Factorio Forums

www.factorio.com
https://forums.factorio.com/

3 and 4 way intersections

https://forums.factorio.com/viewtopic.php?f=194&t=100614

Page 1 of 15

3 and 4 way intersections

Posted: Sun Nov 14, 2021 1:48 pm
by Factoriointersection
-- Info --
Intersections
-- 4-Way --
-- 3-Way --

Re: 3 and 4 way intersections

Posted: Sun Nov 14, 2021 3:36 pm
by Factoriointersection
Reserved

Re: 3 and 4 way intersections [Work in progress]

Posted: Mon Nov 15, 2021 8:38 am
by Tallywort
Great to see a new intersection thread. Hope this one will be as popular as the last

Re: 3 and 4 way intersections [Work in progress]

Posted: Mon Nov 15, 2021 1:53 pm
by FuryoftheStars
Indeed, awesome to see this brought back to life!

Re: 3 and 4 way intersections [Work in progress]

Posted: Tue Nov 16, 2021 6:43 am
by aka13
As always, thank you for your time, great work

Re: 3 and 4 way intersections, Ranked by Throughput

Posted: Wed Nov 17, 2021 1:27 pm
by Factoriointersection
Thanks=)
Now the post is usable.
Still work to be done though=)

Re: 3 and 4 way intersections [Work in progress]

Posted: Wed Nov 17, 2021 5:19 pm
by Koub
In the old thread, there was an additional criterium : the intersection's deadlock proneness (which I found tremendously useful). Do you plan to integrate it at some point ? Or is it out of your scope ?

Re: 3 and 4 way intersections

Posted: Wed Nov 17, 2021 6:10 pm
by Factoriointersection
Koub wrote: ↑
Wed Nov 17, 2021 5:19 pm
 In the old thread, there was an additional criterium : the intersection's deadlock proneness (which I found tremendously useful). Do you plan to integrate it at some point ? Or is it out of your scope ?
After the 0.15.27 update intersections will either deadlock in normal traffic or it will not. So the old deadlock rating is useless. For posted intersections that deadlock we'll comment it:)
aaargha wrote: With the changes made to the path-finding algorithm in 0.15.27 almost all intersections rated from B to E (inclusive) are safe as long as the player does neither, disable the train station the train is heading to while travelling, nor, destroy/build rails/signals that makes the train's path invalid.

Re: 3 and 4 way intersections

Posted: Wed Nov 17, 2021 8:04 pm
by Koub
Haha I feel stupid. I confess I have almost never used trains, I guess the last time was before 0.15.27. Time sure does fly.

Re: 3 and 4 way intersections

Posted: Thu Nov 18, 2021 7:55 pm
by hansjoachim
Koub wrote: ↑
Wed Nov 17, 2021 8:04 pm
 Haha I feel stupid. I confess I have almost never used trains, I guess the last time was before 0.15.27. Time sure does fly.
it does=)

Re: 3 and 4 way intersections

Posted: Sat Nov 20, 2021 12:22 am
by farcast
After looking at junctions for a while, especially the multicross junction, I thought of seeing what would happen if I just spaced out that triple crossing you see in a lot of T-junctions. This "Double Image" junction is the result.
Double Image
RHD Scores
Set 1: 53
Set 2: 59*

You can improve the set 1 score to 57 by breaking the exit block into a 2 car block and a 4-5 car block after it. If you do this, then there needs to be another 6 car block afterwards, as well as only using 5-6 car trains, to prevent deadlocks. I chose to list the score for the more general use case.

* The set 2 score is heavily dependent on two things. The signal spacing leading up to the junction needs to match the length of an inner buffer, and the three inner buffers need to be near exactly the same size. When these two conditions are met, then crossings will synchronize in an extremely satisfying way. My efforts to disrupt the timing revealed this behavior to be stable, meaning crossings will naturally synchronize like this. If either condition isn't met, then this behavior will be unstable, crossings won't synchronize, and set 2 performance will drop to 43.
Synchronized crossing.gif
Synchronized crossing.gif (164.77 KiB) Viewed 83489 times
Extending the buffers to fit 2 trains only improved set 1 TPM from 53 to 57, and had no effect on set 2 TPM, so I decided not to include that since splitting the exit block could get the same benefit, though I haven't tested using both.

Using oversized 1 train buffers increased the set 1 score to 55, but crossing behavior became unstable. I tested again after increasing the signal spacing from 6 cars to matching the inner buffer length, and crossing behavior became stable again. The set 2 score after increasing the signal spacing was 63 TPM while the set 1 score dropped back to 53. At this point I realized ~63 TPM might be the hard limit for crossing performance without using multiple parallel crossing lanes for each entrance, as there was never not a train passing the chain signals at near full speed.

