Trains Keep Committing SPAD

[AdvancedApproach]

@pware it's the same setup on the east end of my route and that problem doesn't occur. Even without a track marker there, the AI trains don't throw a switch that's aligned for the track.

 

[pware]

A SPAD for an AI train, in my experience, will occur if it does not have time to come to a complete stop before a STOP signal. This can occur if:-

1. the signal was at PROCEED (or CAUTION) but suddenly drops to STOP (e.g. another train has taken possession of the section ahead) or
2. the PROCEED signal is directly in front of another signal showing STOP but the first signal has no CAUTION aspect and is too close to the STOP signal.

 

[AdvancedApproach]

A SPAD for an AI train, in my experience, will occur if it does not have time to come to a complete stop before a STOP signal. This can occur if:-

1. the signal was at PROCEED (or CAUTION) but suddenly drops to STOP (e.g. another train has taken possession of the section ahead) or
2. the PROCEED signal is directly in front of another signal showing STOP but the first signal has no CAUTION aspect and is too close to the STOP signal.

The only train that took possession of the succeeding section was the train in question. The path was already clear for usage and the next signal was four baseboards away which is at least 1.5 miles down the line so the distance (or lack thereof) was not an issue. Watch the video again, there's no other oncoming train. In fact, one of the pictures showed the oncoming train had come to a complete stop but cleared that crossover while awaiting the helper units which were roughly 40 feet away from that westbound train when I took the screenshot. There was not yellow padlock symbol, indicating another train took control of that junction. The AI train was the only one present at the time. I know because I intentionally waited for the westbound train and helper units to clear the crossover before taking any further action.

 

[AdvancedApproach]

@JCitron @pware another thing is track markers don't solve everything. On a different thread, I complained about AI Drivers getting stuck at a proceed signal. The support team told me my Track Graph Line (TGL) was in excess of 70 km and AI can only search out to 25 km so if the next destination in their instructions is more than 25 km out, then they'll get stuck and wait for a train to completely clear it even though that track marker is more than 100 miles away. I tested this out in a bug report with the team, and they added a crossover which solved that problem temporarily but that's about it. Once the lead train cleared that track marker it proceeded on as instructed.

Here's some correspondence:

TGLs start and end at junctions, specific track industries and crossings (general rule of thumb - there are some exceptions). As your route has very few junctions and industries, the TGLs are very long. Directional markers apply to the whole TGL as well, so something to be aware of. Another factor to be mindful of is that directional markers are only checked when entering a TGL (not something applying to these cases).

Signals do not start or end TGLs. From how the train reverses to the junction, I suspect it is attempting to get out of that TGL and then head to the Track mark.

The devs have a task to look at it, but aware, it's unlikely to make the next release build. Adding junctions to the route appears to resolve the issue - not saying there isn't an issue - just offering a work around in the meantime.


The cause was the very long TGL - 70km.

The AI command search that should have resulted in the drive forward (searching the destination within the same TGL as the train) was limited to 25km. It was why it behaved as normal adding the junction ahead as explained in earlier emails.

In a future build, the limit will be increased to resolve this issue. It will not be part of the upcoming TLR phase 3 release, but the update after. It only impacts routes such as yours with extremely long TGLs.

I hope this helps.

That's why the track markers are helpful but not in this or every situation for that matter.

 

[pware]

An experiment to try.

1. On the SPAD track after the switch that flips at the turnout just after where the SPAD occurs, place a new trackmark.
2. Insert a new driver command into the AI loco command list for Carlos Beltran Drive Via <name of new trackmark> or Navigate Via <name of new trackmark> in front of the current Drive to *indecipherable name* Diesel Refueling Depot

See if that makes any difference.

On the matter of the long TGLs, that has been a known issue for many years now. Fortunately my longest stretch of single line track is well below the 70km limit.

 

[JCitron]

If it was caused by poor track laying, then this would be a problem at both ends of the route. The approaching signal in my video was approach, which means proceed at a speed not exceeding 30 mph and be prepared to stop short of the next signal.

Not necessarily. Humans do make mistakes you know. I recommend driving the line yourself and find out what's wrong.

 

[AdvancedApproach]

Not necessarily. Humans do make mistakes you know. I recommend driving the line yourself and find out what's wrong.

There's nothing wrong with the line. I just might have to do a little extra work for the AI trains at that particular location.

 

[pware]

There's nothing wrong with the line.

Brave words. I have said that myself many times only to be proven wrong.

Another thought. One reason why the AI will flip a switch into an unexpected direction is because it has determined that either the new direction is a shorter path or that there is a blockage (e.g. another train or a fault with the track or a switch somewhere ahead).

 

[KotangaGirl]

AI drivers flipping switches on facing points to shift to the right drove me nuts when I first started out with Trainz. For some unknown reason it's built into the AI programing as a default and the only way to stop it is to either use direction markers or trackmarks to force the 'Automatic Idiot' to stay where they are supposed to be.

 

[JCitron]

AI drivers flipping switches on facing points to shift to the right drove me nuts when I first started out with Trainz. For some unknown reason it's built into the AI programing as a default and the only way to stop it is to either use direction markers or trackmarks to force the 'Automatic Idiot' to stay where they are supposed to be.

Shifting between lines is the "AI logic" calculating the shortest distance to destinations. Yes, we've said before and even the old Auran guys say to use direction markers and track marks to keep the AI on track, no pun intended.