inserter
Filter Inserter
Posts: 342
Joined: Sun Dec 03, 2017 2:17 am
Contact:

Re: 4-lane RHD intersection with optimized left turn [images+blueprints]

Post by CDarklock »

AlienRaven wrote:
Sun Feb 24, 2019 3:35 pm
Check out this deck: https://drive.google.com/open?id=0B22HA ... Fc5S0wzYjA Its old, but gold in terms of explanation and things haven't changed since then.
That's actually been really helpful. I've been through all the tutorials, incidentally, but they felt kind of... inadequate. I wonder if they might conceivably add a "tutorial mod" interface where the community could write their own tutorials.
Its about space for signals. If one could squeeze in signals and break this into multiple segments it would help. Otherwise game considers this a one segment and it wont let other trains enter while its occupied.
Right. Basically what I was thinking, but I didn't know how to say it effectively. On a real train line, you'd have those tracks passing each other so closely that you can't put a train on each track, let alone squeeze signals onto it so the trains know what to do. But if you spread it out a bit, you could signal everything and improve matters.
Cloverleaf might work, but here it will have same problems, as in real life - connecting traffic on exit\entry from the leafs.
Don't you have that with everything, though? Whenever you have two tracks merging onto one, it seems like that's an issue that has to be handled with signals, and that's essentially every turn in any direction.

I feel like some amount of this can be covered with a simple examination of which trains need to go to which stations. It's definitely most flexible if every train can go to every station from every other station, but is that really necessary?

Rhetorical question, incidentally, it's a question you've undoubtedly already asked and answered for your specific case.
Oh, you are from States. So, maybe you heard how they turn left in Michigan :)
I've lived all over the damn place, so I am familiar with the "Michigan left" - I just forgot it existed until you said something. In DC, they have alternating one-way streets with no left turn, which amounts to the same thing. And I do not believe it is a coincidence that Detroit and DC frequently rank highly on lists of "worst traffic in the US." ;)

It's worth asking why they have these things, though, and I think the answer comes down to human impatience and belligerence - which automated trains don't have. If two organic drivers try to make a left at the same time from different directions, they don't just block one another, they honk and scream at one another while refusing to give ground. This impedes all the traffic for no good reason. But if you signal your intersections properly, your trains won't try to make two lefts at the same time; they'll just obey the signals.

Of course, a properly signaled intersection is going to be slower than an unsignaled intersection. That's the price of like, not running your trains into one another: at least one train in every potential collision has to stop. So in the worst case scenario, how many potential collisions do you have? Because that's the minimum number of trains that have to stop.

And it seems like that's the bottleneck you want to avoid: too many trains stopping at once. So the question I'd ask here is, imagine that your train is approaching the intersection and a left turn would force one or more other trains to stop. If it, instead, takes a right and makes a U-turn... how likely is it that it will still force trains to stop on its way through the junction?

So let's say you've got two lanes running north and two lanes running south. You're on the inner northbound lane, and there is a train alongside you on the other northbound lane. You make a right, forcing this train to stop. You begin to make your U-turn, and once you clear that train's path, it begins to accelerate on the northbound lane again. At the end of your U-turn, does this train force you to stop as well, anyway?

If you swap the trains, so you're in the outer lane, you get the same scenario with the same probability: are the trains alongside one another at the correct distance that one is forced to stop? If one is forced to stop, does it then force the other to stop after the U-turn?

On consideration, the probability that you will make trains coming the opposite direction stop is identical no matter how you cross those tracks. For example, if you would make the train stop when making a left, making a right and a U-turn gives it time to pass. But that also gives a train on the same track enough time to get close enough that it would not have stopped for you to take a left, but it must stop for you to cross the tracks after your U-turn.

So the only thing that matters here, I think, is how likely you are to have two trains alongside one another heading in the same direction. Because that's the only time anything is different from just making the left, unless of course you are running a shedload of trains bumper-to-bumper at speed.

AlienRaven
Inserter
Inserter
Posts: 45
Joined: Wed Apr 26, 2017 7:48 pm
Contact:

Re: 4-lane RHD intersection with optimized left turn [images+blueprints]

Post by AlienRaven »

CDarklock wrote:
Sun Feb 24, 2019 5:01 pm
I wonder if they might conceivably add a "tutorial mod" interface where the community could write their own tutorials.
I think it is already somewhat possible, as described in fff-227 and will be improved in the future by cutscene controller, described in fff-273. Then we only need to wait while someone does that, as after three years of Factorio i find mythelf eventually rechecking rail tutorials from time to time to be sure that my rail spaghetti still makes any sence :)
CDarklock wrote:
Sun Feb 24, 2019 5:01 pm
I've lived all over the damn place, so I am familiar with the "Michigan left" - I just forgot it existed until you said something. In DC, they have alternating one-way streets with no left turn, which amounts to the same thing. And I do not believe it is a coincidence that Detroit and DC frequently rank highly on lists of "worst traffic in the US." ;)
They recently started to introduce it in my homeland as well, thats why I actually started to bother with this small project in the first place :) In fact. i agree with what you said about human nature. I guess it's also a bit safer solution for motorcycle drivers, as from what I see most crashes with them happen due to left turn conflict (not counting stupid behaviour like speeding and wheelies on public roads, ofc).

