Is AI much better in 2019?

[JonMyrlennBailey]

Do they drive from track mark to track mark now without slowing down momentarily as they cross the track mark on a time table? Do they now keep a steady hand on the throttle crossing track marks and junctions? Do they now conform to real-world physics? Is braking and acceleration now realistic under 2019 AI? Does the train gradually slow to a stop now like a real train and not a sports car with disk brakes? Is their range of vision now limited for signals? Do they now wait until they get into normal human range of vision before reacting to/adjusting speed for a given non-green signal color ahead? If you were a human driver of a real-world loco, how far could your naked eyes see a RR signal ahead? 1,000 feet?, 2,000 feet?

 

[KotangaGirl]

Do pigs fly? No AI is still much the same as ever it was, but at least we now have a wider and improved range of driver commands to keep it on task.
With distant signals (Uk signal practice) I sight along the track at about footplate level and at the point where I can just see the distant signal is where I plant an invisible signal. That way AI doesn't do any sudden slow downs while it's still three quarters of a miles away.

 

[JonMyrlennBailey]

Do pigs fly? No AI is still much the same as ever it was, but at least we now have a wider and improved range of driver commands to keep it on task.
With distant signals (Uk signal practice) I sight along the track at about footplate level and at the point where I can just see the distant signal is where I plant an invisible signal. That way AI doesn't do any sudden slow downs while it's still three quarters of a miles away.

Thank you. There's no need for me to run out and buy 2019 any time soon, then. There might be "something even better" two years from now. Always hold off until something really worthwhile comes along worth spending money for. Can routes/sessions created in older Trainz be imported in to 2019 even? It would be so nice if people could get a free trial of these new things, a free test drive, each and every time a "new" and "improved" one comes out.

 

[pware]

Of course it is always a matter of personal opinion and preferences, but I deleted T:ANE from my system almost as soon as I installed TRS19 Platinium and have since upgraded to Trainz Plus. The UDS feature of Platinum was enough for me to dump T:ANE and a decision to support N3V (despite all its critics) was the reason to upgrade to Trainz Plus which is a subscription. That plus the promise of a first look at the latest new features, such as Surveyor 2.0, made it a "no brainer" decision for me to upgrade to Trainz Plus.

I have had no problems in importing routes from T:ANE directly into TRS19 PE and Trainz Plus. All TRS12 routes that I have tried (but there have only been a few) have also imported well but beyond that (TRS2010, 2009, etc) it could be a different issue.

 

[JonMyrlennBailey]

If any newer edition of Trainz doesn't show any marked improvement in graphics motion quality (no shaky slide shows on lineside, chase or free-roaming cameras) and operational behavior (real-world physics), then there's no reason for me to invest in it. My dream is for this PC software toy to look, act and sound like real-world vehicles, aircraft, boats, people, animals, buildings, trees, water, weather and lighting as much as humanly possible and economically feasible enough for the average Joe to buy. Making a consumer-market sim real-world-like is the ultimate challenge of computer science. It all boils down to the 1's and 0's.

There are physical model train layouts with train models that don't conform to the look and feel of real-world physics and geometry. This jittery train has quite sloppy track work, curves too tight and braking too abrupt:

https://www.youtube.com/watch?v=EbkUMCkPWaE

Sloppy model RR layouts are a waste of good money and space.

 

[pware]

There are physical model train layouts with train models that don't conform to the look and feel of real-world physics and geometry.

Sloppy model RR layouts are a waste of good money and space.

"Sloppy trackwork" is not found only in model (or virtual) railways. The "real world" has many examples.

https://www.youtube.com/watch?v=9X2A2f6E5DI

I suspect that what you ultimately want in a railroad, road, ship or flight simulator does not exist yet in any form that is both practical and affordable for the "average Judy or Joe". Perhaps Hollywood is approaching the level you want (I just watched the new movie "Free Guy" which approaches that level). While this will eventually reach consumers I can give a 100% guarantee that it will not run on today's consumer PCs.

A large part of the problem is the current consumer technology. For obvious cost reasons, most Trainz users are well behind when it comes to using the best available optimised hardware. I recently upgraded my old (5 years+) system to the latest best available CPU with more RAM and an RTX video card - it is a whole world of difference. No shaky slideshows or stutters in even the most demanding Trainz routes running on TRS19.

Simulation developers (and I include Route creators) will naturally create for the best system that is currently available because that offers the longest run for their work.

Of course, within a few years (too few unfortunately) my current "new" technology will be obsolete and Trainz creators will be building routes and sessions that will start to have "stutters" and "shaky slideshows". Then it will be time to upgrade, both the hardware and the software, again.

Such is life!

