4-way intersections: Throughput and deadlocks [image heavy]
Posted: Sun May 07, 2017 6:20 pm
by aaargha
0.16 rework in progress
Now that 0.16 feels like it is starting to stabilize it's time to revise parts of this post, mostly regarding deadlocks. As it seems like the pathfinding changes made in 0.15.27 are going to stick I'm going to change the deadlock ratings as all the builds I've seen so far can be signalled to be safe, as long as long as you don't disable stations/break paths.
First out is the updated throughput setup. The new and improved version for 0.16 can be found just above the results.
Regarding the 0.15.27 path-finding changes
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.
However, as there are still some quirks with the current implementation (to be fixed in 0.16), I will wait for those changes to to be made before I make any drastic changes to this post.
Test details
As a part of trying to break things better understand trains I've started to build a test bench for 4-way intersections. What I thought I'd do here is share the results of my tests and allow people to submit new designs for testing. The tests will currently cover deadlock analysis and, unless it's too deadlocky, throughput. I'll also try to do some signal optimization (mostly pruning redundant signals) and add some notes on if there are things that need to be adjusted when changing train length.
Deadlock analysis
This is pretty much a worst case analysis, the only thing I assume about the rest of the train system is that the output block is properly sized. Trains are allowed to change path at the worst time, basically I'll do my best to make it break. While this is mostly manual analysis I will provide examples of any deadlock from my tests if I'm able.
Here is how I grade how "deadlockable" an intersection is.
A - Safe, will not be the source of a deadlock
B - Can deadlock if more than one train changes path inside the intersection, may resolve itself if output block becomes free
C - Can deadlock if more than one train changes path inside the intersection, requires player to resolve
D - Can deadlock if a train changes path in the intersection, may resolve itself if output block becomes free
E - Can deadlock if a train changes path in the intersection, requires player to resolve
F - Can deadlock in regular traffic, usually caused by loops that can be saturated
Note that all this goes out the window if you don't use large enough output blocks.
Also note that in reality B pretty much as good as A. Any problems will be very rare (unless you're super unlucky with your rail system design) and are likely to resolve themselves pretty quickly. C is very close behind as problems are still exceedingly rare, they may require the player to step in though.
Throughput testing
For throughput testing I've written a mod to spawn in trains as needed to ensure a good enough saturation of all input lanes (about 29.35 trains per minute for each input lane, which should be full saturation as the trains can't travel closer to each other). The trains have a randomized destination to make them go straight or turn left/right, the probability of each direction can be controlled by assigning different direction different weights. Each test consists of one run of each set:
Set 1:
Left: 33.3%
Right: 33.3%
Straight: 33.3%
Set 2 (for left hand drive)
Left: 35%
Right: 15%
Straight: 50%
Set 2 (for right hand drive)
Left: 15%
Right: 35%
Straight: 50%
Set 1 is equal probability. Set 2 is to simulate a more "realistic" usage situation where the player has planned the routes a bit.
All tests are done using fully upgraded 2-4 trains running on rocket fuel.
I let it run for 15 min while counting how many trains exit the intersection. When the time is up I calculate the average throughput in trains/min. Note that this is not too exact and small measured differences between intersections may not really mean that much.
For intersections that require lane changers I'm using these that are designed by Tallinu. (Note that I only have 4-lane versions so if your 6-/8-lane design needs lane changers you need to design them yourself)
Notes
If there is a design you want me to test that is not already covered just give me a blueprint string of your intersection. I'm always looking for more meat for the grinder new and interesting designs.
Thought of a good test to add, let me know and I'll see if I can make it work.
Are you the creator of any of the designs here? Most of these are intersections I've tested for various threads so just shoot me a link to the original post and I'll add some credits.
Try it for yourself
If you want to try things out for yourself take a look at the attached save below. For usage instructions see the thread for the mod or the thread linked below.
V2 for 0.16
Controls the same as the previous (signal P selects profile turn off -10k green to start) and only needs Automatic train deployment 0.1.5 to run.
