4-way intersections: Throughput and deadlocks [image heavy]

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.
saturn7
Long Handed Inserter
Long Handed Inserter
Posts: 89
Joined: Wed Apr 27, 2016 6:14 pm
Contact:

Re: 4-way intersection testing: Throughput and deadlocks

Post by saturn7 »

It's interesting how much worse the roundabout is compared to a simple compact intersection. Despite that, I have seen in many threads players promoting roundabouts as the best and only way to do 4-way intersections. Is there some benefit of a roundabout (apart from being able to reverse) that is not apparent here?
AndrewIRL
Fast Inserter
Fast Inserter
Posts: 240
Joined: Fri Mar 24, 2017 2:17 pm
Contact:

Re: 4-way intersection testing: Throughput and deadlocks

Post by AndrewIRL »

saturn7 wrote:It's interesting how much worse the roundabout is compared to a simple compact intersection. Despite that, I have seen in many threads players promoting roundabouts as the best and only way to do 4-way intersections. Is there some benefit of a roundabout (apart from being able to reverse) that is not apparent here?
For what it does, allows trains entering from any direction to exit in any direction, a safe roundabout is neat, pretty, simple and uses the least amount of track vs other solutions. Easy to understand and can be built from memory without blueprints. Prior to 0.15 you had to either download the blueprint string mod (30K dls) or make your first rail crossing manually and then blueprint it. Ease of use for the approximately 800K players who don't have the mod.

Only 8 signals required and supports 20/trains per minute.
Image
Mehve
Filter Inserter
Filter Inserter
Posts: 318
Joined: Sat Aug 06, 2016 9:12 pm
Contact:

Re: 4-way intersection testing: Throughput and deadlocks

Post by Mehve »

I used this particular roundabout implementation for quite some time without a single hiccup. I'm curious how it stacks up in the tester...

Image
Blueprint
aaargha
Filter Inserter
Filter Inserter
Posts: 333
Joined: Wed Dec 07, 2016 8:35 am
Contact:

Re: 4-way intersection testing: Throughput and deadlocks

Post by aaargha »

Distelzombie wrote:Great work!
This one is not in your list. It is similiar to the basic direct cross, but its bigger and has signals.
As far as i know its made by maurojunior2011, but i've seen it in other threads too.
Tested as "Compact attempt". Main throughput limiter is that it's impossible to separate input/output blocks from the same direction. As such any train going straight ahead will block the entire intersection.
Mehve wrote:I used this particular roundabout implementation for quite some time without a single hiccup. I'm curious how it stacks up in the tester...
Tested as "Crossabout". Behaves pretty much the same as regular roundabouts with regards to deadlocks, example in album. Also, is this an original design or do you happen to have a source for it? I think I remember seeing something like it somewhere.

I feel like I should add a warning for people to include names with their submissions, it's probably best I don't name them :D
Mehve
Filter Inserter
Filter Inserter
Posts: 318
Joined: Sat Aug 06, 2016 9:12 pm
Contact:

Re: 4-way intersection testing: Throughput and deadlocks

Post by Mehve »

aaargha wrote:
Mehve wrote:I used this particular roundabout implementation for quite some time without a single hiccup. I'm curious how it stacks up in the tester...
Tested as "Crossabout". Behaves pretty much the same as regular roundabouts with regards to deadlocks, example in album. Also, is this an original design or do you happen to have a source for it? I think I remember seeing something like it somewhere.

I feel like I should add a warning for people to include names with their submissions, it's probably best I don't name them :D
Whoops, missed that listing somehow. I can honestly say that I put it together myself, and have posted the design in older threads before. That said, I don't consider it anything that someone else couldn't have come up with as well. It's basically just a standard roundabout, enlarged just enough that you can install the dedicated left-turn paths so they don't intersect with their mirroring left-turn.
lockthatpl
Manual Inserter
Manual Inserter
Posts: 3
Joined: Thu Aug 11, 2016 3:28 pm
Contact:

Re: 4-way intersection testing: Throughput and deadlocks

Post by lockthatpl »

Hey guys !

