Atomic Systems

From all I can gather from weeks of testing...

1) Any track scheduled to be used by AI traffic(regardless of Type...Special or any other type)... is blocked too the player until ALL scheduled AI traffic has finished using it.

Therefore, it follows that,

a) All AI traffic must use track completely separate from track used by the player, OR All player traffic must have a unique path that does not share the AI traffic track.

If this is the case...and from everything I've tried and read...IT IS!

Then AI Traffic is not an option. AI trains ARE an option...Traffic IS NOT.

This is unacceptable...
It is my sincere hope that RW is seriously working toward AI that follows the same rules as the player...and can (god forbid), tell a red signal from a green one. (I think Atari had this figured out a few year ago!)

What we DON'T need are more bells and whistles(no pun intended), on a scenario editor for trains THAT DON"T RUN!

Don't worry, things aren't as bad as you think they are! I have developed scenarios with about 60 AI trains and a player train weaving in and out of their paths with no problem so there must be something going wrong somewhere. I'm just not sure what.

It is entirely random...

You may have read a week or two ago that I could not get AI movements to interact at all when setting up the scenarios on my Tanigumi tramway. I'm not sure if anyone who has downloaded the route has had any better success - judging by the lack of any fresh activities in the file library it would seem not. I put this down to the single track sections, but then realised that bothmy Heartbeat Moor and Sulitjelma line are single track - where AI trains share the same line in the opposite direction to the player train - and I (mostly) got these working okay. Maybe the AI doesn't like routes with a mixture of double and single track...

The only outside thing I can think of is if the new timetable interface checks things in a different manner to the old system, but surely it's just a different front end on the core?

You can also find that the same scenario will initialise okay on one attempt to run it or not on another. When I uploaded the (since deleted) scenarios for Paddington to Oxford some people were getting AI collisions while others were not.

As the next project I've started is largely single track I'll be waiting with baited breath to see what happens if/when it reaches scenario stage. It gets kind of lonely and boring just driving your own train (or tram).

Otterbear2 wrote:From all I can gather from weeks of testing...

1) Any track scheduled to be used by AI traffic(regardless of Type...Special or any other type)... is blocked too the player until ALL scheduled AI traffic has finished using it.

never use the special setting, this will give the loco its full path and not just so far ahead

Gav

phat2003uk wrote:Don't worry, things aren't as bad as you think they are! I have developed scenarios with about 60 AI trains and a player train weaving in and out of their paths with no problem so there must be something going wrong somewhere. I'm just not sure what.

Once you find out what it is, please let us know!

As things stand at the time of this writing, I would share Otterbear's view. I was not even able to make AI trains cross their paths in a reasonable way. Every now and then, I read about someone's success, but generally, I am not able to reproduce it. I also think, like Vern, that there is a big amount of unpredictability in the whole AI system.

A very special thank-you to all who responded and tried to help me in this.

I have taken my medication and feel a little better now...

Ok,

Firstly, it is the unpredictability of the AI/Scenarios that cause most of the headaches...as you/I continue to TRY to get it to work. Not knowing if your going to succeed or not, and then NOT knowing if it will work as planned the Second time around.

Secondly, Creating a scenario that is 15min's to a couple hours long, with AI traffic, is possible (I think), but when your running a free roam scenario and saving it, and coming back to it for days. (like I do), it would be nice to have a train that is scheduled to go through the area, say once an hour or so, just to add some realism. This is what I had attempted to do - in many various ways. (Portals, Timetabled, looped track, ect).

I play in free roam, on a fairly short track, and I have set it up with plenty of consist to haul back and forth. So, its not really a scenario (IE: There are no timed activities, no special place to put stuff, of pick up ect.).
I just have a handful of trains I can pick from, or jump too, and run all day (or for several days, or indefinitely). So the ability to add ONE train (or more), that would cause me to wait at times would be a GREAT addition for me.

I tried portals, but even though the track is clear, if another train is scheduled to come OUT of that portal and use the track (at ANY time), the track remains closed to the player train.
The same problem arises when, ANY train is scheduled to run on the same track as the player...be it looped or scheduled.
Of course if I add times that certain tasks must be completed...then I can run AI ...but that takes the "FREE" out of Free-Roam...and just makes it a list of tasks to do...basically a scenario, which reduces replay value and becomes boring and only played rarely.

So, its looking (at the moment), like AI traffic is out of the question.
This means that I get tired, and bored, in rather short order, running around on "the table" with no other trains to interact with.

Like I said, Hopefully this is being worked on at RW and will be fixed soon.

Thanks again for all the feedback and support.

Dave

PS:From what I've learned ...Passing Sidings and/or Passing loops are useless...if I have single track running from point A to point B....and I have an AI train scheduled to come "down" that track...The WHOLE track is locked to the player, wether or not there are signals and passing areas. Is this true or am I missing something?

One thing that gets said often is that in freeroam, RSC do not make any promise for AI to function. So you should switch to standard scenarios if you want AI to work. However, it is perfectly simple to run some AI on a separate track which the player does not touch in freeroam scenarios, too. And I had problems with AI (or player) locking the path in standard scenarios (and without any Special train), too. So I am not so sure about the gain.

And because of the mentioned unpredictability, it is pretty difficult (PO-speak for f...rustratingly unattainable) to make progress in this quest. However, someone relayed a twitter about RSC playing with scriptable scenarios. So rescue could possibly be on our way. I for one will certainly way for that or any other form of rescue, already having too much grey hair and too little time for extensive experiments.

