Please N3V Fix this Game

davesnow

Crabby Old Geezer
Please fix this game whereby if a AI Consist is sitting at a Green light and the light says "Clear" that the train will GO ON instead of just sitting there like a knot on a log, keeping all the other trains from going where they need to go.

Thank you

Dave Snow
 
AI Consist is sitting at a Green light and the light says "Clear" that the train will GO ON instead of just sitting there like a knot on a log

More information needed. The train is not moving for a reason that may not be immediately obvious to us.

What does the AI report as the problem? Things such as "no path to destination", "missing junction", "track blocked", etc all of which can occur well beyond the signal block length so the signal can say "clear" but the AI is unable to progress to the next track mark or industry in its orders.
 
If you put the train in 'drive,' if it stops for more than ~30 seconds it will cancel the drive and will not move without user interference. This can clog up the line as other trains with 'drive' enabled stop at red signals waiting for the first train to go.
 
If you put the train in 'drive,' if it stops for more than ~30 seconds it will cancel the drive and will not move without user interference.

I just ran some tests in Trainz Plus and can report that even after 10 minutes of waiting, the "Drive to" command was not cancelled.

The first test was a simple case of two trains needing to access the same long section of single track from different starting tracks. Both trains under AI control. The second train moved forward to the signal that controlled the junction and then had to wait 8 minutes before the first train had cleared the section and it was able to move. During the wait the second driver reported "Waiting for track clearance" then "is stuck and awaiting new instructions".

For the second test I deleted the junction switch lever in front of the second train - the first train was not affected. The second train did not move at all but gave the message "Unable to plot route. Junction missing ahead". After 10 minutes the Drive To command was still active and in its order queue.

So, my original question remains. What is the problem ahead that is causing the AI to refuse to move?
 
No look--- There is not a single train for 20 miles down the line. It's not waiting for track clearance. There is "no path to destination", "missing junction", "track blocked" etc. The train is just sitting there. And another thing--- tell me why some AI trains follow signals all the time and other AI trains follow "trackmarks" all the time. And also tell me why I have two locomotives with the EXACT SAME specifications in their Config, but one train will go the speed limit all the time and the other train will only go 1/2 the speed limit all the time.

Seems to me there needs to be some changes in this game engine.

Running a session with a lot of AI trains can start out great but by the time you've played the session about an hour or two, it becomes a nightmare. Trains missing trackmarks (this happens a lot), going down the wrong track (even WITH a trackmark instructing the train to Bypass the line), getting stuck on a curved section of track (I've found if there is a signal near a curved section of track, and the locomotive has to stop there, sometimes they can't get up enough speed to get going again. The curve is too sharp.), and the old aggravation--- the train just stops and refuses to move--- no reason--- it just will not move. Sometimes I can save the session, exit out of trains and completely close the program--- then re-start Trainz, and go back into the session, the train will start moving.

I still love this game though. It's given this old guy a lot of pleasure along with the headaches.
 
All of that - is why I gave up making routes and running sessions back in the TRS 2004 days. It seems the AI hasn’t improved much in the years since then.
 
Trainz's AI needs MAJOR work. SP5 (I think) changed something about the AI, because I've noticed AI trains bypassing red signals on sidings and on the mainline. It's super annoying and potentially deadly to an op session. There's also the fact that after being stopped for ~30 seconds with the 'drive' command (not drive to), it cancels the command and requires intervention. This is INFURIATING when trying to operate a model railroad or have a railfan session on a route like Coal Country or Mojave Sub.
 
There's also the fact that after being stopped for ~30 seconds with the 'drive' command (not drive to), it cancels the command and requires intervention.

Sorry to keep contradicting you but I can only go by what I see happening.

I just ran a third test, this time using the "Drive" command, which, by the way, is a hangover from the days before the "AI" was introduced (pre TS2004???) and its purpose was to set a train in continuous motion provided the track was clear (a large loop was ideal) and it would obey all speed signs but it would not set any switches.

In driver mode I moved train 1 into a position where it blocked a junction needed by train 2. I then issued the "Drive" command to train2. It moved forward to the stop signal immediately in front of the blocked junction and waited with the message "Waiting for access to junction ...". I waited 5 minutes before moving train 1 past the junction and onto a branch line clearing the junction which I manually reset for train 2. The signal in front of train 2 changed to proceed, the train moved forward and onto the main line. The "Drive" command was not cancelled.

My observations.

I have always stated that the "I" in "AI" has never stood for "Intelligence" - "Idiot" or "Incompetence" could be more accurate definitions. This is also true in other areas of computing, industrial control, problem solving and so called "expert systems". "AI" systems work when everything else operates exactly as predicted - no delays, unexpected events, errors, breakdowns, etc when human intervention is always required. Why do people think that Auran/N3V can achieve with an "AI" system what railroad operators with $$$$$mega to throw at the problem cannot?

