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.
User avatar
hansjoachim
Fast Inserter
Fast Inserter
Posts: 128
Joined: Wed Apr 26, 2017 7:03 pm
Contact:

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

Post by hansjoachim » Sun May 19, 2019 9:02 am

SkiCarver wrote:
Sun May 19, 2019 8:43 am
Similar to some of the designs posted, but it works well.
I have seen that multiple times before.
Works ok but you should preferably keep turnarounds out of the intersection, atleast in busy areas
Updated version of aaarghas intersection testbench. viewtopic.php?f=194&t=46855&p=383830#p383830
My Highscore intersections: https://pastebin.com/dGsJT71Z
Combinator Adaptable merger https://pastebin.com/aLPQQeei

SkiCarver
Fast Inserter
Fast Inserter
Posts: 165
Joined: Sat Apr 06, 2019 9:36 am
Contact:

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

Post by SkiCarver » Sun May 19, 2019 9:40 am

I have looked throughout this threat and it does not appear. As I say, similar, but not the same... Also, there is no roundabout in these junctions. left, forward and right only.

mrvn
Smart Inserter
Smart Inserter
Posts: 3696
Joined: Mon Sep 05, 2016 9:10 am
Contact:

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

Post by mrvn » Mon May 20, 2019 10:47 am

I think you can move the rail signal at the exit points one step inward. There is no gain of the extra chain signal.

SkiCarver
Fast Inserter
Fast Inserter
Posts: 165
Joined: Sat Apr 06, 2019 9:36 am
Contact:

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

Post by SkiCarver » Mon May 20, 2019 11:14 am

mrvn wrote:
Mon May 20, 2019 10:47 am
I think you can move the rail signal at the exit points one step inward. There is no gain of the extra chain signal.
Thanks, but I just checked and they cannot move inward - invalid position.

mrvn
Smart Inserter
Smart Inserter
Posts: 3696
Joined: Mon Sep 05, 2016 9:10 am
Contact:

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

Post by mrvn » Mon May 20, 2019 12:49 pm

SkiCarver wrote:
Mon May 20, 2019 11:14 am
mrvn wrote:
Mon May 20, 2019 10:47 am
I think you can move the rail signal at the exit points one step inward. There is no gain of the extra chain signal.
Thanks, but I just checked and they cannot move inward - invalid position.
You have chain signals there, at least on the 4 lane version, so the position is valid. Simply replace them with rail signals.

SkiCarver
Fast Inserter
Fast Inserter
Posts: 165
Joined: Sat Apr 06, 2019 9:36 am
Contact:

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

Post by SkiCarver » Mon May 20, 2019 12:55 pm

mrvn wrote:
Mon May 20, 2019 12:49 pm
SkiCarver wrote:
Mon May 20, 2019 11:14 am
mrvn wrote:
Mon May 20, 2019 10:47 am
I think you can move the rail signal at the exit points one step inward. There is no gain of the extra chain signal.
Thanks, but I just checked and they cannot move inward - invalid position.
You have chain signals there, at least on the 4 lane version, so the position is valid. Simply replace them with rail signals.
I think perhaps I am not understanding you. I have rail signals on the exits and chains signals on the entrances to the junction. This is how it need to be for the junction to work.

mrvn
Smart Inserter
Smart Inserter
Posts: 3696
Joined: Mon Sep 05, 2016 9:10 am
Contact:

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

Post by mrvn » Mon May 20, 2019 12:58 pm

SkiCarver wrote:
Mon May 20, 2019 12:55 pm
mrvn wrote:
Mon May 20, 2019 12:49 pm
SkiCarver wrote:
Mon May 20, 2019 11:14 am
mrvn wrote:
Mon May 20, 2019 10:47 am
I think you can move the rail signal at the exit points one step inward. There is no gain of the extra chain signal.
Thanks, but I just checked and they cannot move inward - invalid position.
You have chain signals there, at least on the 4 lane version, so the position is valid. Simply replace them with rail signals.
I think perhaps I am not understanding you. I have rail signals on the exits and chains signals on the entrances to the junction. This is how it need to be for the junction to work.
At the exit where the last 2 rails merge you have chain signals before the merge and a rail signal after the merge. You can replace that by 2 rail signals before the merge and none after.

