Disappointed with TRS19 again.....Help pls

marv999

New member
I left trs19 quite a while back, too many problems and just repurchased the game again. I thought building an AI route now will work, but I am wrong, or I am doing something awful wrong. What really ticks me off is one of Trainz's selling marks is, and I quote....."TRS19 takes you way beyond just driving trains, as you set up a fully operating railroad, managing dozens of trains under AI control, following your every instruction." Yeah right!!!

To get re-acquaintd with trs19 I setup a very simple AI controlled route with only one junction, some track marks, invisible speed signs and signals. Using the drive to or navigate to trackmark commands on trackmarks located beyond the junction not aligned with the designated trackmarks would not work. The train would not switch the junction and I get the message "no designated path". This message was constant for anything I tried afterwards. I then went through a process of elimination, removing speed signs, signals etc., without success. I finally removed the junction and the train finally went to the trackmark. During this process I am constantly switching between Driver and Surveyor using Ctl F1 and Ctl F2. I went into Edit Session and reset rules which didn't really do much that I could tell. I replaced the train driver with another and even replaced the train with no change I restarted the game and even rebooted my computer. I did not have this much difficulty with trs19 a year or so ago. I have been playi

I set up another train, track and junction on the same route. The junction was not aligned to its trackmark destination and the train did not switch the junction It stopped at the end of the track, which was very short length from the junction and reversed back out through the junction. It then switched the junction and went forward to the designated trackmark.
 
Hello marv999,

When building an AI route I set all junctions to default to the main line and place a trackmark on the mainline if there is a possibility of the train going through a siding. This way the AI only has to change a junction if it is programmed to go into the siding or yard. How you set the default direction of the junctions makes a big difference to how the AI will perform.

Cheers,
Bill69
 
Familiarize yourself with interlocking towers and mission code manager. It takes time and practice but works well.
 
Thanks guys....

Familiarize yourself with interlocking towers and mission code manager. It takes time and practice but works well.

I may have figured out what is going on, thank god. I was interrupting the test session by ending the session with Ctl F1 and then dragging the train back to its position and then making some session/route changes, and then trying to repeat the same train commands. I think this might be it. I just ran 2 trains and all worked....nice. I also changed the junction default direction to align the main line. No problems so far switching to the diverging track.

I have used interlocking towers in the past. I would like to learn about using mission code manager however. Any help there would be nice.....like pointing me toward some tutorials or something.

Thanks....I feel so much better now. Like building a route....
 
I don't use interlocking towers, way too complicated and in TRS19 there has been issues. I create sessions and usually the AI behaves, but in some odd situations, the AI will not follow some commands. This seems to be related to large routes with a massive number of triggers. Generally I can get half a dozen AI train performing well, and I know of a few people who manage more. Best advise is to keep it simple and don't overdo trackmarks. Expect a working AI to occasionally refuse to work, it's just the way it is!
 
For AI to recognise a path a train/loco must also have at least one signal, or buffer (which acts like a signal), a head of it along the path for it to work. When a path is not recognised other than checking that points have their levers and you have picked the correct track mark from your list etc. perhaps try stopping the schedule (and the train in the process) then moving the train forward along the path one signal, junction, etc. at a time manually before re-selecting 'Continue Schedule'. until AI takes over and you have passed the issue which in turn can hopefully be resolved. Peter
 
You can write and offer suggestions to cure the problem. But in reality there is a bug. My case: A consist arrives to a station or trackmark. Then the command to proceed to another trackmark moves the consist; but there are 2 switches. The first changes ahead, but not the second. The train sits there forever blocking other traffic nd creating a bottleneck. But wait, there is more. When you set all this, it works perfectly, or the train stops, thinks and turn both switches proceeding to next destination, so this means I am setting things as they should. The problem happens on the second time the same consist tries to pass or do this part. Or worst after you save the session and continue it later. If you set turnouts to a proper path for this consist, you screw the others, and or when another consist try to turns switches set for the other path, it would not do it. Clearly it is a bug, and to make things worst, this same set-up was working on previous iterations, but not on this latest one.:'(
 
YOU ARE CORRECT 100%......it is bugged

You can write and offer suggestions to cure the problem. But in reality there is a bug. My case: A consist arrives to a station or trackmark. Then the command to proceed to another trackmark moves the consist; but there are 2 switches. The first changes ahead, but not the second. The train sits there forever blocking other traffic nd creating a bottleneck. But wait, there is more. When you set all this, it works perfectly, or the train stops, thinks and turn both switches proceeding to next destination, so this means I am setting things as they should. The problem happens on the second time the same consist tries to pass or do this part. Or worst after you save the session and continue it later. If you set turnouts to a proper path for this consist, you screw the others, and or when another consist try to turns switches set for the other path, it would not do it. Clearly it is a bug, and to make things worst, this same set-up was working on previous iterations, but not on this latest one.:'(

