AI Problem - Drive To Trackmark

boleyd

Well-known member
wrong-track.jpg



The engine is backing-up. The instruction is to drive to a trackmark at the far end of the coal car consist followed by a couple. Instead AI stops the engine about 50 yards from the switch and sets it to take the engine to the wrong end of the train. It then proceeds on that wrong path at a 1mph pace. Eventually it slows to zero but pressing SHIFT reveals a slow creep with intermittent stops. When the engine gets near the "wrong coal car" it stops and no further movement is possible. There are several instances in the all AI session where similar odd behavior occurs. Track is Jarrah Procedural. No signals. With signals other odd actions occur - usually 7mph, 2mph, ans 1mph slowdowns before coupling at 5mph.
 
The coupling issue is a carryover from T:ANE, which occurred in SP1 and caused the AI to creep up to signals then crawl through them when switching. Then there was SP2... This is where some substantial AI code changes were made, and this caused the AI to ignore track marks, signals, and direction markers!
Tony said this is a problem that will be addressed at a later date as this will require a complete code overhaul to fix the problem, and this was long before we knew of TRS2019. That part was reverted I think in T:ANE SP2 HF1 or HF2 (I can't remember now).

This is why I do my switching manually. The switching is not only faster, but also saner in the long run. Who has all day to wait for an AI driver to creep up to a boxcar at 1 mph, and then have the AI attempt to couple on the wrong end of the consist in the process!

Since T:ANE is the base upon which TRS2019 is built, this means this bug got carried over. Hopefully at some point sooner than later, this will get fixed.
 
FINALLY - someone has the same issue. I was getting ready to reinstall TS19. I even have had a train sit and stutter trying to get into a siding already occupied by a set of coal cars.

Sometimes, rather than changing code where multiple functions have a dependency, it is better to detect the need for using that code (conditional switch) and then branch to new code which is exclusive to the problem. The drawback is when the new code is still dependent on parts of the old stuff that the strategy returns to the the original dilemma.
 
How is the engine supposed to get to "a trackmark at the far end of the coal car consist"? Is there another turnout in the distance we cannot see?

If your scenario/objective is to couple to car3...

================/=====car1=car2=car3====TM========\
===loco==========/==============================\====TM New=========


...place a trackmark at TM NEW

Commands:

Drive to TM New
Couple to car3 (if you know it's name)

-or, if you don't know its name or if it changes-
Drive to TM New
Couple at TM (This assumes the radius of TM covers at least to car3)
 
I used Set Junction to set the switch before the instruction to couple. Junction set properly. I placed the engine about 200 yards from the switch and issued the instructions. The saga began with engine movement at 7mph, then 2 mph, then 1 mph and finally zero/nil. But that did not stop the intrepid driver. He kept the engine moving at a slow walking pace. It may have coupled but 5 minutes was the end of my patience. So, I placed a trackmark ahead of the "to-be-coupled-to cars" as the initial objective. Same as previous.

Setup trackage on a blank board. Slow motion action. Changing the destinations, switch arrangements, using trackmarks etc. did not help.

This is a similar problem I had years ago in Railworks. There a bug was confessed to and a promise made to fix it. Fortunately, I escaped from that drama and gave Trainz another try. I miss Sir Paul (CEO).
 
Did not work for me. Engine creeps at 2mph toward the TM New trackmark. Slows to 1mph and then stops at walking speed - Zero. Repositioning the trackmark and the engine did not help. To reinstall remains an option. Then, if that fails, Ihave to return to Railworks (or whatever it is now - Uck).

Going to setup similar scenarios in other routes. Will focus on $$$$$ routes since that might get the most attention from management when a problem is reported. I have been reluctant to report via the formal process due to the complexity of interrogation when trying to explain the issue. Wasts time since there is a chance, as mentioned, that the problem is known but too dangerous to defuse without risking an explosion.
 
I tried the Hintonville route (Original payware not the clone) and similar unsuitable behavior. Placed 3 box cars on a siding in the Hinton yard. Placed an engine two tracks away and tasked it to couple to the box cars. Started out ok with the engine at 1/2 of the 40mph posted speed. Why 1/2 only the programmers know. NO signals within the yard. As the engine approached the first switch it slowed to 2mph then 1 mph and then zero/nil mph while it crept at walking speed.

I removed, or set, every option to my idea of minimum impact on performance. Set one that enabled debug features to ignore those features. Same unacceptable performance. These settings were both at the program menu Settings and the Launcher level.