 

[Alikiwi]

As someone who makes a lot of sessions, well 81 uploaded to the DLS so far I can assure you the AI in TRS19 is a lot worse than Tane. If it has a destination especially in a yard that has a lot of junctions, and it has to throw just one junction to get there, it'll sail past and keep on going and thereby not reach any of it's destinations. I've mostly given up with sessions in TRS19 and gone back to Tane. It's a rather hit and miss affair, but clearly the AI needs major work and until that's done the game isn't going to improve. The larger the layout, the larger the problem. Also depends how well the route is built especially regarding signalling.

 

[pware]

As someone who makes a lot of sessions, well 81 uploaded to the DLS so far I can assure you the AI in TRS19 is a lot worse than Tane.

I personally do not really see any differences in the AI between the two. But, that said, I must add that the Routes and Sessions I create make very little use of the AI. I don't create commuter or busy city and industrial layouts, mostly isolated, remote single track branch lines and mainlines with very little traffic. The AI, when I use it, works perfectly for me.

The larger the layout, the larger the problem. Also depends how well the route is built especially regarding signalling.

That is the crux of the matter. The use of signalling, including invisible signals, trackmarks, triggers, rules, driver commands, direction and priority marks are the key. Too few creators make sufficient use of these tools and then start to blame the AI for not "thinking" like a real signaller or driver who would have had years of experience behind them.

My thoughts.

 

[JonMyrlennBailey]

As someone who makes a lot of sessions, well 81 uploaded to the DLS so far I can assure you the AI in TRS19 is a lot worse than Tane. If it has a destination especially in a yard that has a lot of junctions, and it has to throw just one junction to get there, it'll sail past and keep on going and thereby not reach any of it's destinations. I've mostly given up with sessions in TRS19 and gone back to Tane. It's a rather hit and miss affair, but clearly the AI needs major work and until that's done the game isn't going to improve. The larger the layout, the larger the problem. Also depends how well the route is built especially regarding signalling.

I use AI mostly to have scenic drivable vehicles (railway trains, streetcars, speeders, boats, trailer trucks, helicopter, horse carriages and even a handcar) running passively in the background. The train rolling stock hand laid on tracks in Surveyor are much prettier than the ones randomly emitted by portals and the hand-placed consists are completely customizable and ridable with camera views. I'm experimenting now with part-time AI-driven trains. AI drives from customer siding to customer siding as well as does the train movement in the yard, turntable and roundhouse while I do the cutting, hooking and dropping of freight cars at sidings by hand. I use Stop Train to halt a time table temporarily. I use Continue Schedule to go back into AI mode again after the customer visit. The only drivable content I have not yet discovered is automobiles, passenger cars for motor roadways.

 

[KotangaGirl]

AI needs to be given a 'chicken feed' trackmark trail through junctions and complex track formations. It's better to be a little too generous with trackmarks than to have not enough of them. My Norfolk (Uk) layout is not that big compared with some at only 40 miles in total length, but the major stations have some fairly complex trackwork and it's by laying a string of trackmarks for each empty stock working between platforms that I get it all to work. I tend to look for simple solutions first before reaching for the more esoteric driver commands, but so far I've been able to find solutions to scheduling problems without raising my blood pressure or resorting to bad language.
Proper signalling is definitely a major thing to get right. I'm a signal junkie* so nothing makes me happier than finding the right sweet spots with signal placement.

* Uk semaphore signals only and the older the better.

 

[Alikiwi]

Talking of junctions, I found if there are more than 4 between trackmarks, then there's a good chance the AI will stop and not know what to do. So trackmarks yes, but on long runs without any junctions, keep them to a minimum or the AI will slow down to 12 mph at each trackmark. Too many, and too few are both issues, you just got to love to hate it lol.

 

[JCitron]

Like a lot of things with Trainz, the underlying AI engine, meaning code, needs a bit more than a lick and a promise sprucing up. Like a lot of things in Trainz, we're expecting a lot of older assets, in this case driver commands, triggered-features, and other nifty things, to work in new code even though they were created more than a decade ago. Many of these commands and things never worked particularly well even when they were first created and now they've been thrown on to code that's further away from what they were originally written for. In all intent and purposes, it's amazing anything works this well at all.

In the end, I do things simply. I find, a liberal splattering of track marks in critical places, such as yards and other places with lots of junctions, does a demonstrable job of keeping the computer-driven trains happy. I have also discovered that after more than a decade of Trainzing, going on two decades of Trainzing, that I can anticipate what the AI will do while I design my routes and plan accordingly. I have also found signaling logically, not always prototypically, works as well, and sometimes having too many signals can get the AI completely frustrated. One of those places where the AI has a snit is a yard. Placing signals with the heads facing in towards the yard tracks, will definitely cause the AI to stop and think about each track. I noticed that this is related to the newer signal code introduced in TANE SP2 when the data model changed to what we have today. After dawn lit up over Marblehead, (Yes, Marblehead is are real place in Massachusetts near Salem, and it's on the Atlantic Ocean), I remedied the stalling by removing the entrance signals, and the AI switcher is happy again.

