Page 23 of 36
Re: 4-way intersections: Throughput and deadlocks [image heavy]
Posted: Sat Jun 30, 2018 10:44 pm
by RulerOfHeck
Hello.
q
I was in need of a very compact 4 lane 4 way intersection, and I did not like the designs listed here: I think I can do better. I did one of my own but I could not make the test map work.
It's weird, a locomotive cannot find a path on the innermost lane,
unless the locomotive starts from half way up the test track?
EDIT: I found one way missing that I fixed - now half of the test trains works. Even though I copied the tracks with blueprints, so it should be completely symmetrical. Still, it works from all directions,
if I start a train from closer to the intersection. This is really annoying
This prevents me from comparing its performance against the other design - can someone please test it for me?
Code: Select all
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
Re: 4-way intersections: Throughput and deadlocks [image heavy]
Posted: Sun Jul 01, 2018 11:39 am
by hansjoachim
Hey! Welcome to this forum!
Test Rocket fuel
set 1 44
set 2 50
set 4 35 (left only)
Your intersection allows U-turns and lane switching, which means that throughput would be even worse if any trains would make a U-turn or switch lanes.
How it can improve
There are missing 4 signals, marked with a red dot
In the yellow area, south-west and north-east (and the opposite) are intersecting. Which makes it so that those two directions can't drive simultaneously.
The two rails marked with green allows lane switching.
If you don't need higher throughput than 120 trains per min, you should use a 2 lane system instead of 4.
Re: 4-way intersections: Throughput and deadlocks [image heavy]
Posted: Sun Jul 01, 2018 9:31 pm
by RulerOfHeck
Thank you!
That's a poor result
But you're right, there is nor reason why north/south shouldn't be able to do a left turn at the same time, I missed that
I (think that I) need lane changing because it will be part of a modular base; this will be in every corner of the modules. At the moment we are using extremely compact 2-lane intersections that only allows straight and right turns. It worked fine - until the traffic hit the ceiling. All the traffic from trains trying to do a left turn has created a situation where we cannot move enough trains to/from these 8-station modules. We need more tracks to hold the traffic, and shorter paths to enable the trains to get out of the way. Probably
I've attached a screenshot of the base; it's just getting started. The middle is the "seed" base that produces the first research and the items needed for constructing the real base.
I think I''ll add lane changing to the "christmas" design and use that for now, it might be enough
Re: 4-way intersections: Throughput and deadlocks [image heavy]
Posted: Wed Jul 11, 2018 6:15 am
by ballzak
Here's a compacted version of 4-lane MULTI-CROSS, with some crossings merged, fewer buffers.
- alt_screenshot_48801625.png (965.18 KiB) Viewed 7906 times
Re: 4-way intersections: Throughput and deadlocks [image heavy]
Posted: Sat Jul 14, 2018 7:32 am
by hansjoachim
Ballzak
What throughput are you getting with it?
Re: 4-way intersections: Throughput and deadlocks [image heavy]
Posted: Mon Jul 16, 2018 1:25 am
by ballzak
hansjoachim wrote:Ballzak
What throughput are you getting with it?
No exact measurements as i couldn't get the test save/map working properly.
Almost as good as the original MULIT-CROSS i'd guess, but that one is just too large for my base.
Re: 4-way intersections: Throughput and deadlocks [image heavy]
Posted: Mon Jul 16, 2018 6:49 am
by hansjoachim
Ballzak
I doubt it
Re: 4-way intersections: Throughput and deadlocks [image heavy]
Posted: Tue Jul 17, 2018 5:37 am
by Tallinu
Actually, that layout looks the same as one of the layouts I built. Though I haven't examined his blueprint or compared it to mine in detail, and I wouldn't expect it to be a perfect match regardless.
I know I hadn't posted all of the variations I'd come up with for the four-lane MC, since I'd been intending to do that once I had a full set of size variations and such, but life got busy and I didn't end up getting back to it.
Point is, that more compact layout, at least in my testing on the version I made, has a surprisingly low difference in throughput compared to the larger layouts with more buffers. It's a small enough that I doubt it would be significant outside a stress testing environment. So unless the signalling is terrible or something I would expect it to work just fine!
Re: 4-way intersections: Throughput and deadlocks [image heavy]
Posted: Tue Jul 17, 2018 1:58 pm
by hansjoachim
Ballzak, When you compressed the intersection, the intersection in the yellow area doesn't get filled every other turn. The intersection in the blue area sends one train and two trains to the yellow intersection every other turn, meaning that it won't regularly get three trains ready. They are of sync and I think that is limiting the throughput compared to the "official" Multicross 4 way 4 lanes 6 cars LHD.
I ran a quick test and got 140 trains/min with set 1. With the same size as you posted. The other one got 171 trains per min.
Re: 4-way intersections: Throughput and deadlocks [image heavy]
Posted: Wed Jul 18, 2018 2:05 am
by ballzak
hansjoachim wrote:Ballzak, When you compressed the intersection, the intersection in the yellow area doesn't get filled every other turn. The intersection in the blue area sends one train and two trains to the yellow intersection every other turn, meaning that it won't regularly get three trains ready. They are of sync and I think that is limiting the throughput compared to the "official" Multicross 4 way 4 lanes 6 cars LHD.
I ran a quick test and got 140 trains/min with set 1. With the same size as you posted. The other one got 171 trains per min.
Yes, including regular signals inside the crossings could possibly shave a few ms, but the crossings would have to become (1.5x) larger.
140 t/m seems pretty good for testing it with 6 cars, which it's not designed for.
Re: 4-way intersections: Throughput and deadlocks [image heavy]
Posted: Wed Jul 18, 2018 2:22 am
by ballzak
Tallinu wrote:Actually, that layout looks the same as one of the layouts I built. Though I haven't examined his blueprint or compared it to mine in detail, and I wouldn't expect it to be a perfect match regardless.