I thought for sure I am doing something wrong, because I considered N3V people are professionals and its gotta be me. However, I just spent many hours attempting to troubleshoot this to pin point a cause and effect. I am a retired engineer and so I think I can recognize a problem. I just finished trying to get the train to pass over two consecutive junctions with junctions set to the trainz favor. Signal light even indicated red over green but I still get the driver message "no path to destination". I moved the track marker between the junctions and the train responded correctly. I have tried different trains, drivers, even clearing the set rules in the session, anything I could think of I have tried. N3V has messed this up big time. I just restarted playing with trainz after a long absence and the AI wasn't messy like this. I quit before mainly because of the graphics and merge problems I had trying to get my route from Tane to trs19 and something serious went wrong, lost the route. After attempts to recover it did not work, I quit.

Not certain what to do now, since my interest is mainly building an AI route.
 
You "moved" the route from T:ANE to 19. I did not. I just copied it: Keeping TANE with everything there was. So, now I can compare. Hence my conclusions, and if I don't like it, simply I go back and nothing lost (other than frustrations and a little money). Now, one day N3V fixes things, and I'll try again. I am also retired engineer and like to keep things as simple as possible.
 
KISS is the name of the game. I love big complex routes, but I found they're impossible to keep up and running as if the session and the AI are made from spinning plates on sticks. The more you keep balancing and running, there's more of a chance that the AI will do their own thing, randomly almost, to screw things up. The problem, I was told by QA, is the AI logic its self does need to be addressed because the multiple data threads within Trainz get bogged down and the AI loses their information. The data bog-down takes place over time as the route runs.

The route that got me in touch with N3V on this is a merger of multiple large routes. I took the old East Kentucky and merged that into Dave Snow's Ozark Valley, which I connected to another route that led to Jointed Rail's Midwest Grain and some other routes I can't remember. The total drive time on this route, if it ever worked properly, would be about 5 hours. The problem, however, is it's impossible to drive the whole length before things begin to fall apart.

The route could run as planned for about 45 minutes to an hour before things started unraveling as if someone pulled a thread out of a rug. First the AI would start to get stupid and sit at stations and signal even if they could drive. The trains were loaded, yet the AI wouldn't proceed to the next stop. Signals were green, yet the AI sat and stared at them. If I pressed the pause key and waited a few minutes, and this only worked once, by the way, the AI would wake up and everything would be back to normal for awhile, well maybe another 30 minutes.

Just after the pause, I discovered that sometimes if I saved the route I could reload it and the AI would drive for another hour or so before stuffing things up, so I would save periodically and save and continue. The problem now is the save and reload bug with everything in what I refer to as an unsteady state like we used to see with TTL logic that didn't have filtering or proper grounding so that when the system powered up the logic was all over the place. During this time, the signals took forever to wake up, meaning they would remain red or stay dark. The AI drivers, however, were off on a free for all and driving along at their own free will like a bunch of monkeys that escaped from the primate building at the city zoo. During this time, they were running red lights, ignoring track marks, blowing through the signals anyway, and derailing. The driver session, however, was reporting SPADs as if it mattered even though the signals weren't awake yet. Eventually, things would settle down, but by then everything had unraveled so far, I couldn't do anything with the session and I'd give up in frustration.

With all that going on, I have found keeping the sessions smaller and routes shorter makes for a more enjoyable time. This is fine for short runs, but that doesn't help us with long North American routes that technically could run for 100s of miles across the open Plains, or through the Rockies and Appalachians.
 
KISS is the name of the game. I love big complex routes, but I found they're impossible to keep up and running as if the session and the AI are made from spinning plates on sticks. The more you keep balancing and running, there's more of a chance that the AI will do their own thing, randomly almost, to screw things up. The problem, I was told by QA, is the AI logic its self does need to be addressed because the multiple data threads within Trainz get bogged down and the AI loses their information. The data bog-down takes place over time as the route runs.

The route that got me in touch with N3V on this is a merger of multiple large routes. I took the old East Kentucky and merged that into Dave Snow's Ozark Valley, which I connected to another route that led to Jointed Rail's Midwest Grain and some other routes I can't remember. The total drive time on this route, if it ever worked properly, would be about 5 hours. The problem, however, is it's impossible to drive the whole length before things begin to fall apart.