Such performance should cause a general outcry, which is not happening. Thus my installation must have a fault. After securing the database I will start a new one to see if there is an issue there.
 
Used AI to go to a location outside the Hinton yard and had to transit several switches/junctions. No issues. the engine maintained full 40mph speed during the process except where a speed reduction was obviously needed. NO COUPLING involved.-- clue....

Manually (no AI) connected 3 box cars to the engine and ran same scenario ok - no delays.

Using Kickstarter 2 system. AI instructed engine to couple to 3 box cars. Crept up at 2mph then finished the couple at 5mph. Instructed to go to a nearby location. As engine was leaving the yard it stopped where it entered the mainline. No further progress.
 
Last edited:
Did not work for me. Engine creeps at 2mph toward the TM New trackmark. Slows to 1mph and then stops at walking speed - Zero.

I just tested this in TANE and T19 Beta on a single baseboard. It works. The default for the turnouts was to the siding. Loco facing so that it has to back past the siding and couple to the front.


Driver commands:
CoupleAt TM
DriveToTM TMNew



The turnouts were automatically set/reset correctly. The loco backed down the track past the second junction, pulled into the siding and coupled. Then the train went to TM New.

If you are getting a creep-to-stop:
Check (and maybe replace) the tracks to at least a mile past the second junction.
Check (and maybe replace) signals to at least a mile past the second junction.

If still in doubt, set up a 1 baseboard test to see if that works.
 
Hi,

I'm using erendir Driver commands (drive to, couple with train,....) and all works well. Just finished 4 hours switching session.
The biggest problem for my is erratic AI reactions. Just hitting Drive via marker slow down instantly for 5 or more mph so all low speed switching is affected.
It is like model RR DCC in the early days.

Kind Regards
Darko
www.vlaki.com
 
Discovery -- If a switch/junction is between the engine and the target "couple-to" car the engine is stopped at the switch where it leaves the main line. If the target car is moved "ahead" of the intervening switch the couple proceeds normally and at expected speeds.

Still need to try Couple at Trackmark.
===============================
Couple at Trackmark operates as it should IF the engine can couple directly to the target car with no intervening switches. However, it also fails if there is a switch between it and the target trackmark.

This appears to be a "BUG". Now I need to figure out the goofy trouble reporting scheme N3V has in-place to discourage its use.
 
Last edited:
As I try different controlled simple scenarios WITH signals, the results vary but eh intervening switch remains an issue. I am not yet sure that it is always involved in all of the faulty scenarios but so far it seems it is. I also have to Try different "brands" of signals. I should note that I develop, or expose, the issues on a track only set of baseboards. Then I construct the same scenario on a real route which is large, but not hundreds of miles of track.This arbitrary setting of red signals, slowdowns to outright stops at red signals when no other trains are active makes the environment much less real and negates the considerable improvements in the visual attributes of the product.

Of course, my copy of the program could be faulty. However, I am not ready for a high risk reinstall.

-or-

Given the antiquity of the AI code, a more likely explanation is the combinations of various values, sets the stage for success or failure. With large AI systems running well the possibility of particular assets, instructions and rolling stock may be needed to inject failure modes.​

Dick[FONT=&quot] near Pittsburgh, Pa. i5-2500K 4.3ghz, 8gb memory, GTX1050 4gb video card[/FONT]:)
 
Discovery -- If a switch/junction is between the engine and the target "couple-to" car the engine is stopped at the switch where it leaves the main line. If the target car is moved "ahead" of the intervening switch the couple proceeds normally and at expected speeds.

Still need to try Couple at Trackmark.
===============================
Couple at Trackmark operates as it should IF the engine can couple directly to the target car with no intervening switches. However, it also fails if there is a switch between it and the target trackmark.

This appears to be a "BUG". Now I need to figure out the goofy trouble reporting scheme N3V has in-place to discourage its use.
boleyd,
Check the diagram in post #4. Check the instructions I used in post #10. There are two intervening junctions between where the loco starts and the couple location. This works perfectly. Tested in TANE & T19 Beta.

The only stop/slowdown is when the loco is approaching to do the coupling...
Engine stops.
Engine moves to the couple at 5 mph.

I'm guessing this is simulating stopping to let a trainman off the loco to guide the engineer during the approach.



Try this on a new blank baseboard. You'll see it works. That tells you that there are other problems along your trackwork.



ETA: I didn't see your post #14 before making this post.

ETA: I just extended the track beyond the siding. I added signals 02 at the diverging points; 04s at the converging points; a few 04's on the track past the siding and a buffer at the end.

No problems.
 
Last edited:
Back
Top