Atomic Systems

Have read and tried several articles on how to time several services using the same sections of track and just cannot get any consistent results. I find this dispatcher very frustrating compared to the Trainz dispatcher which allows complex station and junction movements - it seems impossible in this game to route and time trains close together, even in the same direction. Can get maybe two or three trains timetabled OK then a fourth one takes the whole scenario apart. Have tried combinations of intermediary signals/waypoints to no avail - how far ahead does the dispatcher "look"? How much time does it allow before allowing another train over the same piece of track? Cannot fathom it out...…... Does anyone out there understand the dispatcher's logic?

Bob

Queuing AI trains generally works if they are all of the same train type. Train type translates to priority and if one with a higher priority starts after some others, the dispatcher will try to have zero delay on the higher priority train.

When the player train is added, the dispatcher gets pessimistic or hysteric about safety margins. And if he does accept the player train, and you are running a bit early or late, AI will start to crawl at 1 mph or so and never more come to life, and ignore signals in this state.

So the best practice is to have AI and player on separate paths. You can have AI precede the player train with some headway if you keep things simple as far as this path is concerned. I would assume the preceding AI should better have higher priority than the player, but I am not sure.

When you meet situations where you think that it should work but a group of equal priority AI trains will not run as expected, then it is time to explore the signalling. AI follows these rules:

• Stop signals, i.e., signals that have the isStopping flag in their .bin file set to eTrue, delimit blocks and only one train may enter any block at any time. (If two AI trains start the scenario in the same block, they may crash if the second one runs faster.)
• Junctions span from the unnumbered link 0 of any signal to the farthest numbered link. AI trains will crash when on colliding paths and at least one of the paths is not signalled that way. The junction signals need not be of the stopping type.
• AI always runs to the far end of any marker. Beware of signals at that end which sometimes have their link 0 inside the marker. The train then enters the next block before stopping, and maybe occupies the junction there.
• When you want a train to reverse over points, better use the stop-at command. They way-flag seems to be intended for through running only, or so is my impression. At least I can remember cases where AI would have difficulty breaking at the right spot with it while swapping it for a stop-at marker did the trick.

Thanks for that AndiS - I have really been tearing my hair out with this. It is going to be too complicated for me to set up a realistic operating timetable and as an ex trainzer I am obviously expecting too much. Pity the flexibility of the trainz dispatcher cannot be combined with the superior scenery and overall feel of TS. Even if TS is intended as a simulator rather than a game, surely now we are on 64 bit, the dispatcher function could be improved/simplified, even if it meant a toggle switch between true simulation and gaming mode.

Now that I've bought most of the UK routes from DTG and JT I've got bored with just driving and want to do more - very disappointing.

Bob

From discussions on Matt's stream and comments made by Dovetail on their own streams I would not hold out much hope for major improvements to the dispatcher. That's not because they don't want to or don't have the skills to.

The issue comes if they make major revisions then it breaks tons of 3rd party products and probably most of the Steam Workshop scenarios. I doubt Dovetail would be particularly popular if they broke the entire Just Trains catalogue (and that is just 1 3rd party provider).

I don't work for Dovetail and have no inside information on this - only information from the streams I moderate and others I sometimes watch.

Would be better if you could select the current dispatcher or an updated one in the editor or by some other means, with the current one as default. It would cure the backwards compatibility, but of course even if that were possible, it might not be commercially viable.

Thanks for the further comments and I understand the need to maintain compatability with existing stuff - I've been looking for the streams by Matt and by DTG in case there's something there that I've not thought about - can you point me to them?

There must be many besides me who feel that there needs to be something along the lines mentioned by Holzroller to keep the interest (and additional spending) going.

Thanks again
Bob

Matt : https://www.twitch.tv/mattpeddlesden

Dovetail : https://www.twitch.tv/dovetailgames

Both have previous streams available to play on demand.

Surely it cannot be difficult to have two Dispatchers? Scenarios written with the new one merely have one line somewhere in the scenario files which in essence says "use the new Dispatcher". Without that line scenarios default to using the old one. Simples.
Robin

Thanks for the links Easilyconfused - will have a browse.

Just to clarify, I am not talking about doing anything with DTG/JT/Workshop scenarios, but the ability to model my own timetabled workings from a cloned copy of a route that would not affect anyone else.

Bob

holzroller wrote:Would be better if you could select the current dispatcher or an updated one in the editor or by some other means, with the current one as default. It would cure the backwards compatibility, but of course even if that were possible, it might not be commercially viable.

They already did this with the audio control - the first entry in the blueprint is a choice for Proxy Control Revision between "Version 2" and "Legacy Compatibility". They could do it if they wanted to.

bjs1943 wrote:Thanks for the links Easilyconfused - will have a browse.

Just to clarify, I am not talking about doing anything with DTG/JT/Workshop scenarios, but the ability to model my own timetabled workings from a cloned copy of a route that would not affect anyone else.