All and all, I have found that, outside of outright coding problems, that a good number of our AI problems are self-inflicted by us. There are many occasions, when I've blamed the AI for doing things that I did in fact tell them to do like switch tracks because I did in fact send them to the wrong track marks! This doesn't say that there aren't coding problems either, and Tony has said that the underlying AI code needs to be rewritten. The problem, however, lies in two areas here. First changes to the code will break everything we already have, and second we have time, money, and budgets. This latter issue drives a lot of what businesses do, and even though we have "stupid AI drivers", at least things are still working for now. We have to just anticipate what they may do and bring out the band-aides and bailing wire when needed, and wait patiently until N3V comes up with a better AI engine that will hopefully solve our problem.

So in the end, to answer Jon's question. The AI really isn't much different than it was in TS12. TRS19 has smoother graphics and better running than even TANE had, and TANE was lightyears faster and much better than TS12 was. Like all computer-related stuff, we're forever chasing that perfect combination hardware, software, and reality to give us the perfect train simulator, but as pware said that is impossible even with today's consumer hardware and operating systems. Even if we could run one of those 256 core CPUs with 256 PB memory, super-size SSDs, and multiple high end video cards, we may have a super system, but Trainz or any other game would never run on it because the OS that supports our programs doesn't run on hardware like that.

 

[boleyd]

Also, some equipment does not work PROPERLY with AI. I now change equipment before blaming AI. Just a simple wait for trackmark will not work with some engines in my route. It may in others as timing may be the issue. Small tweaks through the years accumulate potential conditional error traps for the customers. N3V has a major challenge. Is it easier to audit all engines to conform to a known set of proper process or rewrite the AI stuff. I would vote for fixing the equipment and NOT messing with AI. If the AI code is well segmented, using strict rules for intercommunication, then fixing it may work. But if it is a haphazard mess of unmanaged patches then it is like a field of left-over land mines.

 

[pware]

Much, but not all, of the behaviour of "equipment" (loco, signals, etc) is controlled by user created scripts that are a part of the asset. These scripts can vary enormously in how they work, their timings, built-in delays, compatibilities, etc.

It can happen that when N3V make "under the hood" changes to Trainz, the "Trainz Engine" as it is often called, it breaks some of these scripts so the asset no longer behaves as it once did. Some users insist that N3V should never make such changes to prevent this problem but that would mean that Trainz never advances, never gets new features, never improves, etc.

 

[G.M.]

hi Jon,
Q: Do they drive from track mark to track mark now without slowing down momentarily as they cross the track mark on a time table?
A: it helps if you use "via" commands if you want the train to advance without stopping


Q: Do they now keep a steady hand on the throttle crossing track marks and junctions?
A: Who is they?


Q: Do they now conform to real-world physics?
A: no, but we can emulate and take time to tune engine files
there is a new TNI physics addon possibility, but don't see much action in people trying to develop for it
who knows in the future.


Q: Is braking and acceleration now realistic under 2019 AI?
A: no, there is too much difference between "cab mode" and AI


Q: Does the train gradually slow to a stop now like a real train and not a sports car with disk brakes?
A: depends on how the train was made, specially the engine file is important combined with the mass


Q: Is their range of vision now limited for signals?
A: don't understand the question


Q: Do they now wait until they get into normal human range of vision before reacting to/adjusting speed for a given non-green signal color ahead?
A: no, a workaround is use hidden signals.


Q: If you were a human driver of a real-world loco, how far could your naked eyes see a RR signal ahead? 1,000 feet?, 2,000 feet?
A: Every human is different and so are their eyes, then we have the time of day, the weather condition, no easy answer.
in trainz you can help yourself by using the "track-view" rule, this is ergonomical more correct top-bottom
as opposed to the native trackbar underside of your screen left-right.


Trainz may not be all that accurate out of the box, trying to improve that
by making routes/sessions/content, is fun and a challenge
Because it's an open system you can make it as real as you want.
just have fun with Trainz.

 

[pware]

Because it's an open system you can make it as real as you want.

To expand on that, not everyone is interested in achieving total realism in Trainz simulations. There are "fantasy" routes which go to the other extreme and have totally unrealistic physics, impossible track conditions and "weird" train operations. Viva la Difference!