Is AI stupid on purpose?

Bhorton

At work on the SKLR
I was trying to make a session, and the AI used a crossover for no reason (going against track direction markers), then went into a siding, and started moaning at me because it had gotten lost.

Also, the AI started getting even more stupid after about an hour, constantly derailing on junctions that it did not change.

Is there a way to un-derp AI in Trainz?
 
Hi

If you want total control of the AI, use one of the path rules and its associated driver commands. You set the path for the train and the train just drives it which replicates the way things work in real life. The downside is that there is some work to do to set it up but it can be done with some patience. I use it on the ECML route and have no problems getting a train from King's Cross to Newcastle without any detours or headscratching by the AI.

Regards

Brian
 
I too think that AI is stupid. But i have not seen a AI drive across a Direction marker if it was with the point facing in the way you don't want the train to go. Now for the AI crossing from one track to another is that the junction may have been set by befult in that direction.
Make sure the green arrows are poining to the right at crossovers when facing them.
 
Could you elaborate on the Paths rule a little more? I've got DRIVE VIA TRACKMARK all the way across the Central Valley, which to me uses to much of the rulez commands because you can't have to many or you will get an error.

Peace in the middle east!
 
If the AI crosses a direction marker I suspect you have it in the wrong place, or a fairly common error is to have it facing the wrong direction. Post a screenshot.

The AI isn't stupid, it is actually quite clever! Setting things up the way the AI expects it is the trick....
 
Also keep in mind that AI will take the shortest path between 2 points, not the fastest. That means it could go through the yard at 10 MPH instead of the main line at 60 if it is the shortest distance.
 
Hi

With regard to the Path rule and driver commands, the ones that I use are those by brummfondel which are on the DLS. While he is no longer involved with Trainz he still has a website with explanations for many of his rules.

http://www.js-home.org/trainz/

Regards

Brian
 
AI is too smart, and controls you !

AI needs to be totally revamped, so that AI drivers do what ever they are told to do. To follow your exact commands, where a train follows all your dirrections, even if a switch is misaligned, or a signal is red, or even if a collision is immenant.

AI Drivers should blindly follow all your exact commands, in all cases, and should not automaticly stop a train because of misaligned switchs, red signals, or awating track clearance.

AI is very unrealistic.

Real engineers can, and do, make mistakes, and are not locked out of making mistakes ... but in Trainz, AI locks you out, and you can not manually override the control of locomotives and turnouts.

It would be nice if there was a trainzoptions line like: -AImanualoverride ... or a rule that allows an AI train to be manually overridden.
 
Last edited:
It would be nice if there was a trainzoptions line like: -AImanualoverride ... or a rule that allows an AI train to be manually overridden.

Just right click on the train or the driver in the drivers tab and select stop. This gives you control of the AI train and releases the points. Once sorted select continue.
 
AI Drivers should blindly follow all your exact commands, in all cases, and should not automaticly stop a train because of misaligned switchs, red signals, or awating track clearance.

I read that about six times looking for what I was missing......

The AI shouldn't stop for red signals. Oh c'mon!

The AI shouldn't stop for misaligned junctions. You have 'ordered' the train to 'A'. The switch is set for 'B'. So if the train blindly goes to 'B' instead of 'A' you are going to NOT complain about the AI?

The AI shouldn't stop for track clearance! Your goal is a driver who runs headlong into other traffic?


AI is very unrealistic.

and your suggestions above would make it absurd!


but in Trainz, AI locks you out, and you can not manually override the control of locomotives and turnouts.

Rubbish - you can change any and all driver commands at any time. You can take over the train at any time, then you can run through a red signal on the way to 'B' instead of 'A' and run headlong into oncoming traffic all by yourself!


It would be nice if there was a trainzoptions line like: -AImanualoverride ... or a rule that allows an AI train to be manually overridden.

What Stagecoach said, or what I said above....
 
The AI does sometimes make mistakes but there are usually three or four ways to get the driver to obey your commands. One, realign a junction that is diverting the train, Two, place a trackmark that is past the junction a bit closer to it, Three, move a track mark which is just before a signal to a position past the signal Four, Try a different signal, etc. etc. I will admit it is more difficult in TS12 than it is in TS2009 to get them to go the way you want but it just takes a little bit of experimentation.
If you want to see the AI working properly try my 'Auburn route kuid2:243294:22479:3' and the AI session 'Tribute to billegulla kuid2:243294:100340:2

Cheers
Bill69
 
Last edited:
I will admit it is more difficult in TS12 (quote bill 69)

I've used the same ideas with them since using 2006. You know where they are going to act up, and place tm's accordingly.
They will then show you that you have no idea what your doing, and because you weren't keeping an eye on me reverse 20 miles to get to the track mark that was in front of me before I crossed tracks and found I couldn't get back because someone put a tdm there.
 
