Factorio Forums

datarza wrote: ↑

Mon Dec 04, 2023 11:47 pm

Sorry, can not understand the problem, please explain. I can confirm, that turning to the right (from top) blocks the train, which turns to left (from bottom), it is expected behaviour for unbuferred intersections (and for buffered probably as well).

Why not just use the Wide instead? The only advantage I can see for your intersection is that it's slightly smaller, but space isn't a big deal in Factorio.

Avona wrote: ↑

Tue Dec 05, 2023 12:04 am

datarza wrote: ↑

Mon Dec 04, 2023 11:47 pm

Sorry, can not understand the problem, please explain. I can confirm, that turning to the right (from top) blocks the train, which turns to left (from bottom), it is expected behaviour for unbuferred intersections (and for buffered probably as well).

Why not just use the Wide instead? The only advantage I can see for your intersection is that it's slightly smaller, but space isn't a big deal in Factorio.

In general, you have answered the self-question and a little more. Wide B is a bigger intersection, f.e. Foundation and Celtic Nope are much more compact and have the same score. Straights are just the compatibility bonus, put intersection over existing rails. In additional, Wide W does not have U-Turn, therefore, also does not have enter/exit to cityblock from all sides. I do not know how Wide W can be used as 4-way and 6-way intersection, but I know how this discussed intersection can (with some limitation).

What is the purpose of your pickiness?

datarza wrote: ↑

Tue Dec 05, 2023 3:04 am

What is the purpose of your pickiness?

Throughput.

Avona wrote: ↑

Tue Dec 05, 2023 3:17 am

datarza wrote: ↑

Tue Dec 05, 2023 3:04 am

What is the purpose of your pickiness?

Throughput.

and how is it related to previous questions? So, what is the score? Not a big difference in comparing to the original SQUAREABOUT, right?

datarza wrote: ↑

Tue Dec 05, 2023 4:13 am

Avona wrote: ↑

Tue Dec 05, 2023 3:17 am

datarza wrote: ↑

Tue Dec 05, 2023 3:04 am

What is the purpose of your pickiness?

Throughput.

and how is it related to previous questions? So, what is the score? Not a big difference in comparing to the original SQUAREABOUT, right?

So you posted your intersection and didn't ask for any feedback, you just wanted to share and that is good:) We don't need to critics it unless you want to. In general I'll say that the score difference between intersections with blocking opposing turns vs non blocking is substantial but it only matters if the intersections is causing a bottleneck in your train network. You can test it on the testbench if you want to see the impact of changes to the intersection.
viewtopic.php?f=194&t=94795

hansjoachim wrote: ↑

Tue Dec 05, 2023 11:54 am

So you posted your intersection and didn't ask for any feedback, you just wanted to share and that is good:) We don't need to critics it unless you want to. In general I'll say that the score difference between intersections with blocking opposing turns vs non blocking is substantial but it only matters if the intersections is causing a bottleneck in your train network. You can test it on the testbench if you want to see the impact of changes to the intersection.
viewtopic.php?f=194&t=94795

Criticism is good, I want it. Pointing twice at a non-existent issue and confusing left and right is something else. Anyway, thank you for a link.

datarza wrote: ↑

Tue Dec 05, 2023 6:49 pm

Criticism is good, I want it. Pointing twice at a non-existent issue and confusing left and right is something else. Anyway, thank you for a link.

The issue is not non-existent, it is a good advice to improve the junction to avoid blocking-turn, it's not easy to see but most of the junctions on this thread differs from yours on that particular point, the left-right difference is not really a problem, the shape of the rails makes it so that train coming from opposite end cannot both turn in the same direction at the same time without one having to wait for the other one. The "difficult turn" is the "turn left" when train drive on the right side, or "turn right" when trive drive on the left side.

You blueprint is signaled so that train drive on the left side, and if two train coming from opposite direction wants to "turn right" at the same time, they will have to cross the path of each other, forcing one of them to slow or stop.

Sorry if that is not the criticism you expected but it seems like you would have been left with a misunderstanding.

mmmPI wrote: ↑

Tue Dec 05, 2023 7:21 pm

You blueprint is signaled so that train drive on the left side, and if two train coming from opposite direction wants to "turn right" at the same time, they will have to cross the path of each other, forcing one of them to slow or stop.

Yes, it is true. How I mentioned before, this behaviour comes from SQUAREABOUT design.

I dawned now, do you want to hint that this bahaviour can be changed in my design? Honestly, I am not sure if it possible.

Just a driveby here, but...

I believe datarza wants a roundabout design that can be laid over existing tracks without the need to remove any of the existing. As such, I believe what is happening may be inevitable.

