AI drivers see the next signal way too far ahead (in fact, immediately!)

[Smileyman]

As the title says, the AI drivers look too far ahead for the next signal state.
As soon as they go through one signal, they are aware of the state of the next signal, even if it is a mile or more down the track.
And I know it's always been this way, but it's such a pain.

It results in a train passing through a green signal, and if the next signal is amber, it slows down to half the track speed while still in the 'safe block', and if the signals haven't changed, then the same speed for the next block too.

That's the majority of 2 blocks that a train will run at half speed because it started acting on the amber loooong before it should be able to see it.
Is there a way around this?

I looked for a rule that might help, but couldn't find anything, and I discounted driver commands because I'm using EIT and MCM, and I exclusively use 'Autodrive to/via MCPathStop' for the drivers.
I've a feeling it's hard-coded into the game by it's wise developers.

It would be nice if there was a distance we could set for the AI before it 'saw' the next signal, to make it much more realistic.

I doubt there's anything that can be done, but I just wanted to put it out there.

 

[jake07999]

As the title says, the AI drivers look too far ahead for the next signal state.
As soon as they go through one signal, they are aware of the state of the next signal, even if it is a mile or more down the track.
And I know it's always been this way, but it's such a pain.

It results in a train passing through a green signal, and if the next signal is amber, it slows down to half the track speed while still in the 'safe block', and if the signals haven't changed, then the same speed for the next block too.

That's the majority of 2 blocks that a train will run at half speed because it started acting on the amber loooong before it should be able to see it.
Is there a way around this?

I looked for a rule that might help, but couldn't find anything, and I discounted driver commands because I'm using EIT and MCM, and I exclusively use 'Autodrive to/via MCPathStop' for the drivers.
I've a feeling it's hard-coded into the game by it's wise developers.

It would be nice if there was a distance we could set for the AI before it 'saw' the next signal, to make it much more realistic.

I doubt there's anything that can be done, but I just wanted to put it out there.

Have u tried testing with invisible signals? I try to avoid using them but I have had a similar issue to you before in the past, they could be a quick fix for now.

I wanna say track circuit blocks might help too but in my experience, unless u use the path triggers, the signals stay red right up until the loco is a few feet away from it (even though the track ahead would be clear and points are already set)

 

[H222]

I wanna say track circuit blocks might help too but in my experience, unless u use the path triggers, the signals stay red right up until the loco is a few feet away from it (even though the track ahead would be clear and points are already set)

This is definitely not a ‘normal’ signalling experience and I suspect you’re doing something wrong or using the wrong signals. Them only coming off when you crawl up to them is a scripted feature of Bloodnok’s signal script to represent a signal being pulled off and a driver being ‘waved ahead’ up to the next signal only. Try using a distant signal before each yard/station complex if you’re using semaphores.

As for the AI slowing down early - I am pretty sure they use DCC mode - and depending on the signalling sets being used - I have noticed that even whilst displaying a full clear signal state that some signals will show an advanced caution state when you hover over it (“line ahead is clear for two blocks”), meaning they will behave like it’s a caution because they’ve been told it’s a caution, albeit an advanced one. As dumb as the AI is, it can only go on what it’s told. The signalling system I’m using most at the moment (Ranger_51’s NSW colour light signalling system, based on UK practices) doesn’t have this problem - a Green on the starter has the AI fully sending it (up to 160kph), even if it only has Green - Yellow - Stop, as this green is a full clear and not showing in trainz an advanced caution.

Jamie

 

[KotangaGirl]

I solve this problem by sighting the Home signal at track level to find the point where it becomes visible from the driver's point of view and I then place an invisible signal at that spot. I agree with you, the super telescope eyed AI drivers are a total pain.

 

[Smileyman]

Thanks for the feedback guys.

@jake07999 : I haven't experimented with the circuit blocks yet, as everything so far is working fine without them. I do like the idea of invisible signals though.

@H222 : I'll take a look at different types of signals and do some tests.

@KotangaGirl : Wow, that's a very novel idea, and sounds like it would work, especially on long stretches, which is where the problem is the worst.
Forcing the drivers to have human eyeballs. I like it!

Shame that they'll show up in the profile display at the bottom of the screen.
I wonder if the signals are placed on their own layer, with the visibility turned off, whether they'd be invisible in the profile? Hmmm.
Probably not, and if they were, they may not function because the layer is turned off.
Sounds unlikely, but I'll definitely give it a test tomorrow.
And I could always turn all signals off in the profile, as I always forget it's there anyway.

Cheers guys.
Food for thought.

 

[schnackel]

I've a feeling it's hard-coded into the game by it's wise developers.

Similar post look here: https://forums.auran.com/threads/train-only-runs-at-half-speed.171014/

edit the config.txt of the engine-specs you use in a session and add the line
autopilot-caution-speed-multiplier 1
see description here:

KIND Engine - TrainzOnline

online.ts2009.com

 

[Smileyman]

@schnackel : That's a good idea, but as the wiki says, there is a danger of overrunning then, and I'm trying to make my automated sessions as fool-proof as possible.

@KotangaGirl: Tried what you suggested, and it works a treat!
I'll probably only use it on long stretches, but it really makes a difference.
I must have saved around a minute on a 10 mile test.

I did notice that the AI seemed to see more than one signal past the approaching signal on occasion.
Don't know if that's hard-coded or whether it's EIT-MCM.

Either way, I'm really pleased with how this 'hack' has worked out.

Cheers!