Breakdown tolerant tracks

[odalman]

On the positive side, the traffic actually flows in the case that I tested and it only takes up a width of 2 tiles. However, something preventing a train from cutting in front of another train is still the most badly needed functionality. Also, the pathfinder might not recognize the difference between a train at full speed versus a broken down train ahead in a lane. Or did I just not notice it? If a train has a choice of 2 lanes with a train ahead in each, it should choose to come behind the one with highest reliability, all else equal. I tested for a while with 4 backwards signals after each crossover but it did not noticeably discourage lane changes.

[Oncle Boris]

On the positive side, the traffic actually flows in the case that I tested and it only takes up a width of 2 tiles. However, something preventing a train from cutting in front of another train is still the most badly needed functionality. Also, the pathfinder might not recognize the difference between a train at full speed versus a broken down train ahead in a lane. Or did I just not notice it? If a train has a choice of 2 lanes with a train ahead in each, it should choose to come behind the one with highest reliability, all else equal. I tested for a while with 4 backwards signals after each crossover but it did not noticeably discourage lane changes.

Since there are back facing signals every nth-tile on both tracks I'd suspect yapf applies a similar penalty to both lanes, thus making both options equal and triggering unpredictable train behavior.

[odalman]

Yes, the signal length was a problem. Here the amount of forward signals is doubled:

It seems to work better.

Eventhough the pattern of track and signals is symmetrical, the idea is that for any finite distance that the pathfinder looks ahead, there will be at least as many backwards signals in the other lane than in the current lane.

[Oncle Boris]

So with this design you've posted, reducing signal length, trains aren't reserving a path anymore through the switcher, allowing trains on the other line to block them should the pathfinder fancy a lane change.

[Afdal]

Here's how I would do it, if I were mad or playing with the Clipboard patch that lets you copy and paste designs:

High-speed shifters for 3-tile-long trains, at their highest density.

[verzasoft]

Since there are back facing signals every nth-tile on both tracks I'd suspect yapf applies a similar penalty to both lanes, thus making both options equal and triggering unpredictable train behavior.

This is exactly what happens to me. With reference to the picture, brown train cuts the road of the green train. It seems that the pathfinder evaluates as even the two lanes, due to the even number of backwards signals along the full railway.
So it seems impossible to have a "compact" setup 2-tile wide: a middle tile is needed so that only curving path encounters the backwards signal and straight one avoids it.

untitled.PNG

(316.59 KiB) Not downloaded yet

[verzasoft]

Ok, I tried also the "intermediate" switch tile with backwards signals.
It seems that this is not so much penalized by the pathfinder, since trains take both paths with same behaviour.
So it seems that is impossible to achieve a breakdown tolerant double track.