Cleaned up the map and increased the size of the test area to about 2000x2000 tiles. As the distances are much larger now I've added some scripts to teleport the player to different positions, use the commands below to get around.
A savefile with an older version of the test setup, among lots of other tests, can be found here.
Results
Grouped by deadlock rating and roughly sorted by throughput. Names are basically just some bullshit to keep them apart. To see the full images (rendered at 0.5 zoom) just open them in a new tab.
[quote="7 may 2017"]
Initial post. Added: A: Cross, Shuriken, Compact, Wide, 2x 3-way, Knot, R-twister, Roundabout
B: Wide crossover, Rotary, Curly
C: Squareabout
D: Roundabout[/quote]
[quote="8 may 2017"]Fixed BP strings for Shuriken and Wide.
Added RHD variants for Cross, Curly, Knot, Twister, Wide and Wide crossover[/quote]
[quote="9 may 2017"]Added deadlock examples.
Added RHD variant of Shuriken.
Added Direct cross.[/quote]
[quote="10 may 2017"]Added RHD variants for: Compact and Rotary.
Added LHD variant of Squareabout.
Credited impetus maximus for Wide crossover
Tested: Compact Celtic knot[/quote]
[quote="11 may 2017"]Tested: Compact attempt, Crossabout and Wide B[/quote]
[quote="15 may 2017"]Tested: 2x high-perf-3-way[/quote]
[quote="17 may 2017"]Tested: Compact cross, Double diamond, Crushed Celtic knot and Crushed wide
Moved info above pictures as people seemed to miss the RHD blueprint strings
Hunted down some sources[/quote]
[quote="30 may 2017"]Tested: Chunk-aligned cross[/quote]
[quote="09 jun 2017"]Updated results to new throughput test system
Added 8-lane: Christmas, Square twister
Added 6-lane: Christmas
Added 4-lane: Adapted wide, Flower, Flower MKII, Semi-cross 8-car, Simple
Added 2-lane: Cloverleaf, Compact spiral[/quote]
[quote="10 jun 2017"]Added 8-lane: Flower MK3, Compact christmas
Added 4-lane: Christmas, Parentheses[/quote]
[quote="11 jun 2017"]Added 4-lane: Alternating cross, Doubleabout
Added 2-lane: Super compact Celtic knot
Fixed deadlock example for Parentheses[/quote]
[quote="12 jun 2017"]Fixed signalling error for Super compact Celtic knot (LHD)[/quote]
[quote="15 jun 2017"]Added 2-lane: Pleasing, Windmill[/quote]
[quote="20 jun 2017"]Added 2-lane: Buffered compact attempt, Buffered roundabout[/quote]
[quote="26 jun 2017"]Added 4-lane: Fresh cross MK2, Wavy
Added 2-lane: Double wide bubble cross[/quote]
[quote="27 jun 2017"]Added 2-lane: Spinabout[/quote]
[quote="29 jun 2017"]Added 4-lane: Encircled knot
Added 2-lane: Buffered roundabout MK2, Spinabout v3[/quote]
[quote="25 jul 2017"]Added 4-lane: Double squareabout, Multi-cross
Added 2-lane: Inscribed square, Windmill MK2[/quote]
[quote="27 jul 2017"]Added 4-lane: New Gauge Interchange[/quote]
[quote="28 jul 2017"]Added 8-lane: Multi-cross[/quote]
[quote="29 jul 2017"]Added 4-lane: Proper doubleabout[/quote]
[quote="30 jul 2017"]Added 4-lane: Switching Christmas[/quote]
[quote="06 aug 2017"]Added 2-lane: Encicled compact, Minimal, Spiral
Added 4-lane: Spiral
Added 8-lane: Spiral[/quote]
[quote="09 aug 2017"]Implemented improved spawner designed by Tallinu
Added 2-lane: Reversed roundabout[/quote]
[quote="16 aug 2017"]Added 2-lane: Improved-er reverse roundabout[/quote]
[quote="27 aug 2017"]Added 4-lane: Quadabout[/quote]
[quote="25 sep 2017"]Added 2-lane: Windcross[/quote]
[quote="30 sep 2017"]Added 2-lane: Braid, Pasta, Propeller
Enfoced output block size on Inscribed square[/quote]
[quote="28 oct 2017"]Added 2-lane: Magic-roundabout, Multi-Cross, Priority roundabout, Whirlpool
Added 8-lane: Reverse roundabout[/quote]
[quote="11 nov 2017"]Added 4-lane: Traditional[/quote]
[quote="12 nov 2017"]Added 2-lane: Pasta MKII, Pasta MKIII[/quote]
[quote="12 nov 2017"]Added 4-lane: Almost squareabout[/quote]
[quote="28 nov 2017"]Adjusted deadlock rating on Double squareabout
Added 2-lane: No Gods, No left
Added 4-lane: Almost squareabout OCD[/quote]
[quote="09 jan 2018"]Added V2 of the throughput tester[/quote]
8-lane A
Multi-cross
Blueprint string: Left hand drive
Throughput: Set 1: 352 Set 2: 352
Scaling notes: More buffers than I care to count need to be scaled for train size.