I know I hadn't posted all of the variations I'd come up with for the four-lane MC, since I'd been intending to do that once I had a full set of size variations and such, but life got busy and I didn't end up getting back to it.
Point is, that more compact layout, at least in my testing on the version I made, has a surprisingly low difference in throughput compared to the larger layouts with more buffers. It's a small enough that I doubt it would be significant outside a stress testing environment. So unless the signalling is terrible or something I would expect it to work just fine!
After making lots of intersections i think the "MULTI-CROSS" variants are pretty optimal in terms of the fewest number of actual lane crossings and mergers, the only downside is their large size.
I'll keep on trying new designs, but i doubt there's any "magic" intersection which would get better throughput.
Re: 4-way intersections: Throughput and deadlocks [image heavy]
Posted: Thu Jul 19, 2018 5:28 am
by hansjoachim
Ballzak
Have you seen the latest intersections I posted on this forum? (Page 22)
Re: 4-way intersections: Throughput and deadlocks [image heavy]
Posted: Fri Jul 20, 2018 5:02 am
by ballzak
hansjoachim wrote:Ballzak
Have you seen the latest intersections I posted on this forum? (Page 22)
Which one? The last, it's humongous!
I suspect that splitting each lane into more than three, as in the three possible directions, is pointless as you'll get bottlenecks where they merge anyway.
Re: 4-way intersections: Throughput and deadlocks [image heavy]
Posted: Fri Jul 20, 2018 1:13 pm
by hansjoachim
ballzak wrote:
I suspect that splitting each lane into more than three, as in the three possible directions, is pointless as you'll get bottlenecks where they merge anyway.
I guess you haven't seen the merger I made:P It can compress trains up to 33,5 out of a max of 34.(2-4 trains)
https://www.reddit.com/r/factorio/comme ... ng_solved/
As they crush the throughput of every other intersection I don't think it is pointless=)
They are quite big and doesn't beat the multicross well enough in test 1 and 2.
They do win on Left only though(RHD) were the throughput is more than double on the largest ones.
Re: 4-way intersections: Throughput and deadlocks [image heavy]
Posted: Wed Jul 25, 2018 7:17 am
by Tallinu
Yeah, it's pretty much always a trade-off between train throughput and junction footprint (size). Increasing the first inevitably tends to require increasing the second, and shrinking it tends to slow it down.
Optimizing designs to achieve reasonable throughput combined with minimal sprawl for practical usage scenarios is the true purpose, although some of us (like me) have certainly had fun just pushing the limits of the max throughput barrier without much regard to how much space is required (although personally I always preferred to try to keep my designs as tidy and elegant as possible and to avoid careless wasting of space).
And when someone finds the results useful or interesting... so much the better.
Re: 4-way intersections: Throughput and deadlocks [image heavy]
Posted: Thu Jul 26, 2018 10:32 am
by mrvn
Tallinu wrote:Yeah, it's pretty much always a trade-off between train throughput and junction footprint (size). Increasing the first inevitably tends to require increasing the second, and shrinking it tends to slow it down.
Optimizing designs to achieve reasonable throughput combined with minimal sprawl for practical usage scenarios is the true purpose, although some of us (like me) have certainly had fun just pushing the limits of the max throughput barrier without much regard to how much space is required (although personally I always preferred to try to keep my designs as tidy and elegant as possible and to avoid careless wasting of space).
And when someone finds the results useful or interesting... so much the better.
Then how about building a cross that has 1000 train buffers for every way and give right of way for whatever buffer is fullest till it's empty?
Re: 4-way intersections: Throughput and deadlocks [image heavy]
Posted: Fri Aug 10, 2018 9:27 pm
by JackGruff
How does one judge where to put signals outside the intersection?
Or perhaps a better question, can I mess the intersection up or degrade its effectiveness by placing signals to close or too far?
Re: 4-way intersections: Throughput and deadlocks [image heavy]
Posted: Mon Aug 13, 2018 9:02 am
by mrvn
JackGruff wrote:How does one judge where to put signals outside the intersection?
Or perhaps a better question, can I mess the intersection up or degrade its effectiveness by placing signals to close or too far?
Crucial rule is this: After an intersection there must be a full signal with enough space for a full train beyond it. Anything else means a train can stop mid intersection. If there isn't enough space till the next intersection then you need chain signals and both intersection becomes one big intersection.
Otherwise in between intersections you can place as many signals as you like but placing multiple chain signals on a straight is pointless. The closeness of full signals determines how close trains can follow each other. So I guess placing signals everywhere you can would allow optimal (minimal) distance between trains.
On the other hand the crucial rule already added a minimum distance of one train length between trains. That gives you a worst case minimum time between trains when trains start from stand still to enter the intersection. If you do some math you can compute the train speed at each tile and see how many signals you need to keep that time between trains along the way. If that's even possible.
As someone pointed out previously: Since breaking distance isn't linear with speed there is a optimal speed that gives the maximum trains per minute. Placing signals and wiring them to keep trains at that speed might gain you more throughput.
Personally I wouldn't place signals closer than a train length apart and on long distances I do way less.
Re: 4-way intersections: Throughput and deadlocks [image heavy]
Posted: Fri Sep 28, 2018 3:10 pm
by immortal_sniper1
https://factorioprints.com/view/-LG-kkilxJCqOZ398oze
design to be tested
in my its been like 70h and no deadlock (train out of fuel doesnt count )
Re: 4-way intersections: Throughput and deadlocks [image heavy]
Posted: Sat Sep 29, 2018 7:28 am
by hansjoachim
Immortal sniper could you provide a picture of it?