That said, perhaps the discussion can be more about 1) why they want an intersection that doesn't require removing existing tracks (bots trivialize this, don't they?), and 2) the pros and cons of roundabouts (if age old memory serves, these increase complexity and thus overhead with pathfinder, plus they lock the maximum size of a train that can be used in the network without a full redesign).

But maybe in their original thread about the design would be a better place to discuss than here?
viewtopic.php?f=194&t=105361

datarza wrote: ↑

Tue Dec 05, 2023 8:05 pm

I dawned now, do you want to hint that this bahaviour can be changed in my design? Honestly, I am not sure if it possible.

I don't know if it is possible, it is difficult to see if signals would fit with additionnal rails.

The difference in throughput may not be very significant, but if you could make the trains turn before they go past the middle power pole it could allow for trains coming from opposite direction to both "turn right" without interfering which would give better result on the test.

The signals are placed correctly, you seem to not need advice there, so maybe that's why the critics is only about the blocking-turn, it's the little details

FuryoftheStars wrote: ↑

Tue Dec 05, 2023 9:00 pm

Just a driveby here, but...

I believe datarza wants a roundabout design that can be laid over existing tracks without the need to remove any of the existing. As such, I believe what is happening may be inevitable.

Not sure which rails would be considered mandadory, if you divide the juction in 4 corner that are symetric then a train should only use 2 when turning left or right or going straight so that the train coming from the opposite end can use the other 2. ( sometime you have to divide in diagonal ). It create limit for geometry.

FuryoftheStars wrote: ↑

Tue Dec 05, 2023 9:00 pm

That said, perhaps the discussion can be more about 1) why they want an intersection that doesn't require removing existing tracks (bots trivialize this, don't they?), and 2) the pros and cons of roundabouts (if age old memory serves, these increase complexity and thus overhead with pathfinder, plus they lock the maximum size of a train that can be used in the network without a full redesign).

But maybe in their original thread about the design would be a better place to discuss than here?
viewtopic.php?f=194&t=105361

1) i had not understood this, this is personnal choice imo.

2) there is also an increase risk of dead lock with the U-turn junctions, if you have 2 of them, it create a circle/loop and if this circle gets filled with trains it can then get stuck. This require at least 4 junctions for the same risk to occurs in cityblock, but only 2 if they have U-turns ( like roundabouts). It's quite minor risk but it increases when junctions are close to each other in the network.

Roundabouts,i remember had the risk of long train crashing into itself when repathing, but i think this has been changed since then, newer thread have newer information .

mmmPI wrote: ↑

Tue Dec 05, 2023 10:59 pm

2) there is also an increase risk of dead lock with the U-turn junctions, if you have 2 of them, it create a circle/loop and if this circle gets filled with trains it can then get stuck. This require at least 4 junctions for the same risk to occurs in cityblock, but only 2 if they have U-turns ( like roundabouts). It's quite minor risk but it increases when junctions are close to each other in the network.

Sorry, could you please explain this in detail? So, city blocks have four intersections in each corner. And if two of these intersections have a U-turn feature (like a roundabout) and the other two don't have a U-turn feature, could this be a source of dead blocks if there is not enought space for train? Maybe a drawing could be more explanatory? But, as usual, we are planing cityblocks with the same intersection in corners, so problem you mentioned can not be in cityblocks with four u-turn intersections?

FuryoftheStars wrote: ↑

Tue Dec 05, 2023 9:00 pm

Just a driveby here, but...

I believe datarza wants a roundabout design that can be laid over existing tracks without the need to remove any of the existing. As such, I believe what is happening may be inevitable.

That said, perhaps the discussion can be more about 1) why they want an intersection that doesn't require removing existing tracks (bots trivialize this, don't they?), and 2) the pros and cons of roundabouts (if age old memory serves, these increase complexity and thus overhead with pathfinder, plus they lock the maximum size of a train that can be used in the network without a full redesign).

But maybe in their original thread about the design would be a better place to discuss than here?
viewtopic.php?f=194&t=105361

No, they fixed trains so they no longer repath while inside the roundabout and then drive into themself. So you can now have a minimal roundabout and even trains with 16 cargo wagons won't do a 270° turn on them.

Once the train enters a chain signal section it will continue it's current path until it passes a full signal and only then it can go a different way. So unless your roundabout requires the train to do a 270° turn from the start (no 90° shortcut) it will be safe.

datarza wrote: ↑

Tue Dec 05, 2023 11:56 pm

mmmPI wrote: ↑

Tue Dec 05, 2023 10:59 pm

2) there is also an increase risk of dead lock with the U-turn junctions, if you have 2 of them, it create a circle/loop and if this circle gets filled with trains it can then get stuck. This require at least 4 junctions for the same risk to occurs in cityblock, but only 2 if they have U-turns ( like roundabouts). It's quite minor risk but it increases when junctions are close to each other in the network.

