Things about the AI (artificial intelligence) component that are not prototypical.

[KotangaGirl]

I personally couldn't care less about the AI, but it would be nice to have a moon in the night sky........ (runs away and ducks for cover).

 

[obirek]

Trainz for cavemen

One thing that I have learned in my career in IT is "never say never"

I agree. To a cavemen rocket seemed impossible. We are the cavemen of XXI century, yet we dare to judge what cannot be done.

 

[ARhian7]

"Never say never" is so true. In the past I have said "never" over various things and unfortunately it has come back to bite.

As for the AI, I enjoy using them. One day it would be nice to have them all animated in cab and if possible, entering and exiting the loco etc. Anyway, I am just very pleased that they are in Trainz.

 

[PC_Ace]

Gotta say that I think creating AI sessions involving multiple drivers and consists delivering goods and passengers to many interactive industries and stations over a complex route is one the most satisfying logistical and programming challenges that Trainz provides.
Beside creating such routes and sessions, I will work tirelessly in Surveyor mode fine-tuning the AI driving commands and timings, signals, junctions, Trackmarks, etc. to finesse running completely autonomous Driver sessions for hours on end with more than 20 consists at a time. I sit back, as a God-like Fat Controller, watching their progress in both Map View and Chase Mode.
If I ever need to intervene to correct an AI driver error after several hours of runtime, then it is MY Fault! That is my cue to go fix that issue in Session Edit mode to prevent it from recurring.
I'll never let the machine beat me, so when T:ANE and TRS19 are working at their best - as they are right now with Beta Builds 98700 and 98592 - then seamless, continuous AI operation is both possible - and attainable.
Each to their own, however. I enjoy creating and running entire railroad operations scenarios where resources are picked up and delivered on time and in the correct quantities and without any cornfield meets or traffic violations.
Others prefer to drive a loco performing a number of tasks; whilst still others love to spend all their time in Surveyor, creating Worlds of their own.
C'est la vie!

 

[blueodessey]

Great AI discussion in here,

Some thoughts for consideration.......AI is still a Baby in development and probably will be for the foreseeable future.

Now think back when the first simple computers were made, huge tape drives, punch card programing, ones and zeros was all the machine understood? Fast forward to today, do you believe what your computer can do now, Graphics, Cortina and the amazing Graphical Games we have.......Take all this in perspective......How did we get here? A heck of lot people around the world developing hardware,,, Software and all the glitzy gadgets,, such as your awesome Game Mouse........

Now think about AI's humble beginnings,

I would think the simple Transistor was the start of major change with Electronics in it's time, now look at what we can do with Electronics. For example a lot of Ham Operators can literately do their "QSL" via Software and a Computer, even a Smartphone.......Small and portable and amazingly powerful more powerful than the computer that took our astronauts to the Moon!

Now think about how much research and dollars were spent and how many failures were made in the process of developing what we have now........Probably much more than anyone might think........

While I think about, if I recall correctly our Jetliners land themselves at airports, and Pilot is there to monitor the Ride, and take control if needed,,I don't know about you, but that is nothing short of amazing for AI control and performance.

And were back to AI,,,,,,,,How much money does NV3 have to spend on that part of the program, and still maintain a market lead against other competing Software Companies. I'm sure we could belabor the point back and forth to infinity, you can't please all the people all the time, it is simply impossible.........I would think we just have to use AI judicially and understand it's strengths and weaknesses till something better comes along.......?

Also think about the fact, there are a ton of programmers in the world, making every kind of software program imaginable, fancy the fact that maybe a software addon for Trainz could be made, to handle the AI events we need for handling a proper Rail events and tasking. Think about PWare's amazing Path Control Objects he has made for us............And what you can do now for automatic path switching etc.......Nothing short of amazing talent there........

TRC or Boats Crossing control systems we use..........

Take ASSET-X program and the amazing problems it handles in the Trainz world........

And I have been frustrated to the max with AI, but have learned that sometimes there are points of no return, so like a Child, you would never leave them alone running the Railroad, until they grow up and mature to the point we deem responsible for it.s actions. I'll continue fooling around with AI, but I surely won't depend on it a 100% the same way I do with any software program, I learned that the hard way thru 20 yr's of IT work and computer software management and implementation.......

For me, patience is virtue and time will tell.........I pray for the best outcome............And I wouldn't trade Trainz for anything, it still is the best Rail Simulator around for what you can do, and what it will do for you if take the time to look under the hood so to speak..............

 

[cascaderailroad]

I would think we just have to use AI judicially and understand it's strengths and weaknesses till something better comes along.......?

You have gotten AI … You will just have to live with it's inadequacies … They have bigger fish to fry

 

[Axeman3813]

I enjoy pushing the AI in my virtual trainset. Setting up a world is awesome, laying track for airplanes, trains, and ships. 2019 runs great, but the ai trains are having issues with models"industries" that worked within TANE. Not stopping for them, not loading, unloading are a few instances. You can work around this with track markers, but what happened to the industries that have them built in, "seaport".

