AI failing to couple to consist

[ttjph]

Yet another minor scenario problem...

I've got an AI (or possibly AS - artificial stupidity) loco arriving in a yard with a rake of wagons, running round in an adjacent siding, setting back onto another rake of wagons - then failing to couple to them, and b***ering off light loco.

I've checked that I'm using the correct 'couple to' instruction (front, in this case, as it's facing forwards when it first arrives); I've tried moving the wagons closer to the end of the siding where the loco should couple to them; I've checked that it's an outer end of a wagon (Oovee MJAs).

What could I be missing?

[Rockdoc2174]

Something that's caught me out is the loco and the wagons must fit completely inside the marker. If they don't the engine will either come to a stop without coupling or crash into them.

Keith

[AndiS]

It's an ages old limitation of this game that AI has no clue where anything is once that moved from its initial position. So you can only have AI couple to wagons that never moved before.

[ttjph]

Thanks Keith and Andi.

I'm pretty sure neither of those is the problem: the wagons are there from the start, and they fit well within the siding marker - plenty of space for the loco too.

Is it worth trying a 'couple to rear' instead, just to see? I'm sure I've played scenarios where I've got a tick for coupling to the wrong end of the loco...

[malkymackay]

I would try 'couple to rear', just to see what happens. You've nothing to lose by giving it a go.

[Rockdoc2174]

I don't know if it's still a restriction in TS015 but, a couple of years ago, I was trying to develop a very long (about 3 hours) early morning freight scenario for jetgrif's S&Cv7, beginning at Hellifield. The player had to wait for a couple of passenger services before getting the road so I wanted to put something in to watch in the meantime. I had a 4F shunting in Hellyard, assembling a train and taking it towards Skipton. As this was at the start of the scenario it was done in the early stages, too. As I went along I kept adding passing traffic - I was working to a pre-War timetable - and all was well until, suddenly, the shunting part failed. About the same time, AI trains began to appear with red lights leading so I can only assume that there was some kind of boundary I'd crossed with the AI controller in the game.

Try just having the shunting part in a new scenario. It seemed to me that once this AI gets broken removing the last AI consist doesn't clear it up again so cloning your existing work and deleting everything else won't do.

Keith

[deltic009]

From your description it sounds like the AI starts off coupled and moves the freight somewhere and then uncouples to run round. When you give the AI a couple to instruction it will try to run to the exact point that it started with those freight wagons as that's all it can do - it is a silly limitation that has never been worked on sadly, probably one of the biggest glaring errors.

[AndiS]

From your description it sounds like the AI starts off coupled and moves the freight somewhere and then uncouples to run round.

That's the way I understood it, too. But "loco arriving in a yard with a rake of wagons" seems to mean "light loco arriving in a yard where a rake of wagons wait".

I can confirm what Keith said about the marker length. The dispatcher's thinking is this: 1) get the (whole) train inside that marker; 2) uncouple. So if the whole train does not fit in, he cannot succeed, even if it sound ridiculous and often you will have a short marker at some shed where you want the wagon to go to. I haven't experimented too much with that, though. One thing I know: he always goes to the far end of the marker. You might exploit that by making the marker much longer on the end where the engine and the wagons staying with the engine are.

[ttjph]

Sorry, just to be clear there are two sets of wagons. The loco starts coupled to a loaded set, somewhere outside the yard. It brings them into Siding 7 and uncouples. There is an empty set of wagons already sitting in Siding 6. The loco runs back through Siding 5 before (in theory) coupling to the empty wagons in 6 (this is Hedborough North, in the main yard towards the loco depot).

On reflection I'm thinking that if I 'couple to rear' it will probably enter 6 from the wrong end - possibly running through 5 ('go via') and then reversing to do so!

What about coupling specifically to just the end wagon, rather than the whole rake - is that likely to work?

[AndiS]

All right, more actors in the plot than it seemed.

Regarding front and rear, I believe that front is the end where the driver icon is shown in scenario editor. I.e., the initial driver's location. I remember people getting confused about the terminology in scenarios where you start by reversing out of some depot.

Anyway, I would create a copy of the scenario where you are already in that yard and try all permutations of front and rear and wagon numbers/counts. That way, you get hard facts within 15 minutes or so. Maybe it works in isolation but not as part of a bigger scenario. Then it is good for you to find out with minimal time waste. I have no experience with long scenarios. Some people report disturbing things. But getting down to the bottom of such complex settings is very time consuming.


There is another thing that could go wrong: If the number of vehicle numbers used for the wagons in question is insufficient, then some will be used twice which is know to confuse the dispatcher.

[ttjph]

OK, coupling to just one wagon of a consist (even the end one) causes an AI collision and game over.

However, I think I've identified the source of the problem. I sat next to the end MJA and watched the loco (RSDL 66) pull up to it - it looks like the bounding boxes are slightly too short (or possibly long), so when it coupled, the loco and wagons jumped apart by a few inches. It appears this was enough for the coupling operation to fail, and the loco drove off anyway.

Oh well - I'll just leave it sitting in Siding 6 after attempting to couple. The player should be long gone by then anyway.

Thanks for all the suggestions!

[AndiS]

Now that is quite amazing, bad bounding boxes still floating around in late 2014?

This is pretty simple to fix: In the wagon's .bin file, you find:
FrontPivotX - half the length over buffers(*)
BackPivotX - same with a minus sign in front
CollisionCentreX - zero
CollisionLength - less than the length over buffers, by a metre if you want to be on the safe side.

If the wagon is not symmetric along the X axis, then CollisionCentreX will not be zero. In that case, that value needs to be added to FrontPivotX and subtracted from BackPivotX.

(*) To be a bit more precise, you ought to add half the gap between buffers. It is simply the location of the pivot where the next wagon's pivot will sit when coupled up. It is measured from the origin of the body shape.

[ttjph]

TBH I can't be bothered - it's an extra pain for anyone else downloading the scenario; it's only moving scenery; and I'd have to work out whether it's the loco or the wagons. But thanks for the pointers; I may well refer back to them one day!

I included a similar fix in my Class 70 patch after discovering that it was terrible to shunt with; given I'm using the relatively ancient RSDL 66, I'd put a small bet on that being the offending item rather than Oovee's MJA. But it would take time to prove it.

[ttjph]

Just to finish the discussion: telling the AI to finish at Siding 6 causes an AI collision, I presume because, having failed to couple, the loco then heads for the middle (or possibly the far end) of Siding 6 - which of course is occupied by the wagons.

Scenario is now up on UKTS if anyone has the stock and fancies trying it...

Winter Coal Import [48465 bytes] - WinterCoal_141230-2217.zip File ID: 33694 Date: 01 Jan 2015 - 187 Downloads