The route could run as planned for about 45 minutes to an hour before things started unraveling as if someone pulled a thread out of a rug. First the AI would start to get stupid and sit at stations and signal even if they could drive. The trains were loaded, yet the AI wouldn't proceed to the next stop. Signals were green, yet the AI sat and stared at them. If I pressed the pause key and waited a few minutes, and this only worked once, by the way, the AI would wake up and everything would be back to normal for awhile, well maybe another 30 minutes.

Just after the pause, I discovered that sometimes if I saved the route I could reload it and the AI would drive for another hour or so before stuffing things up, so I would save periodically and save and continue. The problem now is the save and reload bug with everything in what I refer to as an unsteady state like we used to see with TTL logic that didn't have filtering or proper grounding so that when the system powered up the logic was all over the place. During this time, the signals took forever to wake up, meaning they would remain red or stay dark. The AI drivers, however, were off on a free for all and driving along at their own free will like a bunch of monkeys that escaped from the primate building at the city zoo. During this time, they were running red lights, ignoring track marks, blowing through the signals anyway, and derailing. The driver session, however, was reporting SPADs as if it mattered even though the signals weren't awake yet. Eventually, things would settle down, but by then everything had unraveled so far, I couldn't do anything with the session and I'd give up in frustration.

With all that going on, I have found keeping the sessions smaller and routes shorter makes for a more enjoyable time. This is fine for short runs, but that doesn't help us with long North American routes that technically could run for 100s of miles across the open Plains, or through the Rockies and Appalachians.
There are two solutions to this mess: Or N3V takes note and make an honest effort to fix, including a report here in the forums, or.. they add an statement disclosing that TRS19 is a 30 minute max playing time per session. Before we go crazy. I am not asking, this should be a request!
 
There are two solutions to this mess: Or N3V takes note and make an honest effort to fix, including a report here in the forums, or.. they add an statement disclosing that TRS19 is a 30 minute max playing time per session. Before we go crazy. I am not asking, this should be a request!

My thoughts exactly. What's so difficult with making a concerted effort to fix things? Oh, wait pretty new stuff is fun! (Snark if you can't tell.)
 
"TRS19 takes you way beyond just driving trains, as you set up a fully operating railroad, managing dozens of trains under AI control, following your every instruction." This is what N3V advertises, and it worked cause it was the booster that made me decide to try the game again.

A flat out lie, that's what really ticks me off.....especially when I feel they do know it!!! There was and there are many problems with the AI through all the different trainz games going way back. I don't think N3V will make the effort. There is just not enough people that do the route building for AI trains. Most play trainz running sessions or making content sessions that players control. So N3V concentrates on where the money is.

John.....I know where your coming from when building large routes, been there. I bought a power house computer just for this game too. Maybe I will live long enough to see N3V address the AI issues and one day build a nice route. For now, going back to some of my survival games.

 
The AI is flawed, yes, but there are solutions. Instead of having trains run DIRECTLY to say a building or trackmark, have them go through various other trackmarks with the 'Drive via trackmark' tool. It is NOT a lie, you just cannot see how to fix it. N3V will eventually smooth it out but we just need to find small fixes. I suggest you post a thread in 'Suggestion Boxcar' for how to fix it.
 
Well I've always said the AI is flacky, and the more of them you have the bigger chance for them to fail. That said, I created a session for the Pittsburgh to Altoona route which is a 4 hour drive on a passenger run and got numerous AIs running, many coming from 2 portals without any real trouble. The only problem I found was if your had to drive on the wrong (left) track to stop at a station and an AI Goods service wanted to pass through at that time, the AI would lock a junction causing a nuisance for the player.

Not all AI problems come late in the game though, some happen at the start, so it's one big mystery really. Hopefully N3V will sort the AI before the next version of the game comes out, otherwise it may not be worth getting. Best advice as said about, KISS!
 
The AI will never meet everyone's expectations, no matter how much time and resources N3V throws at it.

Real AI on real railroads is still not a reality - the only ones I know of (and there will probably be others) are the computer controlled shuttle trains that move between terminals at major airports (Frankfurt comes to mind). But they are a single units running back and forth on a dedicated and isolated track with no junctions and no other traffic - like a horizontal building elevator. Real railroads still need human drivers (even remotely located ones), dispatchers and signallers because real world problems that require human intervention will always interfere with operations.

As Alikiwi points out, the KISS principle is best and always be prepared to do the job of a human dispatcher, signaller and driver when needed.
 