I had 32 trains running at once within TANE, still working with 2 in 2019. At least its relaxing to switch it up and go for a Drive in some of the scenic routes.

 

[llebrez]

Great AI discussion in here,

Some thoughts for consideration.......AI is still a Baby in development and probably will be for the foreseeable future.

Now think back when the first simple computers were made, huge tape drives, punch card programing, ones and zeros was all the machine understood? Fast forward to today, do you believe what your computer can do now, Graphics, Cortina and the amazing Graphical Games we have.......Take all this in perspective......How did we get here? A heck of lot people around the world developing hardware,,, Software and all the glitzy gadgets,, such as your awesome Game Mouse........

Now think about AI's humble beginnings,

I would think the simple Transistor was the start of major change with Electronics in it's time, now look at what we can do with Electronics. For example a lot of Ham Operators can literately do their "QSL" via Software and a Computer, even a Smartphone.......Small and portable and amazingly powerful more powerful than the computer that took our astronauts to the Moon!

Now think about how much research and dollars were spent and how many failures were made in the process of developing what we have now........Probably much more than anyone might think........

While I think about, if I recall correctly our Jetliners land themselves at airports, and Pilot is there to monitor the Ride, and take control if needed,,I don't know about you, but that is nothing short of amazing for AI control and performance.

And were back to AI,,,,,,,,How much money does NV3 have to spend on that part of the program, and still maintain a market lead against other competing Software Companies. I'm sure we could belabor the point back and forth to infinity, you can't please all the people all the time, it is simply impossible.........I would think we just have to use AI judicially and understand it's strengths and weaknesses till something better comes along.......?

Also think about the fact, there are a ton of programmers in the world, making every kind of software program imaginable, fancy the fact that maybe a software addon for Trainz could be made, to handle the AI events we need for handling a proper Rail events and tasking. Think about PWare's amazing Path Control Objects he has made for us............And what you can do now for automatic path switching etc.......Nothing short of amazing talent there........

TRC or Boats Crossing control systems we use..........

Take ASSET-X program and the amazing problems it handles in the Trainz world........

And I have been frustrated to the max with AI, but have learned that sometimes there are points of no return, so like a Child, you would never leave them alone running the Railroad, until they grow up and mature to the point we deem responsible for it.s actions. I'll continue fooling around with AI, but I surely won't depend on it a 100% the same way I do with any software program, I learned that the hard way thru 20 yr's of IT work and computer software management and implementation.......

For me, patience is virtue and time will tell.........I pray for the best outcome............And I wouldn't trade Trainz for anything, it still is the best Rail Simulator around for what you can do, and what it will do for you if take the time to look under the hood so to speak..............

For the most, you are very right. Now consider:

AI is acceptable to me if you understand how it works and what it can do. You set the session and it works. The problem is when you save the session several times. Then it fails. At that point, I wonder if the problem is the AI itself, or the process of saving that fails? Think carefully on that.

Lets do a quantitative evaluation. How many saves before it fails to perform, as when you set it?

 

[blueodessey]

I wish I knew how to program and I would fix that.... Sounds like a variable is out of whack or a if then clause is not right.... Shouldn't be that hard to fix.....

Wish you good luck....

 

[KotangaGirl]

I build small backwater 'scratch in the dirt' minor Uk railways/steam tramways and the way I use AI is to instruct the robot driver to take the train to a particular water tower on the line while I'm doing something else and once the train gets there it stops and waits until I can get to doing whatever is needed. Mostly shunting (switching) as it happens since I have docks and industries along the line. I like to drive my engines which are all steam engines, since my layouts are generally circa 1900-1913 and they are small and characterful, and are in some cases set up in their engine specs to be a bit difficult natured so they want 'knowing' in order to drive them well.
I suppose my small single line tramways with no more than two or perhaps three engines at the most in steam are the complete antithesis of some of the layouts mentioned on this thread with dozens of mile long trains doing their thing with robots in the cab. BUT I like it and it makes me happy, - and that's what Trainz is all about isn't it.

 

[1611mac]

3 month TRS19 user here... I can see fun, joy, or satisfaction (what every you want to call it) from just about any aspect of Trainz. Heck, TRS19 has even helped me with my new railfanning (and virtual railfanning) interests. In TRS19 I like to drive. From the little I've done with Surveyor I've loved it too. There's the "technical" side of things... there's the "artistic" side of things... and there is the "is it accurate" (prototypical) side of things. I loved running the "Dispatching" session in Kickstarter County 2. Once you beat the game requirement it's a challenge to try to beat your own times. Setting there watching the AI drivers "execute your plan" is awesome. Trying to find out what a driver is doing and why... well, that just adds to the challenge.

