AI uncoupling causing crash ... sometimes

Discussion about RailWorks scenario creation.

AI uncoupling causing crash ... sometimes

Unread postby OldProf » Thu Dec 06, 2012 11:47 am

For this scenario writer, inconsistency is the greatest frustration.

I've been working on an elaborate switching scenario for Munich-Augsburg, which involves making up a train in the Pasing Yards, dropping it off in the Pasing West Yard, and then making up another train there for delivery back to Pasing. While the player works in one section of Pasing West, an AI switcher makes up a train in another section of Pasing West, finally dropping off that train and driving away. The result is that both the player and the AI driver wind up a couple of tracks away from each other.

The inconsistency is that sometimes, but only sometimes, when the AI driver drops off its train it proceeds to drive off in the wrong direction, thus crashing into the cars it just dropped off and causing, at best, a scenario-ending train wreck or, at worst, an SBHH error. In either case, the scenario can be resumed from the penultimate save (see below) and the AI driver gets its direction right the second time around. The scenario can then be completed by the player, although he gets the big X for his efforts rather than a nice green check mark.

I've included an instruction that warns the player of this eventuality, but I wish I knew why the magic works sometimes and others not. AI trains have been capable of dropping off cars through at least the last two incarnations of TS, so that alone shouldn't be the cause. At first, my instruction sequence included simply a Drop off instruction followed by a Drive to instruction. When this combination produced wrong-way driving after drop-off consistently, I added a Drive to instruction before the Drop off. That allowed the sequence to work as I intended ... sometimes.

I'd appreciate comments from other scenario writers who have tried something similar. RSC bashers should start their own threads rather than hijacking this one, please.

P.s.: Kudos to the creator of this route for leaving numerous tracks at Pasing West without any markers, thus allowing scenario writers to place their own. !*brav*!
Tom Pallen (Old Prof)

{Win 10 Home 64-bit; Intel Core i7 6700 @ 3.40GHz; 16.0GB Single-Channel @ 1063 MHz (15-15-15-364); 2047MB NVIDIA GeForce GTX 960}
User avatar
OldProf
 
Posts: 2755
Joined: Wed Sep 09, 2009 10:09 am

Re: AI uncoupling causing crash ... sometimes

Unread postby Marleyman » Sat Dec 08, 2012 4:38 am

I was going to say add a marker and give the train a 'way point' to your marker to ensure it heads away from the consist, seems you have done that. Without actually seeing the scenario instructions now I can't think of anything that would be amass. I have the route but I have not created any scenarios for it so cannot comment on limitations on that particular route.

I do know that if an AI train drops off a partial consist in siding A and then has an instruction to pick-up another consist at siding B and drop that in Siding A, it will crash because Siding A was empty at the start of the scenario. The same will be true if siding A is empty and then the AI train makes a consist in siding A and uncouples from it, does anything else, and then has instructions to couple to the consist in Siding A again.
My Railworks 2 Catalogue; http://www.railworks.marleyman.co.uk/
and My RWTS2013 Download Store http://www.railworks.marleyman.co.uk/store/
User avatar
Marleyman
 
Posts: 357
Joined: Fri Dec 10, 2010 4:46 pm
Location: UK

Re: AI uncoupling causing crash ... sometimes

Unread postby OldProf » Sat Dec 08, 2012 10:22 am

Marleyman wrote:... I do know that if an AI train drops off a partial consist in siding A and then has an instruction to pick-up another consist at siding B and drop that in Siding A, it will crash because Siding A was empty at the start of the scenario. The same will be true if siding A is empty and then the AI train makes a consist in siding A and uncouples from it, does anything else, and then has instructions to couple to the consist in Siding A again.


In this case, an AI engine picks picks up cars from sidings A, B, and C, and then drops the entire consist on siding D. As I wrote previously, the pickups happen as ordered and the engine pushes the consist onto siding D successfully. The final drop-off also works as it should. The final step in the sequence has the engine heading off toward a scenario-placed track marker and this is where the problem sometimes occurs. The AI "driver" sometimes gets it right, but other times forgets to move his reverser and crashes at full throttle into the consist he just dropped off.