I want to share with you the intersections i'm using, I made a long time ago and i noticed that some signals are maybe missing. It's a right hand drive and it didn't create a deadclock on my current map. Maybe with a higher througput it will ? :o
4 way cross.png
4 way cross.png (437.41 KiB) Viewed 22858 times
Blueprint string :
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
User avatar
Distelzombie
Filter Inserter
Filter Inserter
Posts: 336
Joined: Tue May 02, 2017 4:27 pm
Contact:

Re: 4-way intersection testing: Throughput and deadlocks

Post by Distelzombie »

aaargha wrote:
Distelzombie wrote:Great work!
This one is not in your list. It is similiar to the basic direct cross, but its bigger and has signals.
As far as i know its made by maurojunior2011, but i've seen it in other threads too.
Tested as "Compact attempt". Main throughput limiter is that it's impossible to separate input/output blocks from the same direction. As such any train going straight ahead will block the entire intersection.
Oops, how did I miss this? At least Im not alone. xD
EDIT: Oh, wait its new. XD XD Oh dude...
Edit: Maybe? Im not sure...

Who did this "cross" you have under A? I would want to include it in my blueprint book. (Changed, for 1-2-1 trains)
Complete 2-Lane system as a Blueprint-Book! The perfect OCD reactor? Testing chained science lab efficiency Please use real prefixes and proper rounding!
User avatar
Distelzombie
Filter Inserter
Filter Inserter
Posts: 336
Joined: Tue May 02, 2017 4:27 pm
Contact:

Re: 4-way intersection testing: Throughput and deadlocks

Post by Distelzombie »

Apparently this 4-way can get stuck:
Even though they only want to move on the outside of the cross they get stuck. I dont think it has to do with the surroundings.
I have no idea why they are not moving though. Should be fine.
Attachments
stuck2.png
stuck2.png (1.86 MiB) Viewed 22708 times
stuck.png
stuck.png (2.1 MiB) Viewed 22711 times
Last edited by Distelzombie on Thu May 11, 2017 3:03 pm, edited 1 time in total.
Complete 2-Lane system as a Blueprint-Book! The perfect OCD reactor? Testing chained science lab efficiency Please use real prefixes and proper rounding!
Aeternus
Filter Inserter
Filter Inserter
Posts: 835
Joined: Wed Mar 29, 2017 2:10 am
Contact:

Re: 4-way intersection testing: Throughput and deadlocks

Post by Aeternus »

The red arrows signify the block that is already occupied, which is why the chain signals stay red as well. There is no room for a signal between the branch that comes from the right instead of left at the exits of this cross.
saturn7 wrote:It's interesting how much worse the roundabout is compared to a simple compact intersection. Despite that, I have seen in many threads players promoting roundabouts as the best and only way to do 4-way intersections. Is there some benefit of a roundabout (apart from being able to reverse) that is not apparent here?
They allow trains to reverse, which makes monodirectional rails possible. IE a T junction or station split off that forces trains to stay on one side of the track, reversing them at the next roundabout. Problem with using roundabouts as high traffic systems is that people forget that a roundabout needs to be able to buffer traffic for smooth flow. Which means it needs to be big. VERY big to function effectively. They work wonderfully when used as such. But for crossing purposes, the "safe" roundabout and the direct cross are about on-par, with direct having slightly higher traffic due to being smaller, but no reversing option.