Bob

Could you clarify the problem you seem to have. It is possible to create timetabled scenarios where all the AI do what is required, interacting with the player train. They must follow the dictum that only one train can be in a signal block at a time (apart from within station limits where a limited set of movements can be possible, under call-on arrangements). It is possible to run a train of equal or higher priority in front of the player train to "block" its passage (ie slow it down, "chasing yellows"). You can have a train in front of lower priority that has to pull into a loop to allow your player train to pass. You can have a train of higher priority following the player train, which has to pull into a loop to allow this second train to overtake. I regularly schedule AI to start by "triggering by distance" rather than by time. I have AI in the opposite direction moving between portals, so that paths do not overlap. I have tried spawning AI from a portal, but find it a bit "hit and miss".

Nick

PS you do not need to clone the route to write your own scenarios!

Hi Nick - my bad terminology with the clone - I meant create my own new scenario. I have tried new timetable and standard scenarios to no lasting avail - even tittled around creating a quick drive.

What I would like to do is timetable and route trains as in real life, i.e. sometimes there will only be 3 minutes between following trains or oncoming trains will cross each other's pathing etc. to make a realistic scenario with numerous AI trains and a player train from which the driver can jump to another train as in free roam. Also I do not want the player train to automatically take priority over an AI train if it is timetabled to follow it. (the prioritising function doesn't always work).

I would like to see a dispatcher that simply follows signals, but everything else is in the control of the modeller - if trains are on a conflicting path then they crash etc. - the modeller is in control, not the dispatcher telling you that you can't do it. (and not always explaining why)

The current dispatcher is far too restrictive - a more resilient one would make a realistic game play and the purist simulation guys could simply choose not to use it.

I've come over to TS from Trainz where I had 70+ trains operating ECML and XC London-Edinburgh to a full day's timetable, but Trainz in their wisdom introduced a timeout restriction with their "New Era". The scenery and general feel of TS is obviously better than Trainz but the modelling ability is very much inferior and I am really disappointed now.

Yes, you can do what you say - SOMETIMES. I have successfully chased yellows, but then find another time the dispatcher will not allow me to follow within 30 minutes. Sometimes can set up several timetabled trains then add another, non-conflicting train and the whole thing develops into a mess trying to remove conflicts.

In short, I want control and not the dispatcher telling me what I can and can't do!

Bob

brysonman46 wrote:

bjs1943 wrote:Thanks for the links Easilyconfused - will have a browse.

Just to clarify, I am not talking about doing anything with DTG/JT/Workshop scenarios, but the ability to model my own timetabled workings from a cloned copy of a route that would not affect anyone else.

Bob

Could you clarify the problem you seem to have. It is possible to create timetabled scenarios where all the AI do what is required, interacting with the player train. They must follow the dictum that only one train can be in a signal block at a time (apart from within station limits where a limited set of movements can be possible, under call-on arrangements). It is possible to run a train of equal or higher priority in front of the player train to "block" its passage (ie slow it down, "chasing yellows"). You can have a train in front of lower priority that has to pull into a loop to allow your player train to pass. You can have a train of higher priority following the player train, which has to pull into a loop to allow this second train to overtake. I regularly schedule AI to start by "triggering by distance" rather than by time. I have AI in the opposite direction moving between portals, so that paths do not overlap. I have tried spawning AI from a portal, but find it a bit "hit and miss".

Nick

PS you do not need to clone the route to write your own scenarios!

I agree with all that, Nick - the critical factors appear to be selecting the priority level of the AI, and making sure that they run from when they start in a different signal block to trains in front and behind. And as for starting from portals, there lies the ultimate in frustration.

My latest trainspotting scenario, at a four track main line station, with another two tracks main line merging from one direction, allowed me to queue AI 'off scene' - I then set start times so that the AI do not occupy the same signal block once they are moving. Trains happily move from the fast to the slow platform lines in the station and are overtaken by another train on the fast line - they can then move back to the fast line, or cross the other three lines to reach the branch, when leaving the station, without interrupting later trains. Trains merge in from the branch, crossing the paths of later trains on the main line. I even have a local shunter crossing all four main lines, and the traffic on those lines resumes once it has crossed.

If you are asking, the location is Hitchin on Cambridge Branch Lines and the ECML. I'd started this before the 64 bit version of TS was released - and could only get 10 trains to run reliably. Now, with 64 bit, I've managed to fit 25 prototypical trains into an hour of action, before the scenario becomes unstable. I built the scenario in stages, adding a few trains each time, and made sure I backed up the scenario files each time into a separate sub-folder, so I could always step back if I had to - and I had to. There are 19 versions of the current scenario, each of which I have run several times, at up to an hour each time - you need plenty of free time for this hobby!

made a lengthy response to Nick's post above which seems to have vanished ???

Bob

Not ignoring your post Nick but I sent a lengthy reply which should have appeared here before Alan's post.

Have since sent an enquiry where it is and that not appeared either.

Have to go out shortly for rest of day so hopefully my reply will appear in the meantime.

Bob