To me, the puzzle here is the inconsistency.
Tom Pallen (Old Prof)

{Win 10 Home 64-bit; Intel Core i7 6700 @ 3.40GHz; 16.0GB Single-Channel @ 1063 MHz (15-15-15-364); 2047MB NVIDIA GeForce GTX 960}
User avatar
OldProf
 
Posts: 2755
Joined: Wed Sep 09, 2009 10:09 am

Re: AI uncoupling causing crash ... sometimes

Unread postby Chacal » Sat Dec 08, 2012 2:56 pm

Could it be that there are several possible paths from the drop-off point to the final AI destination?
Then sometimes the dispatcher would choose a path that goes through the consist (loose consists are invisible to the dispatcher).
Especially if it is reserving a more obvious path for the player train which is nearby,

I suggest you try a few different AI destinations.
Over the hill and gathering speed
Chacal
Site Admin
 
Posts: 6464
Joined: Tue Jul 05, 2011 1:11 pm
Location: Quebec, Canada

Re: AI uncoupling causing crash ... sometimes

Unread postby OldProf » Sun Dec 16, 2012 3:44 pm

Chacal wrote:Could it be that there are several possible paths from the drop-off point to the final AI destination?
Then sometimes the dispatcher would choose a path that goes through the consist (loose consists are invisible to the dispatcher).
Especially if it is reserving a more obvious path for the player train which is nearby,

I suggest you try a few different AI destinations.


Thanks for the suggestion. The AI engine does make its drop-off on a "through", double-ended yard ladder siding, however, the red-line path that appears in the Timetable Editor always shows that engine heading directly for its next destination, which lies in the opposite direction from the dropped wagons and is also its End Destination. I have, of course, considered just eliminating the drop-off and leaving the AI switcher sit on the track connected to the wagons it has successfully picked up and delivered, but I really prefer the drop-off and drive-away. I'll admit that my pride as a scenario writer is at risk here, as well.

For what it may be worth, the AI switcher's starting and ending point is a marker that I placed at the scenario editing level.

Apparently, I'm not the only scenario writer having problems lately. There's a lengthy thread over at the Otto/Woolf forum regarding peculiar behavior by the Timetable Editor lately. Is it possible that RSC, which is reportedly going to eliminate the Timetabled Scenario marker, is also working on a replacement for the Timetable Editor and, in the meantime, is making potentially dangerous changes to the current editor? Good grief! I really apologize for indulging in the very kind of unfounded speculation that I usually deplore.
Tom Pallen (Old Prof)

{Win 10 Home 64-bit; Intel Core i7 6700 @ 3.40GHz; 16.0GB Single-Channel @ 1063 MHz (15-15-15-364); 2047MB NVIDIA GeForce GTX 960}
User avatar
OldProf
 
Posts: 2755
Joined: Wed Sep 09, 2009 10:09 am

Re: AI uncoupling causing crash ... sometimes

Unread postby OldProf » Fri May 17, 2013 11:09 am

Recent additional information: Sometimes, but again, only sometimes, the problem I mentioned does not occur when at least one siding is empty to each side of the siding where the uncoupling and pull-away movement take place. I learned this when I added some static consists to nearby sidings to make the yard area more populated. Does that make sense?
Tom Pallen (Old Prof)

{Win 10 Home 64-bit; Intel Core i7 6700 @ 3.40GHz; 16.0GB Single-Channel @ 1063 MHz (15-15-15-364); 2047MB NVIDIA GeForce GTX 960}
User avatar
OldProf
 
Posts: 2755
Joined: Wed Sep 09, 2009 10:09 am


Return to Scenario Creation

Who is online

Users browsing this forum: No registered users and 1 guest

cron