I make very limited use of AI in my sessions and rarely, if ever, have any problems. Like those childrens toys that can play by themselves, I fail to see the point of setting up a fully automated railroad that runs by itself for hours (or even days) without human intervention - but that is just me. A very large part of the joy of Trainz is getting in and taking control myself.

Individual opinions will always differ, but improving the "AI" system is not anywhere near the top of my list of priorities for Trainz.

My opinions.
 
We all obviously play the game for different reasons and have different guilty pleasures when operating a session. I prefer to sit in the conductor's seat and watch the scenery go by. I get a kick when another train passes going in the opposite direction. Seems realistic to me. I even wave at the other train once in a while! Crazy ain't I.
 
What a load of codswallop.
There is nothing amiss with the A1 driver commands.
I like many other users have multiple consists running in a session.
My route session runs all day, yes with hiccups at time when my command structure or other scheduling errors are present.
I make notes about the errors/pauses and correct them next day, and keep on performing this exercise until all errors have been solved.
Once I have achieved this I then add more scheduled consists.
A1 commands and application are not the only features which can cause problems.
Missing junction switches, disconnected track etc are some of the worst features to be determined as holding up the traffic.
 
Hey Dave,
It's not you, you have been around the block before with this game. People can say whatever which is fine. I do get a little tired of people saying there's obviously something you're missing. And I don't mean you specifically.
 
From my long experience in IT support I can say that yes sometimes the system is at fault and that fact can never be dismissed. But, again from my experience, it is mostly problems with the users or the way that the tools are being used or their expectations of how the software works.

I have been guilty of each of those things myself and I know how frustrating it is. But without clear, reproducible and confirmed evidence we are left in the dark. I would much rather N3V spend their time and resources dealing with clearly defined issues than chasing ghosts.

Someone needs to produce a working example of the AI failing in a specific and reproducible manner that is free of a mass of confusing variables.

My opinions.

As an addendum I seem to recall that N3V has acknowledged a long time ago that the AI is in need of a total overhaul, not just a "quick fix" and that overhaul will be a long term project so don't expect it in a future SP.
 
Last edited:
As an addendum I seem to recall that N3V has acknowledged a long time ago that the AI is in need of a total overhaul, not just a "quick fix" and that overhaul will be a long-term project so don't expect it in a future SP.

Yes, they acknowledged there's an issue with threads getting stuck over time which I worked with the QA Team on ages ago. The program has gotten a bit better where things used die after an hour or so, but the AI does become more and more goofy and starts the usual stupid moves including stopping and sitting at green signals without complaining at all, or sitting at a red signal, which they should do, but the next signal up the chain is green.

I have noticed too that sometimes we complain about the AI doing something but in reality, they're doing exactly what we told them what to do. I've had AI back up due to what I thought was a random reason, but the issue was really self-inflicted. Yes, had a track mark out of order, so the AI was doing exactly what I ordered but not what I wanted. Another time, the AI kept stopping at a signal and not going forward. What I thought was the same as always getting stuck after a period of time, turned out to be a direction marker facing the wrong direction. The AI took that long, meaning an hour to get to that location, then get stuck. I noticed the error message in the queue and that queued me in that one time because in the past I never looked at the messages.

This doesn't say the AI are perfect, but we do help muck up the works too and we need to babysit ourselves sometimes along with the AI.

I were Dave, I would open up the message history and see what the AI says for errors. Waiting for clearance or waiting for junction as they play Rock-Paper-Scissors, can sometimes hint there's a problem. In the past, I used to ignore those messages, but I no longer do so and will go see, or at least check the drivers either on the map or in the driver display to see if they are in fact heading to their destinations.
 
I have read all of the posts related to the trackmarks and AI issues. I have been a Trainz user for quite a long time, and I indeed agree that the AI has not improved, but rather that it has deteriorated with the last versions of Trainz. For instance: I started a new route with very few trackmarks. Lo and behold, the AI did not stop at the prescribed trackmark. It slowed down as if it was thinking about stopping (could there be some intelligence there???? hahaha!) and then merrily went on it's way. Then I thought that maybe the distance was too great to the trackmark. So I put some trackmarks in between. Same thing happened: AI slows down and then continues without stopping. Someone may ask me: "did you use the proper driver command; like drive to instead of drive via?" Yes, after all these years I know the difference. There are no messages anywhere to show what could be the problem.
Bottom line, I have no clue what is going on, except that it is infuriating. By the way, on my new route it is the only AI running.
 
I have been a Trainz user for quite a long time, and I indeed agree that the AI has not improved, but rather that it has deteriorated with the last versions of Trainz. For instance: I started a new route with very few trackmarks. Lo and behold, the AI did not stop at the prescribed trackmark.

That problem is not new. I have seen it happen in T:ANE (and if I risk brain damage by stretching my memory back further, possibly TS12 as well). There is a long history (with many forum posts) of it happening with industry active stations. As I stated in the addendum to my last post, N3V have acknowledged that the AI system is in need of an overhaul but that will be a long term project.
 
Back
Top