I can spend hours/days/weeks running small to medium route sessions and also running my own "live on the fly" sessions on routes like Franklin Ave, philskene HO routes, and others. It really doesn't bother me that much when I run an "older" route (for me that mean pre-TRS19... 4 months ago.. Ha) - graphics and such don't have to be the best.

What I don't personally get are the long routes where you start and then 30 minutes later you stop. Or, if all you are doing is racing to 8 to 12 passenger stations trying to beat a clock. But then, that's just me.

And that is what makes this "game" (simulator) so unique. It can be a whole lot of things to a whole lot of people with different desires and goals.

 

[obirek]

6. Yesterday I drove one of my sessions in TC. I issue command "Drive to trackmark Newcastle 6". What AI does? It drives first to trackmark Newcastle 8, which is on the next track, then it pulls back to the spot it started from and only then it drives to trackmark Newcastle 6, as expected. But when I issue order "Drive to trackmark Newcastle 8", it drives to either trackmarks Newcastle 6 or Newcastle 10, which are on the tracks to the right and left of trackmark Newcastle 8, pulls back and finally finds its way. I thought Trainz is not about the moon and stars, but driving from A to B!

 

[pware]

The big mistake is to assume that the AI "thinks" like normal people (excluding you and I of course). Sometimes (some say often) the AI sees obstacles or paths that we don't see and may or may not exist. Partly for that reason, complex paths are often best navigated by placing intermediate trackmarks and using the "Navigate Via Trackmark" or "Drive via Trackmark" commands. Placing additional invisible signals can also help but can also add to the confusion.

 

[1611mac]

I learned quickly that the AI driver will go "backwards" if it's a shorter path to the destination. Thus, yes, extra trackmarks are helpful to keep the AI train going right!

 

[JCitron]

I learned quickly that the AI driver will go "backwards" if it's a shorter path to the destination. Thus, yes, extra trackmarks are helpful to keep the AI train going right!

Yup. This is a "known" feature with the AI drivers.

I put beginning, middle, and end track marks on a line when I want the AI to stay on course. This works well to keep the drivers running the streamliners and hotshot freights from finding the slowest, most decrepit, broken, yard tracks. The first track mark encountered sets the route, the middle one brings the AI too far from the junction to change anything, and third is the destination.

The method also works well too for WYE setups. I found out the hard way when I had an AI driver decide that it was more fun to take the right leg of the wye, flip the switch the other way, and then run backwards on the rest of the route. By placing a mid-point trackmark, it forced the driver past the point of no return, therefore, no recourse but to go forward and continue on the proper route. This issue occurred in the early days of TS2010 and AI behavior is much, much better now than it was back then. Yes. Seriously it is!

This formula in general works for a lot of things AI-related I noticed including loading and unloading at industries. A strategically placed track mark before the industry, followed by a wait 10 seconds then followed by a slow speed preps the drivers for loading and unloading. I then place a destination track mark after the industry, on the industry lead, to give the AI something to drive after the driver has unloaded, and then that is followed by the exit track mark.

 

[KotangaGirl]

The only way I found to stop AI controlled passenger trains from short cutting through the colliery on one of my layouts was to always keep the yard shunter parked in the way. Nothing else I tried ever worked.

 

[Kennilworth]

Hi

There are various ways to keep the AI on track in addition to trackmarks.

You can use priority markers and set the train priority to match. Track direction markers will prevent a train from taking the wrong leg at a junction. In TS12 or earlier you can use one of the path rules and its driver commands to set the path that the train will follow and in TANE or TS19 you can use Interlocking Towers to do the same thing. If you want prototypical action then path rules or ITs are the way to go as they mimic the actions of a signalman or dispatcher by setting the route for the train.

With a little effort the AI can be brought under control and made to do as you wish. Many people don't have the time or can't be bothered to explore these options but, if you're prepared to put a little time in, the AI can be brought to heel.

Regards

Brian

 

[JCitron]

The only way I found to stop AI controlled passenger trains from short cutting through the colliery on one of my layouts was to always keep the yard shunter parked in the way. Nothing else I tried ever worked.

Having some wagons in a siding or a shunting loco on the lead, also looks good as well, but I have had AI go down the blocked track anyway and then get stuck because the other train is in the way!

Brian's right. Eventually you can get the AI to heel, but they can be like training dumb dogs that still insist on peeing on the carpet.

 

[KotangaGirl]

Brian's right. Eventually you can get the AI to heel, but they can be like training dumb dogs that still insist on peeing on the carpet.

Since I much prefer to drive my engines myself and only use AI to occasionally move a train of wagons from one place to another while I'm busy with something else, so if the 'dumb dog pees on the carpet' my reaction is to shoot it dead and bury it. I can't be bothered with mucking around for hours second guessing what the robot drivers are going to do next.

 

[cascaderailroad]

Brian's right. Eventually you can get the AI to heel, but they can be like training dumb dogs that still insist on peeing on the carpet.

I prefer to place Pee Pee Training pads all over creation … and only buy small dogs like Chihuahua's