I myself have grown quite fond of the roundabout T junctions... I use those a lot these days. No locks thus far, using mainly 2 locs 4 wagons for cargo movement.
Last edited by Aeternus on Thu May 11, 2017 3:19 pm, edited 1 time in total.
User avatar
Distelzombie
Filter Inserter
Filter Inserter
Posts: 336
Joined: Tue May 02, 2017 4:27 pm
Contact:

Re: 4-way intersection testing: Throughput and deadlocks

Post by Distelzombie »

Aeternus wrote:The red arrows signify the block that is already occupied, which is why the chain signals stay red as well. There is no room for a signal between the branch that comes from the right instead of left at the exits of this cross.
Sorry I dont understand. (It shouldnt be red since there is nothing) I Also edited the post above with another picture thats even weirder.
Complete 2-Lane system as a Blueprint-Book! The perfect OCD reactor? Testing chained science lab efficiency Please use real prefixes and proper rounding!
Aeternus
Filter Inserter
Filter Inserter
Posts: 835
Joined: Wed Mar 29, 2017 2:10 am
Contact:

Re: 4-way intersection testing: Throughput and deadlocks

Post by Aeternus »

I'll illustrate where the problem is by editing your own pic:

Image

The red circle is just a single block, only one train may be on any section of this rail, even if they don't directly intersect. The blue lines show where chain signals should be, to break up the parallel tracks so that they don't block eachother - but this intersection does not have room for chain signals at those spots. Widening this design and adding the needed chain signals should resolve the problem
User avatar
Distelzombie
Filter Inserter
Filter Inserter
Posts: 336
Joined: Tue May 02, 2017 4:27 pm
Contact:

Re: 4-way intersection testing: Throughput and deadlocks

Post by Distelzombie »

Ok. How do you see that without testing??
I had it run for a while and put signals here and there and let it run, changed stuff... it works now, apparently. Because im an idiot. (I removed these two signals in the pic because reasons) I should delete my posts here since they are not related to this thread. Were never.
Attachments
stuck.png
stuck.png (2.22 MiB) Viewed 22704 times
Complete 2-Lane system as a Blueprint-Book! The perfect OCD reactor? Testing chained science lab efficiency Please use real prefixes and proper rounding!
User avatar
Distelzombie
Filter Inserter
Filter Inserter
Posts: 336
Joined: Tue May 02, 2017 4:27 pm
Contact:

Re: 4-way intersection testing: Throughput and deadlocks

Post by Distelzombie »

lockthatpl wrote:Hey guys !

I want to share with you the intersections i'm using, I made a long time ago and i noticed that some signals are maybe missing. It's a right hand drive and it didn't create a deadclock on my current map. Maybe with a higher througput it will ? :o
4 way cross.png
The Blueprint sting is wrong, lockthatpl.
Complete 2-Lane system as a Blueprint-Book! The perfect OCD reactor? Testing chained science lab efficiency Please use real prefixes and proper rounding!
aaargha
Filter Inserter
Filter Inserter
Posts: 333
Joined: Wed Dec 07, 2016 8:35 am
Contact:

Re: 4-way intersection testing: Throughput and deadlocks

Post by aaargha »

Mehve wrote:Whoops, missed that listing somehow. I can honestly say that I put it together myself, and have posted the design in older threads before. That said, I don't consider it anything that someone else couldn't have come up with as well. It's basically just a standard roundabout, enlarged just enough that you can install the dedicated left-turn paths so they don't intersect with their mirroring left-turn.
Distelzombie wrote:Oops, how did I miss this? At least Im not alone. xD
EDIT: Oh, wait its new. XD XD Oh dude...
Edit: Maybe? Im not sure...

Who did this "cross" you have under A? I would want to include it in my blueprint book. (Changed, for 1-2-1 trains)
Those were added because your submissions, just a bit of poor wording on my part. Sorry for the confusion :)
I don't remember where I saw that design originally, the one posted is made by me from what I remember of the original design. If you manage to find it, please let me know.

@zytukin: Sorry for not addressing you earlier, was a bit short on time. At the moment I do not have systems that can handle more than 2-lane intersections. But as there are a few other requests for multi-lane intersections I'll get around to building some, it'll probably take at least few days though. Also, if you have a blueprint string for your intersection that you can share I'd be most grateful.

@Distelzombie: Glad you got it worked out. As for seeing where the blocks end, it's mostly just following the rails (crossing ones as well) as far as possible until you hit a signal. The parts you can get to are in the same block. Also in 0.15.10 they added a debug option called show-rail-blocks which displays this in-game.