Version tested: 0.15.31
Original by: Tallinu
Flower MK3
Blueprint string: Right hand drive
Throughput: Set 1: 183 Set 2: 159
Scaling notes: More buffers than I care to count need to be scaled for train size.
Version tested: 0.15.18
Original by: 6180339887
Christmas
Blueprint string: Left hand drive
Throughput: Set 1: 87 Set 2: 96
Version tested: 0.15.18
Original by: aaargha
Compact christmas
Blueprint string: Right hand drive
Throughput: Set 1: 85 Set 2: 85
Version tested: 0.15.18
Original by: 6180339887
Square twister
Blueprint string: Left hand drive
Throughput: Set 1: 48 Set 2: 60
Deadlock notes: Some stopping possible due to path changing, but I'm pretty sure it won't deadlock.
Version tested: 0.15.18
Original by: Rapier31
B
Spiral
Blueprint string: Left hand drive
Throughput: Set 1: 55 Set 2: 64
Deadlock rating: B
Version tested: 0.15.32
Original by: SirCaptainCrumpet
D
Reverse roundabout
Blueprint string: Right hand drive
Throughput: Set 1: 84 Set 2: 92
Scaling notes: All the buffers need to be adjusted to fit the longest trains, posted is 6 car
Version tested: 0.15.37
Original by: ridesdragons
6-lane
Christmas
Blueprint string: Right hand drive
Throughput: Set 1: 84 Set 2: 86
Deadlock rating: A
Version tested: 0.15.18
Original by: aaargha
4-lane A
Multi-cross
Blueprint string: Left hand drive (untested 14-car RHD variant)
Throughput: Set 1: 171 Set 2: 155
Scaling notes: all buffers should be scaled for longest train, posted is 6-car
Version tested: 0.15.30
Original by: Tallinu
Flower MKII
Blueprint string: Left hand drive
Throughput: Set 1: 101 Set 2: 100
Scaling notes: all 96 buffers should be scaled for longest train, posted is 6-car
Version tested: 0.15.18
Original by: aaargha
Flower
Blueprint string: Left hand drive
Throughput: Set 1: 88 Set 2: 99
Scaling notes: all buffers should be scaled for longest train, posted is 6-car
Version tested: 0.15.18
Original by: aaargha
Semi-cross 8-car
Blueprint string: Right hand drive
Throughput: Set 1: 84 Set 2: 93
Deadlock notes: Some stopping possible due to path changing, but I'm pretty sure it won't deadlock.