The AI is flawed, yes, but there are solutions. Instead of having trains run DIRECTLY to say a building or trackmark, have them go through various other trackmarks with the 'Drive via trackmark' tool. It is NOT a lie, you just cannot see how to fix it. N3V will eventually smooth it out but we just need to find small fixes. I suggest you post a thread in 'Suggestion Boxcar' for how to fix it.

That is exactly how I build my sessions. I took a dive into triggers and what not, but they broke way too often due to session corruption after spending days setting them up. The first time I set them up it was fun and interesting. The second time around, due to glitches I couldn't figure out, was a bit more than a tad annoying. The third time was hell and a complete drudge. After that, I never bothered again. It wasn't worth being more than annoyed, so instead I use the Schedule Library and build sessions just like I did in TRS2004 using track marks, direction marks, and wait commands. The good thing with the Schedule Library is the library can be copied and pasted between sessions, so if one bites the dust, I can easily copy over the Schedule library from the old one to the new and then quickly setup my consists again. The only pain in the backside then is setting up the consists on sidings and configuring interactive industries if there are any, and I've pretty much eliminated those too because if the session craps, they have to be setup again as well. To load freight, I use the instant load and don't bother with the animated industries other than power plants and coal mines. The rest of my freight cars, i.e. tanks, covered hoppers, and boxcars, carry "stuff". Simple.

In the end, KISS is the name of the game here. In TRS2004, things worked well as crude and clumsy as that version may appear compared to today's wiz bang pretty graphics. It worked quite well once the usual sessions tweaks were done. My big custom route, not the one I mentioned here for example, could run for hours at a time and I could really get into the operations without spending 99.9999% of the time chasing down stuck AI. I know a lot has changed along the way, but fundamentally things are still the same. The problem is due to all the requests for pretty shiny globules, the basics have been lost, and so has the refinement in the product. In a sense this reminds me of those fancy restaurants that want to please everyone rather than focusing on what they do best and then put the extras on later.

In all seriousness, the company needs to take a step back and see where things are heading and perhaps take the time and fix things. It's not like we don't complain loudly here, and usually about the same issues over and over. Maybe it's time to put away the toys and fix these issues? It's these basics that need to be addressed before going forward again at the quick pace we've been seeing. This doesn't have to be a complete stop development kind of fix, but let's not rush out the service packs either. I think one of the problems is the development cycle is driven by marketing rather than by engineering. When sales and marking get their fingers into the mix, the pressure is on to meet the market expectations and sales numbers rather focus on the quality of the product. Once that occurs, the downward spiral begins. I say this from experience at the companies I worked at in the past including the Polaroid spin-off.
 
If things were kept as simple as TRS2004 and no shiny advancements were made, the competition would have blown past Trainz years ago and made it irrelevant.
Consumers demand pace is kept with the latest technology and it would be poor business to deny them. The Artificial Idiot always needs it's hand held and told specifically what to do and generally works quite well despite the increase in complexity of Trainz. If the Artificial Idiot ever gains intelligence, N3V could market it to rail companies worldwide for driverless trains.
Graeme
 
AI can only ever be as smart as those who program it - including our users.
Most issues occur because people don't use the right command at the right time, or place track marks strategically, or fail to fully consider the behaviour options perceptible to the AI driver within the route & sessions created by Trainz users.
I never blame AI for failures on my watch. It is always something I failed to do. Correct those things - and Drivers behave themselves assiduously, every time.
TRS19 is easily the best performing, most attractive, and photo-realistic version of Trainz so far to date.
Let's get these minor issues/ bugs dealt with quickly so that we can all enjoy this railroad simulator supremacy together. :)
 
If things were kept as simple as TRS2004 and no shiny advancements were made, the competition would have blown past Trainz years ago and made it irrelevant.
Consumers demand pace is kept with the latest technology and it would be poor business to deny them. The Artificial Idiot always needs it's hand held and told specifically what to do and generally works quite well despite the increase in complexity of Trainz. If the Artificial Idiot ever gains intelligence, N3V could market it to rail companies worldwide for driverless trains.
Graeme

The problem is we still have the same underlying logic which has never been improved and the old code is now overburdened by pretty shiny things. It's like taking a small cottage and adding in spires, gables, large wraparound porches, huge doors, windows, and a gazebo. The underlying cottage, without structural changes, will collapse. Right now, Trainz is collapsing under its own weight.

I saw the changes occur over time with AI getting worse and worse as versions were "improved". Things worked pretty much the same from TRS2004 to TS12 SP0. After that, things got worse and TANE came out. TANE carried on the same issues and created more, and now TRS2019 is doing more of the same.

If time was spent shoring up the AI code, along with making the pretty shiny things, then maybe everything would work better.
 
Back
Top