Am I alone in thinking the AI needs much more "I"?

[styckx]

Funny when I read that I didn't see him say can't or didn't want to improve ai. I read it as him saying this is how it is now, and this is what it does and does't do.

For the future? The game has been growing and growing since 2007 using this method of ai. If it was that huge of a burden the game would have floundered by now. Its not perfect or dynamic but it works and obviously fills the void good enough. AI while not perfect defintely isn't imo a high priority problem.

[AndyM77]

Get the signalling perfect (all regions) and for us the users to learn to place them correctly, might help improve the current AI, because at times, a single placement error can cause the AI to go "What? ".

It would be great to have and AI do things "On the fly", but with lots of paths I can see lots of users complaining that their CPUs aren't up to the job (particularly since Railworks "Multithreading" isn't really all there, by which I mean that it's not designed for proper "Multicore" usage - i.e a thread running just physix on core 1, passing rendering info to GPU via core 2, etc..).

I think that "Career Mode" will lead to many improvements, but it's going to take time, and lots of biscuits..

[Acorncomputer]

I can agree with most comments here.

There are unlikely to be any significant 'improvements' to to the way AI works but ongoing development of the Career System should fine tune the existing dispatcher.

Adam's post is very good and tells us how it is and makes no promises. This is fine by me and I think it will actually help to lower expectations so that we can learn to accept the limitations and better understand the way the dispatcher works and get the best from it.

Want we want is a long way from what we are going to get but what we have is still pretty good.

[bigvern]

It would be great to have and AI do things "On the fly", but with lots of paths I can see lots of users complaining that their CPUs aren't up to the job

Just checked the specs on Zusi 2 which does all of these AI/timetable/dynamic routing etc. and it requires an 800 Mhz processor, so I don't think it's CPU cycles which is preventing this type of realism in RW.

[sanyix]

Get the signalling perfect (all regions) and for us the users to learn to place them correctly, might help improve the current AI, because at times, a single placement error can cause the AI to go "What? ".

It would be great to have and AI do things "On the fly", but with lots of paths I can see lots of users complaining that their CPUs aren't up to the job (particularly since Railworks "Multithreading" isn't really all there, by which I mean that it's not designed for proper "Multicore" usage - i.e a thread running just physix on core 1, passing rendering info to GPU via core 2, etc..).

I think that "Career Mode" will lead to many improvements, but it's going to take time, and lots of biscuits..

Pathing is not a big thing... since this is a railsim, it have limited paths to be tested, and it's a piece of cake for a cpu. I think with a reasonable path test/refresh rate (1sec) it would use insignificant cpu time.
With multithreading, the first problem is the physix, because it is limited to one thread on cpu (thanks to nv), and outdated too. There are many other better !FREE!physics engines.

[sanyix]

It would be great to have and AI do things "On the fly", but with lots of paths I can see lots of users complaining that their CPUs aren't up to the job

Just checked the specs on Zusi 2 which does all of these AI/timetable/dynamic routing etc. and it requires an 800 Mhz processor, so I don't think it's CPU cycles which is preventing this type of realism in RW.

Or just check OpenTTD it computes everything alone between waypoints, and it's dynamic, and needs how much cpu? (i play with it max 15% cpu usage on an athlon x2 downclocked to 1 ghz )

[JasonM]

I prefer the suprise factor in a railsim, just like real life you never know what is going to happen on a journey, where as with Railworks you know roughly what is going to happen in a scenraio and even more so if you have played through it once before.
That is what makes Zusi so much better in the simulating front, no scenarios just choose a train from the days timetable and off you go.
Still this is Railworks so whever we like it or not that is how it is.

[AndiS]

Thanks to Adam for the precedented precision!
Like said by others: Dash hopes if needed and take it as a concise description of what is there.
The problem with dynamic recalculation is not computing power, it is the complexity of the code.
And like I said before, they will not get away with this state of affairs for multi-player mode. But multi-player is quite a way down the road, next stop is career mode.
You can count me in with those who desire Zusi3 realism with RW graphics, released now. But it does not help me. Adam's statement did, in a way.

[FoggyMorning]

Thanks to Adam for the precedented precision!
Like said by others: Dash hopes if needed and take it as a concise description of what is there.
The problem with dynamic recalculation is not computing power, it is the complexity of the code.
And like I said before, they will not get away with this state of affairs for multi-player mode. But multi-player is quite a way down the road, next stop is career mode.
You can count me in with those who desire Zusi3 realism with RW graphics, released now. But it does not help me. Adam's statement did, in a way.

Multi player mode will surely simplify the code rather than complicate things
The computer will not have to work out any paths in advance, only tell the player when the signal they are approaching is safe to pass or not

The questions will come from what happens when a player decides to take a half hour break while occupying a section of the ECML

But as you say, all this is far in the future at present

[bigvern]

You can count me in with those who desire Zusi3 realism

Does that mean you have seen the "grail" of train sims in action, Andi?