SkiCarver
Fast Inserter
Fast Inserter
Posts: 165
Joined: Sat Apr 06, 2019 9:36 am
Contact:

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

Post by SkiCarver » Mon May 20, 2019 1:07 pm

mrvn wrote:
Mon May 20, 2019 12:58 pm
SkiCarver wrote:
Mon May 20, 2019 12:55 pm
mrvn wrote:
Mon May 20, 2019 12:49 pm
SkiCarver wrote:
Mon May 20, 2019 11:14 am
mrvn wrote:
Mon May 20, 2019 10:47 am
I think you can move the rail signal at the exit points one step inward. There is no gain of the extra chain signal.
Thanks, but I just checked and they cannot move inward - invalid position.
You have chain signals there, at least on the 4 lane version, so the position is valid. Simply replace them with rail signals.
I think perhaps I am not understanding you. I have rail signals on the exits and chains signals on the entrances to the junction. This is how it need to be for the junction to work.
At the exit where the last 2 rails merge you have chain signals before the merge and a rail signal after the merge. You can replace that by 2 rail signals before the merge and none after.
Ahh, I am with you now and yes, I could do that. thanks.

pallrider
Burner Inserter
Burner Inserter
Posts: 7
Joined: Mon Jun 03, 2019 2:22 am
Contact:

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

Post by pallrider » Mon Jun 03, 2019 2:40 am

I have created a "compact" 2 lane intersection designed for 2+8 trains (actually, it was originally 3+8). I wanted it to take up less horizontal and vertical space than some other designs. It is supposed to be symmetric (though this may not be totally true anymore).

The way I use it is as a mainline intersection, with my resource sinks on the north and south, and mainline running east-west. It has less buffers than the ideal intersection, but it still doesn't perform at the level that I was hoping. The main issue (for my use) is that the east-traffic is constrained.

Could anyone provide feedback on this intersection? Is there some basic things that I could improve before I start investigating circuit logic?
Blueprint
Attachments
Screen Shot 2019-06-03 at 12.31.42 pm.png
Screen Shot 2019-06-03 at 12.31.42 pm.png (1.49 MiB) Viewed 919 times

User avatar
hansjoachim
Fast Inserter
Fast Inserter
Posts: 128
Joined: Wed Apr 26, 2017 7:03 pm
Contact:

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

Post by hansjoachim » Mon Jun 03, 2019 3:35 pm

Hey, I have looked at it
There is one major throughput limiter and that is in the center for all 4 straight paths.
In circle 1 straight driving trains have to wait for potentially 4 different trains to pass. Therefore they can only pass every fifth time on average with heavy load.
If you were to make 4 buffer, marked red, straight driving trains would only have to wait for potentially 2 trains a time, almost doubling throughput for straight driving trains.

Trains driving left have a similar issue in circle 2 where they have to wait for potentially three other trains before driving. A buffer there would increase throughput.
You could remove the buffers marked with green for straight driving trains if you want to save space. It will affect throughput, but shouldn't effect it too much.

If you want I can make a version with the changes and show a throughput comparison.

Edit: This intersection would drastically improve throughput, especially straight through traffic, at about the same size:
https://pastebin.com/phGeh5yr
Attachments
20190603_173301.png
20190603_173301.png (1.33 MiB) Viewed 878 times
Updated version of aaarghas intersection testbench. viewtopic.php?f=194&t=46855&p=383830#p383830
My Highscore intersections: https://pastebin.com/dGsJT71Z
Combinator Adaptable merger https://pastebin.com/aLPQQeei

pallrider
Burner Inserter
Burner Inserter
Posts: 7
Joined: Mon Jun 03, 2019 2:22 am
Contact:

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

Post by pallrider » Tue Jun 04, 2019 1:03 pm

Thanks for that blueprint, I get it showing about 50 trains per minute vs the 36 that I was getting on the other one. That is a really nice increase, thanks. I'll continue working on mine too. Thanks for your feedback, that was really useful.

pallrider
Burner Inserter
Burner Inserter
Posts: 7
Joined: Mon Jun 03, 2019 2:22 am
Contact:

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

Post by pallrider » Thu Jun 06, 2019 6:58 am

@hansjoachim, I really like your intersection that you suggested, but I really needed it to fit in the same footprint as my intersection, so I modified it to fit
Screen Shot 2019-06-06 at 4.53.23 pm.png
Screen Shot 2019-06-06 at 4.53.23 pm.png (1.16 MiB) Viewed 798 times

User avatar
hansjoachim
Fast Inserter
Fast Inserter
Posts: 128
Joined: Wed Apr 26, 2017 7:03 pm
Contact:

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

Post by hansjoachim » Thu Jun 06, 2019 2:20 pm

Cool, I like it!
There are two change I want to make.
Trains from three or more lanes should merge in the same block. The current merger favors right turns, also throughput of the merger is higher if they merge in the same block. Moving right turns looks the easiest.