I have found too that if you place in a Wait... command periodically, like after pulling into an industry and after unloading, they seem to remember their commands and aren't as goofy. I wonder if this has something to do with threads.

I have to admit though when they gain control of a junction, because they ignore the human driver, it becomes a wrestling match, or as I should say a punch-out at the engine house. There's nothing worse than backing up a train only to find yourself going down the wrong line because an AI train gained control of the junction a mile away! I've been able to solve some of these issues, though, by placing invisible junctions in the way. These act as stops that force the AI to look closer to rather than way down the line for the next junction.

In the end I've decided not to second guess them. I let them get themselves mixed up then later on go back and fix the problem. In the past I'd try to second guess and come up with a "plan" to avoid AI annoyances right off, but that never seemed to work - it only made things worse.

John
 
I can confirm that the load/unload proceedure itself is threaded, as it's in the GenericPassengerStation.gs and BaseIndustry.gs files.

Shane
 
'Wait' commands certainly help, and they are also realistic. A train doesn't pull up, uncouple, move forward etc without a 'wait' between, though I have never found the need for one before or after a 'load'/'unload' command.

There is a very handy driver command 'Release Junction', iirc by atilabarut. If it is enabled in a session it can be slipped into an AI drivers command line whenever necessary, I used it a bit in the original Clovis session where it was possible for an AI driver to claim a crossover lever while a considerable distance away.

I often think that session AI issues are caused by over-specifying commands, too many trackmarks etc. I have a 'Belen Dispatcher' session in TRS04 which has run (with saves) for nearly a real week. The route trackmarks etc are as per the released route. Portals at both ends spawn every 30 or 35 minutes and no train has a command line more complex than: one Drive Via TM, one Drive to TM List, one Wait and than a Drive to Portal. The Dive Via is to avoid the Jarales cutoff which is a shorter route and therefore the preffered AI route, the TM list puts the train at a pitstop refuel station where it waits to simulate refuelling, then off to the portal.

As I said, it has run for four or five REAL days (I forget, haven't fired it up in ages!), replicates the prototype traffic density on the BNSF Transcon over a distance of 120 miles, is totally in the hands of the AI and total human intervention is none! I just sit back and watch the trainz sort it all out for themselves, including numerous 3-way meets that I've managed to watch so goodness knows how many snuck through behind my back!
 
Last edited:
I think the OP is quite justified in calling the AI "stupid", especially in TS12 with the locked junctions nonsense.

Recently I tried my route with only one train on it. It stopped at a junction somewhere, and the message on the signal was "line ahead is occupied". Is that not "stupid"? There were no other trains in the route!

So it then changed the points to the other direction, and off the train went, to a dead end that could not possibly lead to the destination. Realizing that it had done something "stupid", it backed out again, but then tried the same thing again! Anything but the proper action!

Sorry, but in my mind, that is the definition of "stupid". The present moronic behaviour of the AI has reduced my enjoyment of Trainz quite considerably.

Mick Berg.
 
There should be a happy medium between manually driving in Cab mode ... and AI's ludicrously whimsical automatic control.

I have tried STW (SwedishTrainzWorks), Drive, Just Drive, Autodrive, Autopilot ... and many other drive rules.

Presently I have AI Trainz that are impossible to stop, I can not start them, I can not abandon schedual, I can not jump onboard, and I can not assign a driver to jump onboard, and I can not throw turnouts ahead of the AI Train.

There should be a rule such as "Artificial Manual Intellegent Drivers".

AI controls you, and makes you jump through hoops to get things to function.

You do not control AI ... You are a slave to AI's nonscencical whims and programing.

Don't you just hate it when your automobile gets a "Check Engine Light" and completely shuts down your car engine, all because of one small resistor failing in a sensor, and you have to call a tow truck when you are 700 miles from home. What if airplanes were that way ? The Hal 2000 computer: "I'm sorry Dave ... I can't do that" ... CRASH

http://www.youtube.com/watch?v=R_qE6gMlou8
 
Last edited:
Sometimes, I repeat sometimes, for no aparent reason, you set everything to work the way you want it and it works! the train goes back and forth from A to B, uncouples, passes to the other end, couples and continues. So far, so good. Then, in one of these instances, the AI stops and refuses to do what it was doing all along before. You may say I changed something.. No, by the time I get to this I have made many tests and trials. AI sometimes acts stupid.. it switches points behind the last car so it derails, can't see beyond 3 or 4 points and sits there not knowing what to do, slows down several miles from a caution or reduced speed and can even switch points ahead so another train derails.. among others. It needs to be revamped, perhaps a thread with all the observed faults to bring them to the attention of N3V?
 
Back
Top