@lockthatpl: That blueprint string is for a blueprint book that does not contain the intersection in question.
User avatar
Distelzombie
Filter Inserter
Filter Inserter
Posts: 336
Joined: Tue May 02, 2017 4:27 pm
Contact:

Re: 4-way intersection testing: Throughput and deadlocks

Post by Distelzombie »

aaargha wrote:The parts you can get to are in the same block. Also in 0.15.10 they added a debug option called show-rail-blocks which displays this in-game.
It looks like a bug to me. The block way to big and the inherent rails have nothing to do with each other!
Complete 2-Lane system as a Blueprint-Book! The perfect OCD reactor? Testing chained science lab efficiency Please use real prefixes and proper rounding!
lockthatpl
Manual Inserter
Manual Inserter
Posts: 3
Joined: Thu Aug 11, 2016 3:28 pm
Contact:

Re: 4-way intersection testing: Throughput and deadlocks

Post by lockthatpl »

I'm using the little arrow that points on the top right corner in my blueprint book is this right ? :o Or is there another manipulation to do that i'm not aware ?

Maybe this one will work (this one is from the blueprint alone, not in a book) :

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
aaargha
Filter Inserter
Filter Inserter
Posts: 333
Joined: Wed Dec 07, 2016 8:35 am
Contact:

Re: 4-way intersection testing: Throughput and deadlocks

Post by aaargha »

Distelzombie wrote:It looks like a bug to me. The block way to big and the inherent rails have nothing to do with each other!
Unfortunately that's how signals work at the moment, pic below is the same principle. What you're proposing would be a major change to signalling rules, signals would need to be way smarter than they are now for it work.
Attachments
regular.png
regular.png (266.87 KiB) Viewed 23724 times
aaargha
Filter Inserter
Filter Inserter
Posts: 333
Joined: Wed Dec 07, 2016 8:35 am
Contact:

Re: 4-way intersection testing: Throughput and deadlocks

Post by aaargha »

lockthatpl wrote:I'm using the little arrow that points on the top right corner in my blueprint book is this right ? :o Or is there another manipulation to do that i'm not aware ?

Maybe this one will work (this one is from the blueprint alone, not in a book) :
That one worked, I think you can do individual blueprints from a book if you click the arrow when editing a blueprint. The provided signalling is indeed faulty, the picture below is after about 3 minutes of traffic. It was pretty funny to see how close it got to deadlocking a few times but always seemed to right itself at the last second, until it didn't :). I've added an entry with corrected signalling as "Wide B", as it's basically the same as "Wide".
Attachments
regular.png
regular.png (1.64 MiB) Viewed 23709 times
lockthatpl
Manual Inserter
Manual Inserter
Posts: 3
Joined: Thu Aug 11, 2016 3:28 pm
Contact:

Re: 4-way intersection testing: Throughput and deadlocks

Post by lockthatpl »

Well i think my old train network wasn't so heavy traffic as your test ^^

Thank you for giving a look and updating my blueprint, i will update mine aswell :)
AndrewIRL
Fast Inserter
Fast Inserter
Posts: 240
Joined: Fri Mar 24, 2017 2:17 pm
Contact:

Re: 4-way intersection testing: Throughput and deadlocks

Post by AndrewIRL »

This is great work! :D
aaargha wrote:Throughput testing
Here we use a simulated high traffic situation with trains coming and going in all directions. For the test we're using 24 fully upgraded 2-4 trains running on rocket fuel.
I was wondering how throughput would be affected with 1-4-1 trains. In theory the lower speed of 1-way vs 2-way trains should mean a significant decrease in throughput.

Code: Select all

LL-CCCC,  ratio = 2, top speed = 259.2
L-CCCC-L, ratio = 6, top speed = 134.6
The 1-4-1 train has only 52.93% the speed of the 2-4-0 so I would predict that the maximum throughput will be somewhere in the range: 20-26 down from 64 for the highest throughput intersection.

Maybe you would be willing to test just one intersection, the big 64 throughput cross?
Locked

Return to “Railway Setups”