Bonus RHD Double Double Image scores (without split exits)
Set 1: 54
Set 2: 55
Set 3: 74

Blueprints are for 6 car trains and 7 car 27 rail tile signal spacing, but would also work with 6 car signal spacing. After figuring out that I can switch between set 1 and set 2 without waiting for trains to clear out, I found out that the tolerance for self synchronization is even less than I thought. Signal spacing really needs to be near exactly the same as the buffer length. I think it might actually be buffer length + chain signal blocks, but in this case there's not much difference.

Re: 3 and 4 way intersections

Posted: Sat Nov 20, 2021 3:41 am
by mrvn
Wouldn't it be better to have 4 lanes between the 2 Ts?

Re: 3 and 4 way intersections

Posted: Sat Nov 20, 2021 8:20 am
by farcast
mrvn wrote: ↑
Sat Nov 20, 2021 3:41 am
 Wouldn't it be better to have 4 lanes between the 2 Ts?
Good idea! I was going for minimal buffers, but since buffers need to be there anyway, I might as well make good use of the space.
Double Double Image v2
New RHD scores
Set 1: 62
Set 2: 56
Set 3: 77

Maybe there's a better way to do it, but it's hard to make changes without making crossings unstable. I get the feeling this concept doesn't scale very easily, if at all.

Re: 3 and 4 way intersections

Posted: Sat Nov 20, 2021 7:34 pm
by Kano96
Hey, nice intersection! The synchronized crossings are mesmerizing.
Do you want to have it added to the post? If yes, which version? V2 obviously performs better, but I also like the simplicity of the original.

Re: 3 and 4 way intersections

Posted: Sat Nov 20, 2021 9:37 pm
by mmmPI
very inspiring :)

Re: 3 and 4 way intersections

Posted: Sat Nov 20, 2021 11:47 pm
by farcast
Kano96 wrote: ↑
Sat Nov 20, 2021 7:34 pm
 Hey, nice intersection! The synchronized crossings are mesmerizing.
Do you want to have it added to the post? If yes, which version? V2 obviously performs better, but I also like the simplicity of the original.
I would love for it to be added, thanks! I think v2 should be used, since it allows for simultaneous left & right turns from one T to the next, which is situationally better. It still has synchronized crossings, so it's just better all around.

Re: 3 and 4 way intersections

Posted: Sun Nov 21, 2021 7:04 am
by DaveMcW
Branch Predictor

4-Way, 2 Lane, Unbuffered.

Image



This intersection attempts to keep a pair of compatible lanes running at high speed as long as possible. It works best when many trains in a row are going to the same destination.

The main intersection is 58x58, but the train detector requires the 7-tile-spaced signals to extend past the end of the first train. So the blueprint for 6 wagon trains is 114x114. There is probably room to optimize these numbers by adjusting the train detector.

Re: 3 and 4 way intersections

Posted: Sun Nov 21, 2021 3:12 pm
by Factoriointersection
DaveMcW wrote: ↑
Sun Nov 21, 2021 7:04 am
 Branch Predictor

4-Way, 2 Lane, Unbuffered.

Image



This intersection attempts to keep a pair of compatible lanes running at high speed as long as possible. It works best when many trains in a row are going to the same destination.

The main intersection is 58x58, but the train detector requires the 7-tile-spaced signals to extend past the end of the first train. So the blueprint for 6 wagon trains is 114x114. There is probably room to optimize these numbers by adjusting the train detector.
That is pretty cool:) it gets a score of 53 which is better than an 2 lane unbuffered intersections. Regulary turning signals of, with combinators, causes trains to repath and try to avoid the intersection.

Re: 3 and 4 way intersections

Posted: Sun Nov 21, 2021 5:59 pm
by Lubricus
I have an design that I haven't seen any other use and shuldn't be to bad. 4-way unbufferd
https://factorioblueprints.tech/api/str ... c6df053472

https://factorioblueprints.tech/bluepri ... 450d3521dd

Re: 3 and 4 way intersections

Posted: Sun Nov 21, 2021 8:08 pm
by causa-sui
2-lane, 4-way buffered by Tallinu. RHD.



Results in testbench 4.2.1, factorio 1.1.46 (build 59110, linux amd64): 45/48/45/46
All times are UTC
Page 1 of 15

Powered by phpBB® Forum Software © phpBB Limited