Crawling AI

[jp4712]

I wonder, could you friends try something for me please? What I like to do in RW is to create scenarios. This is going okay for me, but one problem I'm having is the syndrome whereby an AI service crawls at 0.1mph or is even welded to the track. I read somewhere that this may be caused by another service 'grabbing' the track; or possibly by the despatcher getting so busy that he 'forgets' a train; or, I have observed that it can be triggered by an AI train being too long. The latter convinces me as being at least one explanation; I put together a scenario on the default Hedborough route, this issue occurred, but when I removed some trucks from the train the problem went away.

I think it would be helpful to me (and to RS.com I suppose!) to really get to the bottom of this. So, colleagues, if you get a crawling AI on a scenario, could you maybe try shortening the train and seeing if that really is the issue? If you could perhaps post here what impact this had, if any, it might help resolve this little glitch.

Thanks

Paul

[RudolfJan]

I just submitted a bug report to RC.com. Too long consist cannot be the only cause. I have a crawling consist problem, which occurred just after changing the start time of the AI.

Regards

Rudolf

http://www.hollandhiking.nl

[thecanadianrail]

i only get this problem when i put more than 20 freight cars on the track. after 20 every car slows the train down by 1-2 MPH untill you hit 5 MPH.
mabey someone should create a update so the AI drives at line speed nomater the train length or change the number of cars for the maximum train length to at least 50 cars. i guess RS.com really dosen't care about these american routes because in a eauropean route almost every train is about 20 cars long. no long-hauls except for north america and australia.

[OldProf]

i only get this problem when i put more than 20 freight cars on the track. after 20 every car slows the train down by 1-2 MPH untill you hit 5 MPH.
mabey someone should create a update so the AI drives at line speed nomater the train length or change the number of cars for the maximum train length to at least 50 cars. i guess RS.com really dosen't care about these american routes because in a eauropean route almost every train is about 20 cars long. no long-hauls except for north america and australia.

I just don't buy this theory that long AI trains cause creeping. Why? Simply because I have played scenarios, particularly on the Barstow-SanB route, that include v-e-r-y long AI trains which do anything but creep. In my scenario creation experience, creeping is more often the result of two or more trains, often including the player engine, interfering with each other's paths. And this is certainly not a case of the so-called dispatcher "forgetting" a train: it's a computer, not one of us mere humans, after all. When creeping occurs, I try adjusting departure times, inserting specific pathing instructions--sometimes using scenario-specific track markers, and juggling the classification of the trains in question: not all at the same time of course, since this has to be dealt with patiently and thoughtfully. As I work, I usually create numerous clones of the scenario I'm writing and I have at times discovered that a version that refuses to work due to stuttering on one particular day will work at a later time. This can easily become confusing, so I try to make copious notes as I work.

Let me provide an example. I'm currently developing a scenario for the RS Falmouth route that combines shunting in the yard and dock areas at Falmouth and features a good deal of AI traffic. Yesterday, I tried changing a couple of the player train's instructions and an AI passenger train that had previously run well began creeping. After about an hour of frustration, I realized that the amount of time the player train took to accomplish a series of interconnected tasks had changed, thus changing the time at which the player train and this AI passenger approached each other. It took several adjustments to the AI passenger's departure time, but the near meet finally ran smoothly again. By the way, interference between two trains often relates not to the actual time at which events occur as someone plays the scenario, but the times predicted for those actions that are visible in the timetable scenario editor. Although potentially frustrating for the scenario writer, this makes good programing sense because the program cannot predict the actual pace of play by a player, so it is set to predict times. Have you ever watched AI trains move, especially if they engage in drop-offs? They dash into place at the track speed limit, come to an abrupt halt, uncouple at the speed of light, and dash away again at the speed limit: I defy any human player to duplicate this behavior!

Tom Pallen

[micaelcorleone]

Some promising news from RS.com's Facebookpage:

Comment by James Penca:
Tell me if I am wrong. When I see trains go by here in the US they can have 100+ cars on it. In RW2 I feel like cars are too heavy or something because the longest consist I have pulled is around 60. I know I could pull more but I feel like i couldnt go too much more. And most scenarios have you taking 20 cars from A to B. 20? how about 120?

Answer by Paul Jackson, CEO of RS.com:
Hi Guys, we are doing a ton of work on the Physics engine at the moment AND working on some cool US scenarios. Will take some time of course, but I am thrilled by what I am seeing.

[SithMurcielago]

That is some encouraging News.

[thecanadianrail]

yes, very encouraging! well thats them, alwayse making things better!

[Terry]

I have created scenarios that involve very long trains (100 loaded coal hoppers or so) and not had a problem. I've created some pretty long AIs as well - maybe not 100 cars but some are 50 or more, I'm sure.

It can take a while to get a big player train moving from a stop but give it enough power on the front end and increase the throttle slowly until you see the cars all moving forward. As the last car begins moving you may want to slowly increase the throttle. I think I used three ES44AC units to pull the 100 cars but I might have been able to get away with two.

Glad to see the continous work on the part of the RS.com guys - very encouraging indeed! :)

[captkilljoy]

Well from what I can tell is that if you have more than one engine only the one that you are in uses the Simulation file, I figured this out after I got the SW1500 pack. The SW1500's smoke effects are set up to work off the Tractive Effort, if you have more than one yes they will all smoke but not like the one you are sitting in. So if you have a 100 car train and 4 engines only one of them is really doing all the work. So till RS or someone with scripting skills can make it that all the engines use the Simulation file there is not a whole lot that can be done and if I am wrong sorry to waste your time reading this.

P.S. this is also stated in the SW1500 engine script,


-- If this is a player train and this vehicle is the driven one, then send
-- RPM values in both directions along train. Note the Driven control is set
-- by the simulation script, which only runs on the vehicle the player is driving.
-- Only send messages if values have changed.

[arizonachris]

Hi, Cap'n! Welcome to the Forum. I think there may be 3 different issues being talked about here. First off was an issue with AI trains, when creating a custom scenario, not behaving correctly. (and look at the date of this Topic: June of 2009) Then someone mentioned the game's physics, which has always been an issue. Can one SD40 pull 100 cars? So that's the second issue.

Now, the smoke you are talking about, I know the topic has come up about the effects (sound and smoke) of multiple unit consists not being correct, and I know someone made an add on pack to fix it, but for the life of me I can't find it. I don't drive "in cab" too much, unless it's a British or European route, I use the "2" key and "fly" around, and since day one I have noticed that the head loco is the only one that looks and sounds like it's under power. All the other loco's are at idle. Now, whether that is truly affecting the physics of tractive effort, I don't know. So I guess that's the third issue. Could be a nice experiment, if someone had the time to do it.

Me, I'm just gonna have fun with Railworks as it is, and wait for the next Big Update.

[micaelcorleone]

I want to clarify this:

-my post was mainly related to the physics
-which has also something to do with the crawling AI issue


What I've expreienced is that all locos in a consist supply power to the train, but some locos only have idling sounds and constant smoke.
There have been some major improvements in the last half year. The new SD70 engines throttle all up in sound, at least the ones behind the lead engine (my impression, can't tell if this is true).
Also the smoke was improved with the new engines.

The crawling AI is the bug that if you create an AI train and go over 55 cars or so it will get slower for every new car you add regardless how many engines you add to the train. If a certain number of cars is reached the train will alsways do a maximum of 5 mph or so. If you drive the train yourself, everything is fine, but the AI can't.