Sorry, could you please explain this in detail? So, city blocks have four intersections in each corner. And if two of these intersections have a U-turn feature (like a roundabout) and the other two don't have a U-turn feature, could this be a source of dead blocks if there is not enought space for train? Maybe a drawing could be more explanatory? But, as usual, we are planing cityblocks with the same intersection in corners, so problem you mentioned can not be in cityblocks with four u-turn intersections?

Any time you have a circular path you can count the number trains that fit onto that circle (one per full signal unless they are spaced too close). If you ever manage to get that number of trains into the circular path with all of them wanting to continue going along the circle then they deadlock. Minimal example:
Both trains want to make an u-turn but they can't enter the roundadbout because the other train prevents them exiting it. A design with u-turns makes for some really small circular path. Without u-turns the train would have to go around a full block, which would increase the circluar path by 4. So the minimum number of trains to deadlock goes from 2 to 6.

Note: This holds true for any circular path in your network. In a city grid you will have tons and tons of such circular paths, some smaller going just between two roundabouts. Some larger spanning multiple city blocks. Trains can also re-path when they are stuck at a chain signal. So instead of a u-turn one of the above trains might decide to go the long way and drive around a full city block. They way is longer but there isn't a train in the way blocking it's progress. So the path finder gets a lower cost for the longer path, eventually, theoretically.

Note 2: Cost should go up the longer a train blocks a path from what the devs said. So in a grid design it should always un-deadlock over time but maybe I haven't waited long enough in the cases where I did get grid lock.

datarza wrote: ↑

Tue Dec 05, 2023 11:56 pm

Sorry, could you please explain this in detail? So, city blocks have four intersections in each corner. And if two of these intersections have a U-turn feature (like a roundabout) and the other two don't have a U-turn feature, could this be a source of dead blocks if there is not enought space for train? Maybe a drawing could be more explanatory? But, as usual, we are planing cityblocks with the same intersection in corners, so problem you mentioned can not be in cityblocks with four u-turn intersections?

The problem can be seen in the wiki :
https://wiki.factorio.com/Tutorial:Trai ... #Deadlocks

When "too many trains" are on a circle/loop, it can create a deadlock where all the blocks are occupied :


This can happen around 1 city block, because it create a square/circle if you count the 4 corners. But this is quite rare, it means you have trains all around the cityblock( all 4 side are busy).
If the junctions have U-turns, it is possible for trains to get stuck only using 1 side of the cityblock because it create smaller loops, no need for 4 corner, only 2 , it is more risky for small blocks because it makes it possible to create deadlock with only 2 trains when very unlucky.

Thanks to all for the clarification.

mrvn wrote: ↑

Wed Dec 06, 2023 12:14 am

FuryoftheStars wrote: ↑

Tue Dec 05, 2023 9:00 pm

Just a driveby here, but...

I believe datarza wants a roundabout design that can be laid over existing tracks without the need to remove any of the existing. As such, I believe what is happening may be inevitable.

That said, perhaps the discussion can be more about 1) why they want an intersection that doesn't require removing existing tracks (bots trivialize this, don't they?), and 2) the pros and cons of roundabouts (if age old memory serves, these increase complexity and thus overhead with pathfinder, plus they lock the maximum size of a train that can be used in the network without a full redesign).

But maybe in their original thread about the design would be a better place to discuss than here?
viewtopic.php?f=194&t=105361

No, they fixed trains so they no longer repath while inside the roundabout and then drive into themself. So you can now have a minimal roundabout and even trains with 16 cargo wagons won't do a 270° turn on them.

Once the train enters a chain signal section it will continue it's current path until it passes a full signal and only then it can go a different way. So unless your roundabout requires the train to do a 270° turn from the start (no 90° shortcut) it will be safe.

I wasn't thinking they still crashed into themselves, but rather that too long of a train might not be able to use them properly/cause a deadlock on themselves. But I might be wrong about that.

datarza wrote: ↑

Wed Dec 06, 2023 1:40 am

Thanks to all for the clarification.

You apparently edited your post between when I read the thread and attempted to reply, because this post is soo much shorter.

Anyway, I did want to note two things of what I saw you originally posted. First, the second of the two images you posted is solvable without a roundabout. But second, if you are approaching 10k city blocks, my first concern would be decreasing strain on the pathfinder (assuming I'm not mistaken on that, of course) from the roundabouts than being concerned about the trains taking a lap around the block to get going in the right direction. But, if your factory is still running well (UPS wise) and you like the roundabouts, then you do you.