Scaling notes: Tailored to 8-car trains, most of the intersection would have to be resized to fit larger trains
Version tested: 0.15.18
Original by: VFaalcatnodriiro
Alternating cross
Blueprint string: Left hand drive
Throughput: Set 1: 73 Set 2: 80
Scaling notes: all buffers should be scaled for longest train, posted is 6-car
Version tested: 0.15.18
Original by: Tallywort
Traditional
Blueprint string: Right hand drive
Throughput: Set 1: 61 Set 2: 71
Version tested: 0.15.37
Original by: Divaya
Fresh cross MK2
Blueprint string: Right hand drive
Throughput: Set 1: 56 Set 2: 59
Deadlock notes: Some stopping possible due to path changing, but I'm pretty sure it won't deadlock.
Version tested: 0.15.23
Original by: Xtrafresh (submitted via PM)
Blueprint string: Right hand drive
Throughput: Set 1: 54 Set 2: 55
Deadlock notes: Some stopping possible due to path changing, but I'm pretty sure it won't deadlock.
Version tested: 0.15.31
Original by: KuuLightwing
Wavy
Blueprint string: Left hand drive
Throughput: Set 1: 41 Set 2: 47
Deadlock notes: Some stopping possible due to path changing, but I'm pretty sure it won't deadlock.
Version tested: 0.15.23
Original by: Rapier31
Simple
Blueprint string: Right hand drive
Throughput: Set 1: 40 Set 2: 40
Deadlock notes: Some stopping possible due to path changing, but I'm pretty sure it won't deadlock.
Version tested: 0.15.18
Original by: ManimMond
Adapted wide
Blueprint string: Right hand drive
Throughput: Set 1: 35 Set 2: 35
Deadlock notes: Some stopping possible due to path changing, but I'm pretty sure it won't deadlock.
Scaling notes: The long part of the right turns can be made into its own block if the trains are shorter than 6 cars.
Version tested: 0.15.18
Original by: flyingaidsman
Blueprint string: Right hand drive
Throughput: Set 1: 45 Set 2: 45
Version tested: 0.15.18
Original by: Stumpyofpain
New Gauge Interchange
Blueprint string: Right hand drive
Throughput: Set 1: 36 Set 2: 44
Version tested: 0.15.31
Original by: THE kilroy
C
Almost squareabout OCD
Blueprint string: Right hand drive
Throughput: Set 1: 58 Set 2: 59
Version tested: 0.15.37
Original by: DarkElder
Almost squareabout
Blueprint string: Right hand drive
Throughput: Set 1: 56 Set 2: 55
Version tested: 0.15.37
Original by: DarkElder
Double squareabout
Blueprint string: Left hand drive
Throughput: Set 1: 52 Set 2: 59
Version tested: 0.15.30
Original by: 19wolf, Improved by Tallinu
D
Encircled knot
Blueprint string: Right hand drive
Throughput: Set 1: 44 Set 2: 50
Version tested: 0.15.23
Original by: Nexarius
Quadabout
Blueprint string: Right hand drive
Throughput: Set 1: 42 Set 2: 48
Version tested: 0.15.33
Original by: Geregori (submitted via PM)
Proper doubleabout
Blueprint string: Right hand drive
Throughput: Set 1: 40 Set 2: 48
Version tested: 0.15.31
Original by: f_luis
Doubleabout
Blueprint string: Right hand drive
Throughput: Set 1: 29 Set 2: 37
Version tested: 0.15.18
2-lane A
Pasta MKIII
Blueprint string: Right hand drive
Throughput: Set 1: 97 Set 2: 91
Scaling notes: All the buffers need to be adjusted to fit the longest trains, posted is 6 car
Version tested: 0.15.37
Original by: Zijkhal
Pasta MKII
Blueprint string: Right hand drive
Throughput: Set 1: 92 Set 2: 90
Scaling notes: All the buffers need to be adjusted to fit the longest trains, posted is 6 car
Version tested: 0.15.37
Original by: Zijkhal
Windmill MK2
Blueprint string: Left hand drive
Throughput: Set 1: 91 Set 2: 90
Scaling notes: All the buffers need to be adjusted to fit the longest trains, posted is 6 car
Version tested: 0.15.30
Original by: Tallinu
Multi-Cross
Blueprint string: All the different versions and read-me can be found here.