The thing is, even in my town this was introduced diferently in different places. It's inconvinient, when you have to ride a mile for next U-turn. On the other hand if there are additional trafic lights nearby, that allow a safe U-turn (before pedestrian crossing for example, which blocks both directions) I had the feeling that such left turn was even faster, than a classic yield opposite traffic before turning method.
CDarklock wrote:
Sun Feb 24, 2019 5:01 pm
Rhetorical question, incidentally, it's a question you've undoubtedly already asked and answered for your specific case.
Somehow, after reading your post I feel like it was a psychologist session :) I guess, I'll just build couple more junctions as a test ground with running trains and will try to compare if left turn is actually that much of a problem, perhaps there will be a better (and more compact) solution to this.

Also, I guess next time I'm out for beers, I'll raise one for you :mrgreen:

User avatar
CDarklock
Filter Inserter
Filter Inserter
Posts: 342
Joined: Sun Dec 03, 2017 2:17 am
Contact:

Re: 4-lane RHD intersection with optimized left turn [images+blueprints]

Post by CDarklock »

AlienRaven wrote:
Sun Feb 24, 2019 7:29 pm
Then we only need to wait while someone does that, as after three years of Factorio i find mythelf eventually rechecking rail tutorials from time to time to be sure that my rail spaghetti still makes any sence :)
I've had the game for almost two years, but I've only been actively playing for about a month. It's complex and confusing and I've only recently gotten around to really digging into it.
I guess it's also a bit safer solution for motorcycle drivers
Oh, yeah, it would be. I guess I am as guilty as every other driver of completely forgetting motorcycles exist until I can actually see one.
Somehow, after reading your post I feel like it was a psychologist session
Yeah, I tend to get either that, or "arrogant jerk." Same thing, I guess.
Also, I guess next time I'm out for beers, I'll raise one for you :mrgreen:
Same. :lol:

User avatar
hansjoachim
Fast Inserter
Fast Inserter
Posts: 241
Joined: Wed Apr 26, 2017 7:03 pm
Contact:

Re: 4-lane RHD intersection with optimized left turn [images+blueprints]

Post by hansjoachim »

AlienRaven 4 lanes 4 way intersection
Set 10 = 73 trains/min
Deadlock rating F

Christmas intersection unbuffered 4 lanes 4 way intersection
The most basic and small 4 lanes 4 way intersection
Set 10 = 81 trains/min
Deadlock rating A
https://pastebin.com/YuTezdVq

Hans intersection 2 lanes 4 way intersection
set 10 = 118 trains/min
Deadlock rating A
https://pastebin.com/4u1WpAvy

Max for 2 lanes 4 way. 120-140 Trains/min with 1-3 trains
Max for 4 lanes 4 way 240-280 trains/min with 1-3 trains

So I took a look at your intersection.
I got it to run on the test bench to get a result in set 10 (random traffic)
But it deadlocked in set 9 ( left only)
In Aaarghas deadlock ratting this gets an F for "Deadlocks in regular traffic" Also if you use larger trains it can deadlock with only one train.
There are multiple reasons for letting left turns go to left first and not to the right then to the left. IRL you can let cars/trains take a right turn first then go to the left if you use tunnels/bridges.
Now there are a lot of changes you could do to drastically improve the intersection, like add signals where the trains switches lanes.

The Christmas intersection unbuffered 4 lanes 4-way intersection is way smaller intersection and support all train lengths and is rated A in Aaarghas deadlock rating. You should use this instead.

For reference, this 4 way 2 lane intersection gets with 1-3 trains in set 10: 118
https://pastebin.com/4u1WpAvy
Its optimized for 6 cars, so it could be smaller for 4 cars. If you don't make intersection that beats 2 lanes intersections you don't need to jump to 4 lanes.
Attachments
Deadlock.jpg
Deadlock.jpg (108.48 KiB) Viewed 15563 times
Last edited by hansjoachim on Thu Feb 28, 2019 9:55 pm, edited 1 time in total.

AlienRaven
Inserter
Inserter
Posts: 45
Joined: Wed Apr 26, 2017 7:48 pm
Contact:

Re: 4-lane RHD intersection with optimized left turn [images+blueprints]

Post by AlienRaven »

hansjoachim wrote:
Thu Feb 28, 2019 7:15 pm
So I took a look at your intersection.
Thanks for testing, results are interesting. I guess the bottleneck is 3 sections in the U-turn, so I'll have to stick with other ones from the list with proved no deadlocks :)

User avatar
planetmaker
Fast Inserter
Fast Inserter
Posts: 180
Joined: Mon Jan 21, 2019 9:30 am
Contact:

Re: 4-lane RHD intersection with optimized left turn [images+blueprints]

Post by planetmaker »

The root of the problem is "loops". If you have loops you can always find situations where you have trains fed into them up to a point where they block eachother. The straight tracks and left turns together form a closed loop where trains in principle could go-around forever. Or in other words: it doesn't follow the principle of split before merge.
Maybe you can get by without immediately merging straight and left-turn tracks by placing them adjacent, so only needing slightly more space in the central intersection.

Post Reply

Return to “Railway Setups”