(I will say that I have a personal bias against roundabouts because of rl. I used to work nearby to a roundabout and there's another large town/small city/whatever you want to call it that I often travel through that's been converting all of their signaled intersections to roundabouts lately. All I can say is that while these are fairly decent for the traffic wanting to enter the roundabout directly (putting aside the fact that most people in these areas apparently don't know how to drive in them), they're a F'ing PITA for traffic trying to enter the road downstream from them.)

FuryoftheStars wrote: ↑

Wed Dec 06, 2023 2:19 am

You apparently edited your post between when I read the thread and attempted to reply, because this post is soo much shorter.

Sorry, I wanted to think a little bit and move that question to another thread in future, but thank you for answer, appreciated. I would say, that city blocks do not need 2-way intersections. Honestly, cityblock needs 4-way (6-way is better) with feature to change the rail on intersection to all directions. These additianal ways can be used as train enter/exit to/from cityblock. Also, it is better to have left hand traffic, because right hand traffic consumes two additional outside cells for placing semaphores. That is why I worked on adoption of SQUAREABOUT (ROUNDABOUT has not so good cityblock's enter/exit position) to 4-way and 6-way and posted here my creation. U-turn can be an additional feature, that gives comfortable enter and exit. I am still thinking if it can harm a lot.

FuryoftheStars wrote: ↑

Wed Dec 06, 2023 2:19 am

mrvn wrote: ↑

Wed Dec 06, 2023 12:14 am

FuryoftheStars wrote: ↑

Tue Dec 05, 2023 9:00 pm

Just a driveby here, but...

I believe datarza wants a roundabout design that can be laid over existing tracks without the need to remove any of the existing. As such, I believe what is happening may be inevitable.

That said, perhaps the discussion can be more about 1) why they want an intersection that doesn't require removing existing tracks (bots trivialize this, don't they?), and 2) the pros and cons of roundabouts (if age old memory serves, these increase complexity and thus overhead with pathfinder, plus they lock the maximum size of a train that can be used in the network without a full redesign).

But maybe in their original thread about the design would be a better place to discuss than here?
viewtopic.php?f=194&t=105361

No, they fixed trains so they no longer repath while inside the roundabout and then drive into themself. So you can now have a minimal roundabout and even trains with 16 cargo wagons won't do a 270° turn on them.

Once the train enters a chain signal section it will continue it's current path until it passes a full signal and only then it can go a different way. So unless your roundabout requires the train to do a 270° turn from the start (no 90° shortcut) it will be safe.

I wasn't thinking they still crashed into themselves, but rather that too long of a train might not be able to use them properly/cause a deadlock on themselves. But I might be wrong about that.

The train pathing I think will still crash into itself. You can design a roundabout where a 90° turn isn't possible forcing the train to do a 270° turn and it will happily do it. Iirc the pathing will happily take the train through a block it already reserved / plans to reserve because sometimes that is the only way. You have to design your rail network so that the paths that loop over themself are never the cheapest / only option or such that the loop is always longer than the train.

The train will never deadlock though. It will just crash into itself. I think that is still the way it's implemented.

Note: the term roundabout is used loosely here. What is really mend is any network with a path that crosses itself. If the whole path has chain signals then it is save from re-pathing breaking your train. If the loop is larger than the train then it's safe from the train crashing into itself no matter what. An a train will never take a path that makes a complete loop and then takes an exit it could have used before.

So a roundabout is safe because it's all chain signals. A grid without u-turns is safe because the trains generally are shorter than the circumfrence of a city block. But if a 16 wagon ore hauler decides to enter your city grid that might no longer be the case and with the loops in a city grid containing signals (that or not chain signals) things can go BOOM.

PS: see viewtopic.php?f=6&t=110069 where I suggest they fix chain signals to consider train length and fix this issue.

I tested the Delta intersection by Hans Joachim, my setup can be seen in the attached image, and got the following results:

Set 1: 46.98
Set 2: 38.11
Set 3: 43.34
Score: 42.81 (42.8 ± 0.5 with uncertainty from https://mods.factorio.com/mod/Testbenchcontrols/faq)

I set the TPM goal to 29, and the test time to 240 min.

So my question is: Which measurement is correct, and why?
Also, it kind of bothers me that the 2-lane unbuffered 4-way intersection with the highest score is so asymmetrical. Additionally, I think Symmetrical Cross should be listed above Symmetrical, as it scores slightly higher for some spacings, but this is less important.