Throughput: Set 1: 88 Set 2: 89
Scaling notes: Available in 6-, 8-, 10- and 14-car variants. Tested is 6-car.
Version tested: 0.15.37
Original by: Tallinu
Inscribed square
Blueprint string: Left hand drive
Throughput: Set 1: 84 Set 2: 71
Scaling notes: All the buffers need to be adjusted to fit the longest trains, posted is 6 car
Version tested: 0.15.34
Original by: aaargha
Whirlpool
Blueprint string: Right hand drive
Throughput: Set 1: 72 Set 2: 73
Deadlock notes: Some stopping possible due to path changing, but I'm pretty sure it won't deadlock.
Scaling notes: All the buffers need to be adjusted to fit the longest trains, posted is 6 car
Version tested: 0.15.37
Original by: Tallinu
Pasta
Blueprint string: Right hand drive
Throughput: Set 1: 66 Set 2: 73
Scaling notes: All the buffers need to be adjusted to fit the longest trains, posted is 6 car
Version tested: 0.15.34
Original by: Zijkhal
Propeller
Blueprint string: Right hand drive
Throughput: Set 1: 62 Set 2: 64
Scaling notes: All the buffers need to be adjusted to fit the longest trains, posted is 6 car
Version tested: 0.15.34
Original by: Zijkhal
Windcross
Blueprint string: Left hand drive | Right hand drive
Throughput: Set 1: 63 Set 2: 59
Scaling notes: All the buffers before and after the intersection need to be adjusted to fit the longest trains, posted is 6 car
Version tested: 0.15.34
Original by: Zijkhal
Cross
Blueprint string: Left hand drive | Right hand drive
Throughput: Set 1: 58 Set 2: 62
Scaling notes: All the buffers before and after the intersection need to be adjusted to fit the longest trains, posted is 6 car
Version tested: 0.15.18
Original by: tallinu
Variants: 8-car by Zijkhal
Blueprint string: Left hand drive | Right hand drive
Throughput: Set 1: 51 Set 2: 54
Scaling notes: Buffers need to be scaled for longest train
Version tested: 0.15.21
Original by: RulerOfHeck
2x high-perf-3-way
Blueprint string: Left hand drive
Throughput: Set 1: 50 Set 2: 54
Scaling notes: Input and output buffers should be scaled for longest train
Version tested: 0.15.18
Spinabout v3
Blueprint string: Right hand drive
Throughput: Set 1: 47 Set 2: 53
Deadlock notes: Some stopping possible due to path changing, but I'm pretty sure it won't deadlock.
Scaling notes: Buffers need to be scaled for longest train which can be a bit tricky at times with how compressed the design is, posted is 6-car
Version tested: 0.15.23
Original by: AntiBlueQuirk
Spinabout
Blueprint string: Right hand drive
Throughput: Set 1: 47 Set 2: 52
Scaling notes: Buffers need to be scaled for longest train which can be a bit tricky at times with how compressed the design is, posted is 6-car
Version tested: 0.15.23
Original by: AntiBlueQuirk
Shuriken
Blueprint string: Left hand drive | Right hand drive
Throughput: Set 1: 48 Set 2: 50
Scaling notes: Buffers before the intersection should be scaled to longest train.
Version tested: 0.15.18
Blueprint string: Right hand drive
Throughput: Set 1: 39 Set 2: 37
Version tested: 0.15.18
Original by: leoch
2x 3-way
Blueprint string: Left hand drive
Throughput: Set 1: 36 Set 2: 38
Scaling notes: Middle part should be adjusted to longest train, rail signals on north and south exits are for 6-car trains
Version tested: 0.15.18
Blueprint string: Left hand drive | Right hand drive
Throughput: Set 1: 32 Set 2: 41
Deadlock notes: Signalling on the circle and the entrances/exits connected to it is pretty delicate, adding additional signals may make it very unstable.