[AndiS]

I did not see the grail in action.
But going by what I grasp on the forums, the user manual will be a handbook of German railways, worth the licence price alone.
And like everyone I saw the screenshots with the OLE. And I have no doubt that the superelevated curved switches will be correct to the nanometre.
But there are quite some parts of the grail still in the works. And steam is clearly less of a priority.
And while nothing is released, all is speculation.

Multi player mode will surely simplify the code rather than complicate things
The computer will not have to work out any paths in advance, only tell the player when the signal they are approaching is safe to pass or not

What I meant was AI in multi-player mode. AI trains will have to adapt to the new situation depending on the actions of all the players.

There are threads on multi-player mode where lots of opinions regarding issues and solutions regarding player behaviour were shared.

Getting back on topic, I would like to make a stupid joke about Is, but then again, I don't want to look stupid besides Adam's clarifications. So I could cautiously say that the AI needs an F for flexible, and I hope that it will come by and by, e.g., many agree that scenarios need some means of variation. But for the near and medium future, the AI is precomputed, maybe intelligently so, obviously looking artificial at times, but certainly not flexible, and not meant to be.

[sniper297]

Styckx, it's not a high priority simply because the overwhelming majority of users merely cruise around admiring the scenery, couldn't care less about actual operations, so it's not a problem for them. Except it IS a problem for some of them, they just aren't aware that it's a problem.

http://forums.uktrainsim.com/viewtopic. ... 06&t=95432

FOOTY FANS: "This guy gave up since he don't like the AI traffic, I don't care about AI traffic, that's his problem."

Even the Footy Fans complain about lack of free content - last time I checked the one and only free US steam engine was Ron's, so if Ron can't get AI traffic, you don't get any more of his steam engines, it's now your problem by proxy. That's what community is all about, I don't care about headlight projection or cars that stop at crossing gates, but if someone who is making content that I want cares about it, then by extension I'd BETTER care about it too if I want his stuff.

But moving on, my question;

http://forums.uktrainsim.com/viewtopic. ... &#p1164124

What I would like is for whoever is the CEO this week to make a definitive statement, "No, we have absolutely no plans of ever fixing the AI "traffic" (background scenery!) because we think it's good enough as is." I'll accept that and buy Trainz instead, go away and quit bugging everyone about it.

http://forums.uktrainsim.com/viewtopic. ... 9#p1216099

There it is, Matt, my one and only question - will they or will they not make a new scenario editor that's at least as easy to use as the MSTS activity editor, with at least as much versatility and control for the AI traffic as MSTS has?...(snip)...If the answer is "No, the AI traffic is designed as animated background scenery for multi track routes and that's all it will ever be", then I will do what the footy fans want, delete it from my hard drive and shut up about it.

One question, that's all I want answered, "further enhanced" is not an answer.

Has finally been answered - so I deleted it from my hard drive, now I'll shut up about it. If someone had answered my question a year ago I wouldn't have wasted my time and everyone else's about the subject.

Via con Dios.

[boleyd]

Don't loose hope. RS is connecting up with the Demolition Company Gmbh game producer in some sort of distribution agreement on Steam. Soon we will be able to demolish structures, load the debris into RailWorks hopper cars and whiz off to the landfill. My question is, will the Demolition Company scores be added to the RW scores and published?

Now we are getting into the fun stuff. Since this is a Green activity additional points should be accrued for avoiding spillage of the waste material. Who needs fancy signals when we can eventually merge two leading games into one master scenario and have it properly scored!

[Easilyconfused]

That is not the case and although I can see some humour in the post it is not conducive to introduce such patently false information into the forum since some people may take it as fact. Please discontinue this line of discussion before posts start getting culled.

They are 2 separate products that do not interface in any way. Once I finish writing the review for demolition simulator it will be passed onto the site admins for publishing in the review section.

[bigvern]

I think the quote from (RS)Adam(Zone) should probably be extracted and stickied at the top of the Signalling and Scenario forums so anyone asking about the "AI" can be directed to it.

However I do think they need to rethink the approach as with career mode, if you're hurtling along a single track route with passing loops, but future developments with career mode decide to delay you at "B" for ten minutes with police attending a ticket irregularity, in reality the signaller would arrange with the driver to put the signal back and run the opposing train you should have crossed at "C" to cross you at "B" instead.

Cause and effect. As being discussed in the career threads, you cannot provide a realistic driver's working environment unless the signalling and control functions in exactly the same way as the real signaller and Train Running Controller (me!) decide how to react to disruption and out of course running.

I also saw the mailshot this morning re Demolition Simulator, I guess it helps RS.com revenue stream with commisions on sales but it's like most of those other farming/tractor type games which have now appeared - an interesting diversion for five minutes but not of much interest to the train simmer. Now if they partnered with TML...