Second is adding chain signals here at the red marks
Because that would allow trains to go into the other waiting bay if they firstly tried to go into a full one.
Attachments
20190606_162418.png
20190606_162418.png (812.83 KiB) Viewed 769 times
Updated version of aaarghas intersection testbench. viewtopic.php?f=194&t=46855&p=383830#p383830
My Highscore intersections: https://pastebin.com/dGsJT71Z
Combinator Adaptable merger https://pastebin.com/aLPQQeei

pallrider
Burner Inserter
Burner Inserter
Posts: 7
Joined: Mon Jun 03, 2019 2:22 am
Contact:

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

Post by pallrider » Wed Jun 12, 2019 6:23 am

Great feedback. Here is the updated version
Screen Shot 2019-06-12 at 4.22.06 pm.png
Screen Shot 2019-06-12 at 4.22.06 pm.png (1.62 MiB) Viewed 659 times

pallrider
Burner Inserter
Burner Inserter
Posts: 7
Joined: Mon Jun 03, 2019 2:22 am
Contact:

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

Post by pallrider » Wed Jun 12, 2019 6:31 am

I've also been working on a larger intersection that tries to make as much use of parallel lanes as possible. I'm not really comparing to other intersections directly (since I have changed my train config): but I was able to get a score of 81 when using nuclear powered 3-7 trains.

The main bottleneck is now at the merge outlet. Any feedback that anyone has on improving that would be appreciated.
Screen Shot 2019-06-12 at 4.30.16 pm.png
Screen Shot 2019-06-12 at 4.30.16 pm.png (633.43 KiB) Viewed 657 times

User avatar
hansjoachim
Fast Inserter
Fast Inserter
Posts: 128
Joined: Wed Apr 26, 2017 7:03 pm
Contact:

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

Post by hansjoachim » Wed Jun 12, 2019 9:11 pm

The copy blueprint isn't working. The first intersection looks really good now!

On the second, three lanes parallel one.
Merging is a bottleneck no doubt.
I got a combinator based merger you could try.
Its in my signatur.

I have two changes
Cutting the extra right turn waiting bays. They just makes merging harder. The point of having three lanes is to improve throughput in the intersection. Right turns aren't intersecting with anything so they don't need more lanes.
They only works as waiting area and hardly improves throughput.

The Yellow lines. They might reduce throughput in the intersection but the merger is the bottleneck so they will probably not effect throughput. The change will reduce rails used and make merging all the rails easier. It would have to be tested.

How does it compare to the other intersection? they have the same car length so they should be easy to compare.
Attachments
Alternativ Pallrider.png
Alternativ Pallrider.png (416.54 KiB) Viewed 618 times
Updated version of aaarghas intersection testbench. viewtopic.php?f=194&t=46855&p=383830#p383830
My Highscore intersections: https://pastebin.com/dGsJT71Z
Combinator Adaptable merger https://pastebin.com/aLPQQeei

vanatteveldt
Filter Inserter
Filter Inserter
Posts: 920
Joined: Wed Nov 25, 2015 11:44 am
Contact:

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

Post by vanatteveldt » Fri Jun 14, 2019 9:26 am

This might be patched really quickly, but it must be possible to use this 'feature' to improve intersections quite dramatically, no?

https://www.reddit.com/r/factorio/comme ... on_design/

User avatar
hansjoachim
Fast Inserter
Fast Inserter
Posts: 128
Joined: Wed Apr 26, 2017 7:03 pm
Contact:

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

Post by hansjoachim » Fri Jun 14, 2019 9:36 am

It can drastically improve small intersection, but in mine intersection and some others the limitation of throughput is merging lanes.
I consider using this as cheating and taking away challenge.
Updated version of aaarghas intersection testbench. viewtopic.php?f=194&t=46855&p=383830#p383830
My Highscore intersections: https://pastebin.com/dGsJT71Z
Combinator Adaptable merger https://pastebin.com/aLPQQeei

vanatteveldt
Filter Inserter
Filter Inserter
Posts: 920
Joined: Wed Nov 25, 2015 11:44 am
Contact:

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

Post by vanatteveldt » Fri Jun 14, 2019 10:37 am

hansjoachim wrote:
Fri Jun 14, 2019 9:36 am
It can drastically improve small intersection, but in mine intersection and some others the limitation of throughput is merging lanes.
I consider using this as cheating and taking away challenge.
Yeah it would be more of a gimmick / proof of concept. But in a way you can see this as a railway bridge, just imagine the S-piece being elevated :)

User avatar
hansjoachim
Fast Inserter
Fast Inserter
Posts: 128
Joined: Wed Apr 26, 2017 7:03 pm
Contact:

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

Post by hansjoachim » Fri Jun 14, 2019 10:49 am

Yeah it looks like multiple levels.
But if factorio where to include bridges and make making high throughput networks easier, there should be some drawbacks like in OpenTTD where curves and bridges slows down trains significantly.
Updated version of aaarghas intersection testbench. viewtopic.php?f=194&t=46855&p=383830#p383830
My Highscore intersections: https://pastebin.com/dGsJT71Z
Combinator Adaptable merger https://pastebin.com/aLPQQeei

Post Reply

Return to “Railway Setups”

Who is online

Users browsing this forum: No registered users