Thanks AndiS,

AndiS wrote:One thing that gets said often is that in freeroam, RSC do not make any promise for AI to function. So you should switch to standard scenarios if you want AI to work.

I hate the idea of predictable or programed running. (although it is fun at times).
I wonder if I could create a 12-16 hr scenario and set an AI train to run once an hour down the track that way?

AndiS wrote: However, it is perfectly simple to run some AI on a separate track which the player does not touch in freeroam scenarios, too. And I had problems with AI (or player) locking the path in standard scenarios (and without any Special train), too. So I am not so sure about the gain.

This does work...but then so does...watching a video of a passing train...once an hour. Not exactly interaction.
What this means for a route developer (such as myself), is, that I essentially have to create TWO or MORE separate tracks for my route, so that the occasional AI train can drive by. I think this is unacceptable for both of us.

AndiS wrote:...And because of the mentioned unpredictability, it is pretty difficult (PO-speak for f...rustratingly unattainable) to make progress in this quest. However, someone relayed a twitter about RSC playing with scriptable scenarios. So rescue could possibly be on our way. I for one will certainly way for that or any other form of rescue, already having too much grey hair and too little time for extensive experiments.

Ditto!

One often wonders how long this wait will be...
Too me, its like building a Formula one race car...when the rules are going to change every other month. Route development and scenarios, although easier at the moment, are still too buggy to invest massive amounts of time in production.
We run the danger of creating so many work-arounds that we tie the hands of the developers so that major enhancements (that we want), will destroy so many "hacked" routes that they (RW) have to forgo implementing them.
This, I believe, was a major contribution to the demise of MSTS.
Problems have already started to crop up with payware not able to work with new enhancements that RW have implemented.
I'm kinda glad that the User development side of things has not grown as fast as some might like..for this very reason.
RW (or any developer for that matter), Does not want to strip a paying customer of an asset/route that they have paid good money for...so they are forced to use "buggy" coding and such, That they KNOW they can fix, but would essentially render purchased products obsolete and unusable.

I hate the idea of predictable or programed running. (although it is fun at times).
I wonder if I could create a 12-16 hr scenario and set an AI train to run once an hour down the track that way?

Sorry to say the kind of thing that works the least. Jim (the snipper) went wild about it a year ago. Programming a dispatcher that makes up a plan once and then goes home was simply easier and they had some budget cut back then, or cost explosion or whatever. Not that I know what Kuju did back then, but everyone knows that the outcome was a "simplified solution".

One often wonders how long this wait will be...

1) Three months for the scriptable scenarios, whatever that will bring. Could be the door opener for a really clever dispatcher, if they get the foundation right. Could be a big plug for an even bigger hole, if fates are against us. Only time will tell.

2) A year or two for multiplayer to become a reality. By that time, they need a really good solution for an automatic dispatcher (i.e., run by the game, artificially intelligent or just plain functioning), one that handles all kind of unpredicted courses of actions. The prefabricated stuff will go to the bin then, because with multiple players, interaction is all and 3 or 4 players waiting for some AI train to pass, or otherwise looking at the clear road for half an hour just because the dispatcher planned it to happen in a certain order will not sell this new multiplayer feature which CEO Paul Jackson hailed in an interview.

Multi-player???

Well at least the writing is on the wall now.

We now know what demographic they are striving to satisfy. (lets just say if your over 13 your input is no longer wanted although greatly needed as Beta testers for your teenage kids.)

Multi-player...is a way of saying..."We want more money...NOT...more quality".

Maybe they can join forces with TOM CLANCY and churn out hundreds of games that aren't worth the money it cost for the cardboard box!



-One seriously disappointed rail-sim fan.

If you think support is bad now...wait till you have a problem with a single player aspect of a multi-player game.
Who do you think is going to get more development time...the 3 guys needing single player support, or the 3,000 kids whining about their "scores" not being right?

If this Multi-player stuff is true...its a truly sickening turn of events.

It may be getting slightly off topic now but I also really fail to see how multiplayer would add any value to a train driving sim. Now as a bit of a crusty sod I've never really seen the point of multiplayer in non-railway games but what is it really going to offer? One poor guy trying to fulfil the role of signaller or despatcher, getting irate messages from bored kids sat at red signals and trying to moderate the whole thing. Your interaction is effectively limited to a virtual "wave" as you pass a train going the other way (assuming the scenario loaded in the first place). No real scope for co-operative or competitive play unless we're reducing it to a racing style game. The inevitable kids ignoring red signals (pretty much what the AI does now, I guess) in a deliberate attempt to cause wrecks or crashes.

I've been around these type of sims in one form or another for 12 - 13 years now and the thought of running some sort of multiplayer activity has frankly never seemed relevant. Of course the same noises pop up on the Trainz forum but, unless I'm missing some big picture here, I really don't see the point.

I am in favour of a career mode or even a structured driver's "diagram" where you can even get off one train and walk to another in order to take it over. However that can all be accomplished quite nicely in SP mode.

On another point which came up, the issue of the continuing AI lacking leading to less enthusiasm for producing long routes, I must nod sagely to that one. While a number of people (like the Danish gang) have bucked the trend and released extensive lines, even without all the other factors at play I really can't see the point in constructing a long single track 100+ miles line when the signalling AI can't even match TTDLX/Bahn/Rail3D.

The track that I had worked for months to just be able to develop would have had at least 750 miles of MAIN LINE track. (not including yards and sidings). As it stand now...everything is at a dead stop until we figure out what RW is doing.
As of right now, the same enjoyment can be achieved on a 3 mile Loop.