Scaling notes: Blocks in the "circle" must fit longest trains, posted is for 7-car
Version tested: 0.15.23
Original by: aaargha (inspired by mrbaggins)
Blueprint string: Left hand drive
Throughput: Set 1: 20 Set 2: 19
Deadlock notes: Only safe as long as the circle is longer than the trains that will use it, otherwise it may crash.
Version tested: 0.15.18
B
Braid
Blueprint string: Right hand drive
Throughput: Set 1: 61 Set 2: 62
Deadlock notes: Needs 10 trains in the right place, two of which change path, to deadlock
Scaling notes: All the buffers need to be adjusted to fit the longest trains, posted is 6 car
Version tested: 0.15.34
Original by: Zijkhal
Blueprint string: Left hand drive | Right hand drive
Throughput: Set 1: 27 Set 2: 30
Deadlock notes: There exists at least a 4 train right (on RHD) to straight/left deadlock. There are probably more possibilities but this one is annoying as heck to reason about. Signalling is delicate, altering it may drastically worsen deadlock rating.
Scaling notes: Sides need to be scaled for longest train.
Version tested: 0.15.21
Original by: mrbaggins
Blueprint string: Left hand drive | Right hand drive
Throughput: Set 1: 38 Set 2: 41
Deadlock notes: Even though only one train needs to change path there needs to be another train already on the leaf for the deadlock to occur.
Scaling notes: The "leaves" need to be scaled to longest train.
Version tested: 0.15.18
Original by: Dr. Walrus (submitted via PM)
Blueprint string: Left hand drive
Throughput: Set 1: 39 Set 2: 39
Version tested: 0.15.32
Original by: maciekmm
No Gods
Blueprint string: Left hand drive
Throughput: Set 1: 36 Set 2: 39
Scaling notes: all buffers should be scaled for longest train, posted is 6-car
Version tested: 0.15.37
Original by: aaargha
Blueprint string: Left hand drive
Throughput: Set 1: 30 Set 2: 31
Version tested: 0.15.18
E
Nothing here yet. F
Magic-roundabout
Blueprint string: Right hand drive
Throughput: Set 1: 73 Set 2: 68
Deadlock notes: Multiple circles that can be saturated and deadlock. The circuitry makes it mostly safe though.
Scaling notes: All the buffers need to be adjusted to fit the longest trains, posted is 6 car
Version tested: 0.15.37
Original by: ERJHolton
Priority roundabout
Blueprint string: Right hand drive
Throughput: Set 1: 49 Set 2: 59
Deadlock notes: While I cannot guarantee that the circle cannot be saturated the circuitry makes this design very stable.
Scaling notes: The more trains that can fit on the circle the better it will perform. If only a few trains can fit on the circle the risk of deadlocks rapidly increases.
Version tested: 0.15.37
Original by: ERJHolton
No left
Blueprint string: Right hand drive
Throughput: Set 1: 49 Set 2: 55
Deadlock notes: Can deadlock if the intersection gets saturated, needs a lot of trains to do that though.
Scaling notes: all buffers should be scaled for longest train, posted is 6-car
Version tested: 0.15.37
Original by: Jap2.0
Double wide bubble cross
Blueprint string: Left hand drive
Throughput: Set 1: 31 Set 2: 38
Deadlock notes: While it can deadlock with regular traffic this one is really stable, should be comparable to A or B rated
Scaling notes: All buffers need to be scaled for largest train, inner circle need to be larger than longest train to avoid crashes
Version tested: 0.15.23
Original by: FrodoOf9Fingers (this variant via PM)
Re: 4-way intersection testing: Throughput and deadlocks
Posted: Mon May 08, 2017 3:44 pm
by Zonk
Wew nice work!
Re: 4-way intersection testing: Throughput and deadlocks
Posted: Mon May 08, 2017 4:31 pm
by AndrewIRL
Wow, great work.
In terms of throughput / (rails used x space) then it looks like the compact is the best. Still, my favourite is the r-twister because it looks pretty!
Re: 4-way intersection testing: Throughput and deadlocks
Posted: Mon May 08, 2017 4:36 pm
by Shokubai
These are awesome. Great work.
