The latest AI balk in TS12: well cars!

On my "Double Gridiron" Model Trainz layout I initiate the session with a unit freight train on the main line: two orange V/Line S Classes by Auran pulling a string of 25 of Dave Snow's ethanol tank cars followed by two intermediate engines: Union Pacific SD70ACE Number 8485 by n8phu and BNSF GE ES44DC Number 7695 by the same author followed by 28 more Dave Snow ethanol tank cars of all the same markings: GATX DES.

The above train is given a schedule to run over the mainline from the library and does so continuously without ever missing a beat. All day long.

The damndest thing happens, however, when the freight cars on the train are changed.

The tank cars are put in my yard while different cars altogether are assembled to this train with the same engines.

The new cars are various intermodal well cars by boweavil and gary-evans: DTTX and TXX with various loads and markings: some cars are double stack, some single and some empty.

The new consist has 29 well cars following the head engines and 25 cars following the intermediate helper pair.

The new well car train is given the same schedule and starts at the same place on the line the original tank train started.

The weird thing is that NOW the well car train stops and backs up when it gets to a certain part of the main line. There are no trackside objects, track marks or junctions at this part. This occurs about 100 feet past a signal but the signal can't be the culprit since it is green when it happens and no other AI train balks at it. Usually if there is a problem on the track where one AI engine balks, every other engine and driver will balk there just the same.

No other AI train on this line balks at this part of the line: just the newly-formed well car train right after the tank cars are traded out for well cars.
Merely changing railcars and nothing else creates this newest AI balk. Supernatural phenomenon? The head engine AI driver even remains the same too.

What could possibly happen?

Is this particular well car content jinxed?

Have certain non-powered rolling stock content known to give AI trouble?

Also, note that the well cars have a longer wheelbase than the tank cars so the new well car train is longer overall and probably heavier than the old tank train.

Should I try changing the tank cars for something else besides well cars?

Should I try shortening the overall length of the well car train so it's no longer than the original balk-free tank train and see what happens?



Does train weight and overall length sometime affect AI performance negatively? Could the well car content just have corrupt data in it?

Follow up sometime later after doing some AI tests.


It appears that AI was balking at overall train length.

Reconfiguring the original tank train to a well car unit freight of the same length to start a session seems to have made the AI behave normal.

The shortened well car train seems to be constantly passing that spot without hesitation on my mainline where it was balking before when the train was longer and pulled from the yard.

I made the well car consist a couple hundred feet shorter than they were when they were in the yard and AI started to act goofy following a swap-out operation with the tank consist.

The only question remains is will the train continue to act normal when the well cars are swapped out for the tank cars which are now in the yard.

Time will tell. Was it a train length issue or was it the yard marshalling operation that made my AI train go bonkers?

Apparently my well cars aren't inherently bad content.

That would actually make sense. It's a possibility that your longest consist was sitting between two signals, causing confusion. Remember the scripting and what not that's going on. In real life railroads don't like to stop trains in between blocks either because it can screw up traffic flow. To get around this situation, consists are usually no longer than their passing sidings and signals can accommodate.

The spacing of signals is quite a complex issue because having too long a distance between signals can cause many short trains to stack up as they wait for passage into the next block, while too close a distance with too long a train can cause signals to either remain yellow causing too slow an operation since the drivers will now run under caution, or for signals to remain in stop position, red, causing too many trains to wait for passage.

JCitron said:

That would actually make sense. It's a possibility that your longest consist was sitting between two signals, causing confusion. Remember the scripting and what not that's going on. In real life railroads don't like to stop trains in between blocks either because it can screw up traffic flow. To get around this situation, consists are usually no longer than their passing sidings and signals can accommodate.

The spacing of signals is quite a complex issue because having too long a distance between signals can cause many short trains to stack up as they wait for passage into the next block, while too close a distance with too long a train can cause signals to either remain yellow causing too slow an operation since the drivers will now run under caution, or for signals to remain in stop position, red, causing too many trains to wait for passage.

Click to expand...

Well, that long consist that was balking would have been between 4 signals spaced at 1,000 foot intervals in the spot is was reversing the train. It backed up even when it was 100 feet past the lead signal in that particular block that was green. Signals at 1,000-foot intervals go for several miles along the layout. No long train has ever balked at any of those other close-interval parts of the track. When the well car train was at its longest it only seemed to balk past this one particular signal.

The 1,000-foot interval spacing for signals is designed to minimize yellow conditions and train-stacking. I'm therefore restricting AI's forward vision of the main line. In real-word railroading, signal status ahead is limited by as far as human eyes can see. How far ahead can the naked human eyes of a train crew read an upcoming signal from the lead engine cab? 1,000 feet? 2,000 feet? More? Less?

My shortened well-car train now runs past this formerly-troublesome block without ever missing a beat. The shorter train has a lighter gross train weight as well. I don't know if train weight can cause AI malfunctions. The train goes up an upgrade in this part. AI may stall and back up if the train is too heavy for the loco power supplied. The F7-type units at the head of the train aren't particularly powerful.
When my well-car train had about ten more loaded cars than it now has, it was accelerating rather sluggishly just before it stopped and backed up.

Now, I will try switching out my well cars for the tank car consist of equal length in the yard and see if the new train also runs balk-free.


Follow-up.

Ok, no problem when the well car train is switched out for the tank car train of the same length.

It appears to be a train length/weight/loco power problem.

Another thing. The back of the shorter train now clears the 50 MPH sign (leaving a 25 MPH zone) just BEFORE it reaches that spot where it balked before when it was a long train. This is on a slight upgrade and now the diesel locos can actually rev up before reaching the former spot where it would stop and back up. The train has to get up just enough speed (gumption) before reaching the questionable spot on the line so as not to bog down and stop.

The solution: restrict train length for this particular AI-driven train. This particular section of oddball track with its grade and speed limit sign placement is idiosyncratic for AI function. Something to do with speed and AI throttle control. An excessively long AI train may stall on upgrade and balk if approach speed is too low and loco power is weak.

If your train is 5280' long, try spacing the signals @ 5300' apart