Factorio Forums

Okay so, as the title suggests, I came across an academic question which annoys me a bit

Basically, I wanted to implement an 8 way intersection, where you would have (rotational) symmetry AND no roundabout to avoid throughput problems. After thinking about it a little bit, I think this intersection necessarily implies to have a roundabout OR that you would need a ridiculously enormous intersection to avoid rails merging together on some key spots. Since I didn't have the courage to try and implement this ridiculously large intersection, my academic problem would be: what would be the most compact intersection that verifies the imposed conditions? Maybe if some of you work in maths, you could use your knowledge of topology to answer this, who knows?

I just think it could be pretty interesting to submit this problem here, I'd be quite curious about the solutions people may come up with on this forum, so here it is!

Bugger wrote: ↑Sun Apr 14, 2019 12:56 pmavoid rails merging together on some key spots.

You need to define this better. Is a 4-way intersection acceptable? How about a 3-way intersection? The highest throughput intersections use only side merges and lane swaps, and it indeed results in a ridiculously enormous intersection.

May I introduce you to a forum post that analyzes and compares many different intersections?

viewtopic.php?f=194&t=46855

A roundabout increases the throughput
A basic intersection only allows one train at a time which is the worst you can have.

I think you should reorganize your rail grid just to make the need of a 8 way intersection obsolete.
my complete base is one big roundabout. Trains barelly have to stop.
For example: enough space to let a train speed up before merging in.
When all trains go in one direction there are almost no intersections.

Simplest thing I can think of is to invert all of the "diagonal" lanes (flip half of the roads between left and right hand drive), join the inverted lanes with the then-aligned lanes on either side (either merge or run them along-side), and feed it into your choice of 4-way intersection (might want to base it on an intersection which uses twice as many lanes as your network and which incorporates lane switching and turnaround).

JimBarracus wrote: ↑Mon Apr 15, 2019 6:40 am A roundabout increases the throughput

From what has been measured in the topic seePyou linked, it's exactly the contrary, roundabouts seem to be amongst the worst alternatives.

Koub wrote: ↑Mon Apr 15, 2019 8:23 am

JimBarracus wrote: ↑Mon Apr 15, 2019 6:40 am A roundabout increases the throughput

From what has been measured in the topic seePyou linked, it's exactly the contrary, roundabouts seem to be amongst the worst alternatives.

He is talking about the whole train system being one giant loop. Not a cross that uses a roundabout at its core.

With one big giant loop the throughput is limited by the number of trains you can put through a single rail line. The merging costs you some more in case there is contention and a train needs to break. But still that's often more throughput than even a single crossing has.

DaveMcW wrote: ↑Sun Apr 14, 2019 5:31 pm

Bugger wrote: ↑Sun Apr 14, 2019 12:56 pmavoid rails merging together on some key spots.

You need to define this better. Is a 4-way intersection acceptable? How about a 3-way intersection? The highest throughput intersections use only side merges and lane swaps, and it indeed results in a ridiculously enormous intersection.

What I meant by this, is that if you merge some exit tracks, then applying a rotational symmetry would result in a roundabout inside the intersection. For example, let's say we work on the West exit. Then, it would be quite natural to merge the tracks coming from SE, S and SW. But then, if you apply the same scheme on the exit to the North, East and South, you would have a roundabout at the center of the intersection, allowing the possibility of a U-turn for your trains. I'm not sure if I am clear by saying this, but maybe i'll post a picture to illustrate this later.