Am curious...the center sections of your cross seem to be missing several chain signals to break up the blocks. At a glance it looks like your N-S/E-W traffic would block trains moving the opposite way through.
Also damn your left hand drive
Re: 4-way intersection testing: Throughput and deadlocks
Posted: Mon May 08, 2017 5:13 pm
by aaargha
Shokubai wrote:These are awesome. Great work.
Am curious...the center sections of your cross seem to be missing several chain signals to break up the blocks. At a glance it looks like your N-S/E-W traffic would block trains moving the opposite way through.
I think it's about as good as it can get, some of the blocks look a bit large but in those blocks the rails cross so they'll block each other anyway. N-S/E-W traffic runs fine for opposing trains as far as I can tell. However, if you spot a missing signal somewhere I'll take a look.
Shokubai wrote:Also damn your left hand drive
Hehe, it kinda ended up like that as the deadlock tester I built was for LHD, now that my systems can handle both I really should get to including more RHD variants as well. Currently I only have the squareabout (which can't be converted as is to LHD).
Re: 4-way intersection testing: Throughput and deadlocks
Posted: Mon May 08, 2017 5:38 pm
by iceman_1212
Very nice
p.s. i laughed when i saw the curly intersection
Re: 4-way intersection testing: Throughput and deadlocks
Posted: Mon May 08, 2017 7:40 pm
by impetus maximus
about throughput...
ask yourself, how many trains can be going in different directions at once in the four way intersection?
if it's less than three (max), you should rethink your intersection.
Re: 4-way intersection testing: Throughput and deadlocks
Posted: Mon May 08, 2017 10:29 pm
by aaargha
Thanks for the kind words!
I've updated the post with RHD variants for most intersections. I'll have to take a look at a few of the ones left to see if it's possible to fit the signals in some reasonable manner.
impetus maximus wrote:about throughput...
ask yourself, how many trains can be going in different directions at once in the four way intersection?
if it's less than three (max), you should rethink your intersection.
I feel like I'm misunderstanding you but wouldn't that depend on which directions the trains are comping from and which exit each train is taking?
For example (in an LHD system) most intersections allow four trains to turn left at the same time, one from each direction. Going straight ahead or turning right is at best two as the paths intersect. And for most designs it's possible to do combinations more or less well. For the Cross design it's possible (if it all lines up) to have four trains turning left while two trains can go right/straight allowing for a six train "burst", so to speak.
Gah, time to stop over analysing and go to sleep.
Re: 4-way intersection testing: Throughput and deadlocks
Posted: Mon May 08, 2017 10:58 pm
by impetus maximus
i don't know what a LHD is. load haul dump?
i'm saying for a 4 way intersection of a two way rail system (not double two way etc) AT MOST in a best case scenario, three trains should be able to pass each other without stopping. of course depending on the direction trains having to cross other lines this can, and will be less.
Re: 4-way intersection testing: Throughput and deadlocks
Posted: Mon May 08, 2017 11:39 pm
by iceman_1212
impetus maximus wrote:i don't know what a LHD is. load haul dump?
left hand drive
Re: 4-way intersection testing: Throughput and deadlocks
Posted: Tue May 09, 2017 12:40 am
by impetus maximus
of course. lol thanks iceman_1212. i should know that, i use left hand drive (live in the US)
Re: 4-way intersection testing: Throughput and deadlocks
Posted: Tue May 09, 2017 5:26 am
by Aeternus
I'm surprised that the basic direct cross isn't on here:
Compact, can't deadlock since the entire intersection is one block, but very low throughput. Should be used on low-traffic sections that you don't want to risk jams at (rail sections far away from your main factory that just service some mining outposts).
Re: 4-way intersection testing: Throughput and deadlocks
Posted: Tue May 09, 2017 1:00 pm
by Shokubai
Aeternus wrote:I'm surprised that the basic direct cross isn't on here:
****
Compact, can't deadlock since the entire intersection is one block, but very low throughput. Should be used on low-traffic sections that you don't want to risk jams at (rail sections far away from your main factory that just service some mining outposts).