Of course 4-ways and 3-ways are acceptable, but I was just playing around with an 8-ways intersection. Then, I wanted to take it a bit further and tried to optimise it a little, so I wanted to avoid having a roundabout inside the intersection. I did not work on this for too long, but I thought it'd b interesting to see if someone could come up with an 8-ways intersection without a central roundabout!

seePyou wrote: ↑Sun Apr 14, 2019 10:48 pm May I introduce you to a forum post that analyzes and compares many different intersections?

viewtopic.php?f=194&t=46855

Yes I have been reading this thread for quite a while now, and even used the throughput tester map for some of my designs

Eketek wrote: ↑Mon Apr 15, 2019 8:22 am Simplest thing I can think of is to invert all of the "diagonal" lanes (flip half of the roads between left and right hand drive), join the inverted lanes with the then-aligned lanes on either side (either merge or run them along-side), and feed it into your choice of 4-way intersection (might want to base it on an intersection which uses twice as many lanes as your network and which incorporates lane switching and turnaround).

That's an interesting idea, maybe I'll try something with it later!

JimBarracus wrote: ↑Mon Apr 15, 2019 6:40 am A roundabout increases the throughput
A basic intersection only allows one train at a time which is the worst you can have.

I think you should reorganize your rail grid just to make the need of a 8 way intersection obsolete.
my complete base is one big roundabout. Trains barelly have to stop.
For example: enough space to let a train speed up before merging in.
When all trains go in one direction there are almost no intersections.

I really don't want to be pedantic, but that's not really the point of my thread. First, (local) roundabouts have a lower throughput than intersections, and basically that comes from the fact that you increase the number of paths crossing each other. Also, basic intersections do not allow only one train at a time if you make the signals correctly!
Finally, I generally use only 4-ways or less in my train networks, but I was just having fun by trying to design a nice 8-ways intersection.
However, the idea of the global roundabout is quite nice, but I think it's not very different from the popular grid design and I don't often implement this because I don't find it super interesting. It's very efficient, but a bit boring (in my opinion of course).

Only problem with having a roundabout inside the cross is with trains longer than the size of the roundabout. So lets assume you want your 8 way intersection to be able to handle L*128+C*1024 trains. A roundabout would allow trains to make a 270° turn and crash into them self.

That said I think I only just got what you mean by 8 way intersection. It's no 8 parallel tracks but you want (one?) tracks coming an going from the 4 major and 4 minor compass points. Right?

mrvn wrote: ↑Sun Apr 21, 2019 8:43 pm That said I think I only just got what you mean by 8 way intersection. It's no 8 parallel tracks but you want (one?) tracks coming an going from the 4 major and 4 minor compass points. Right?

Code: Select all

\ | /
_\|/_
 /|\
/ | \

https://mods.factorio.com/mod/entity-symmetry


Like this

This seems straigth forward: I cann see no way a train could turn around.

Qon wrote: ↑Tue Apr 23, 2019 12:05 pm
https://mods.factorio.com/mod/entity-symmetry

Image

Like this

Nice! Very elegant.
Problem: you cannot make a hard right from the cardinal directions(0 to 315, 90 to 45, 180 to 135, 270 to 225).
It looks like an easy fix.

alienwombat wrote: ↑Tue Apr 23, 2019 4:02 pm This seems straigth forward:

I cann see no way a train could turn around.

Also very interesting...
Problem: train cannot turn 90 degrees from 45, 135, 225 and 315.
Signaling the additional turns might be difficult.

-Zaka

The general pattern for designing a high-throughput intersection seems to be:

For each incoming line:
1- branch off separate tracks for each direction it could go to (normally 3, here 7)
2- provide a buffer
3- go to destination while crossing a minimum number of other tracks
4- provide a buffer
5- merge

I see no reason why that wouldn't work here? Of course, why on Nauvis anyone would ever want such a thing is a different question...

Why would you even need an 8-way intersection. Any situation that requires something like that is inherently flawed.

You should never get yourself into a situation where you need more than a 4-way intersection and if you really want a hardcore train-based megabase then you might even think about avoiding 4-way intersections too and only go for 3-way intersections using a honeycomb-like pattern where only 3 egdes meet. That is my advice with experience in civil engineering and urban planning.

But apart from that I have posted one of those monstrosities above years ago because someone else had the same idea but it obviously is useless.