Bad signaling...Too large of a block. Max 1 train anywhere in the intersection?
Also you need standard signals instead of chains on the exits.
Re: 4-way intersection testing: Throughput and deadlocks
Posted: Tue May 09, 2017 2:58 pm
by Aeternus
Yea, that was an old screeny. Entrance chain, exit standard signals. The whole thing is a single block by design. As I said, very basic, low thoughput, but very safe. Use in low traffic areas only, where only deadlock resilience is a must.
Re: 4-way intersection testing: Throughput and deadlocks
Posted: Tue May 09, 2017 5:20 pm
by impetus maximus
here is the left side drive, 4-way intersection i came up with. i moved the left turns to the outer part so up to three trains can pass each other.
yes, she's a big one.
its.my.4way.png (1.01 MiB) Viewed 526572 times
i ran 2 trains from each direction, going to each direction. 12 total for 24+ hours. 0 deadlocks.
i use double headed trains. these singles are just to Illustrate direction.
Re: 4-way intersection testing: Throughput and deadlocks
Posted: Tue May 09, 2017 6:16 pm
by aaargha
Another small update: Added example pictures of deadlocks for any intersection not rated A. Tested the Direct cross intersection. Finished making RHD blueprints for all intersections that do not require major restructuring to fit signals where I need to.
impetus maximus wrote:here is the left side drive, 4-way intersection i came up with. i moved the left turns to the outer part so up to three trains can pass each other.
yes, she's a big one.
its.my.4way.png
i ran 2 trains from each direction, going to each direction. 12 total for 24+ hours. 0 deadlocks.
i use double headed trains. these singles are just to Illustrate direction.
That design is tested, look at "Wide crossover". The flaw with this design, in my eyes, is that the right turn cross more paths than needed, two trains cannot turn left right at the same time. For an improved variant take a look at "Wide". If you want to see how your intersection can deadlock there is and example picture linked in the OP.
Also, if you're the creator of this design, and want to be sourced for it, send me a link to the original post and I'll add it as source.
Re: 4-way intersection testing: Throughput and deadlocks
Posted: Tue May 09, 2017 7:14 pm
by impetus maximus
aaargha wrote:
That design is tested, look at "Wide crossover". The flaw with this design, in my eyes, is that the right turn cross more paths than needed, two trains cannot turn left right at the same time. For an improved variant take a look at "Wide". If you want to see how your intersection can deadlock there is and example picture linked in the OP.
Also, if you're the creator of this design, and want to be sourced for it, send me a link to the original post and I'll add it as source.
i'll see if i can improve it. i posted that design on April 3rd.
Re: 4-way intersection testing: Throughput and deadlocks
Posted: Wed May 10, 2017 6:09 pm
by aaargha
Added a few more LHD/RHD blueprints. All intersections, except the roundabouts and 2x 3-way, now have blueprints for both LHD and RHD.
impetus maximus wrote:i'll see if i can improve it. i posted that design on April 3rd.
Thanks, I don't really remember where I've found all of them:)
Re: 4-way intersection testing: Throughput and deadlocks
Posted: Wed May 10, 2017 11:39 pm
by zytukin
Posted this on your reditt topic but will post here as well.
I don't use 4 ways anywhere in my factory so don't actually use this.
This came from modifying the standard T intersection I use everywhere in my factory which I carried over to my 0.15 factory from my 0.14 factory.
When the original T intersection was used in my 0.14 factory, I never had a dead lock despite having 200+ trains running at once, including when I sent 40 trains to get copper and iron at once when I started a Green Circuit production area.
As it is pictured it is signaled for my current factory's 2-8 trains.
For the sake of reference, this is the T intersection that I use. Pic is from my 0.14 fac and it's signaled for the 1-4 trains it used.
Re: 4-way intersection testing: Throughput and deadlocks
Posted: Thu May 11, 2017 12:03 am
by Distelzombie
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.