4.363 Are there issues with the AI scripting logic?

[Gibsonm]

I'm unsure if its a 4.357 or 4.363 thing, but the AI scripting seems more random than it used to be.

 

The things I'm having issues with are I would have thought very straight forward.

 

e.g. Unit A embarks if Unit X has reached waypoint Z.

 

What I'm finding is that in the Mission Editor running a test at normal speed:

 

Test run 1: This works. I go off to "fix" something else.

 

Test run 2: This doesn't work work. I then spend time trying to work out why it doesn't.

 

Last  night this worked.

 

This morning it doesn't.

 

Happy to PM AAR files (last night and this morning if required).

 

I've resorted to removing any of the time issues ("Before checking  wait" etc.) in an attempt to simplify things and am now moving units on the map so the delay is created by them covering extra ground as opposed to waiting for the timer.

 

I have spent several frustrating weeks, with several iterations per day, trying to sort this out.

 

It is especially frustrating as you need  to run the entire scenario the entire length every time (because what worked 10 mins ago may no longer work), not just check whatever changes you make.

 

I can't release these items if sometimes the scripting works and sometimes it doesn't.

 

Edited September 19, 2022 by Gibsonm

[Ssnake]

At the moment I don't have an answer, but needless to say, this doesn't sound good, if true.

Are you certain that the units don't actually embark, but then fail to actually drive forward?

(The battle position symbol at their location would still be there if the decision to embark wasn't made, but gone if it was en route but not making progress traveling). 

[chrisreb]

I would echo what @Gibsonm says. Units act differently with same triggers which seems illogical for computer controlled units.  We can both run exactly the same scenario and get very different behaviour.  Next time I will have a particular look at the battle position locations.  Be interested to see what others are experiencing.

 

In addition , probably since last couple of updates spawning is not always consistent.

 

I have units due to spawn given a condition OR after mission time exceeds X. On occasion they simply do not spawn. I can send an AAR when I next encounter it.

[Gibsonm]

6 hours ago, Ssnake said:

At the moment I don't have an answer, but needless to say, this doesn't sound good, if true.

Are you certain that the units don't actually embark, but then fail to actually drive forward?

(The battle position symbol at their location would still be there if the decision to embark wasn't made, but gone if it was en route but not making progress traveling). 

 

In this specific example there is no battle position, its just a "naked" waypoint with an embark if  ... conditional route.

 

Waypoint "205" is marked (poorly) with the fluro green pen.

 

What is supposed to happen:

 

3C comes through waypoint 156 (aqua), moves through 480  (where 2C's route connects) and then proceeds through 205 (green fluro) and beyond.

 

 

Last night's run:

 

 

3C hits 156, then 480 and 205 (triggering  2C to follow) and beyond.

 

Same file (some edits elsewhere) tested this morning:

 

 

3C does what is supposed to do.

 

2C stalls at 507.

 

Rest of C Coy (that moves sequentially) goes nowhere because 2C stopped at the waypoint with the "embark if 3C has reached 205" command.

 

Edited September 20, 2022 by Gibsonm

[Volcano]

The only way this could be tested would be if you provide a stripped down test scenario here so we can look at the logic. That is, have a unit with just the route properties in question present, where the unit can be observed.

 

There are always cases where the route type and overall situation may affect how a unit behaves, so if enemy is present or other units are getting in the way (traffic jam) them try to include those as well.

 

Despite all that, I'd say that the logic itself has NOT changed. The addition of the colored tactics (black, yellow or flashing etc) is just a visualization of what was already done. However, there could be a bug affected things with pathfinding, or some other such issue. A test scenario would help figure that out.

[Gibsonm]

OK and if I strip the other 500+ waypoints, routes and units away and then it works in isolation then what? There is no guarantee that this micro version displays the behaviour or that if the "fix" will work back in the full sized scenario.

 

I understand it saves you the 30 mins of playing time to get to that point, but this is a small portion of a whole.

 

This seems to be just a bucket of extra work for me to hopefully fix one small issue, when maybe its caused by unit density or some other factor that a bare bones test bed doesn't represent.

 

Not to mention the few other places where similar things happen / don't happen on a seemingly random basis.

[Ssnake]

If it never happens with a stripped-down version of the scenario, that would be a clue in itself.

We don't know when, where, or why it happens. You know THAT, and where it happens. Lead us to the case, and we'll investigate.

[munckmb]

Can you share the scenario? If you want I can do some experimenting.

[Gibsonm]

16 hours ago, munckmb said:

Can you share the scenario? If you want I can do some experimenting.

 

No, I'm afraid not.

[Gibsonm]

18 hours ago, Ssnake said:

If it never happens with a stripped-down version of the scenario, that would be a clue in itself.

We don't know when, where, or why it happens. You know THAT, and where it happens. Lead us to the case, and we'll investigate.

 

I'll post it and links to the earlier AARs down in the relevant Forum section.

[Volcano]

On 9/20/2022 at 4:40 PM, Gibsonm said:

OK and if I strip the other 500+ waypoints, routes and units away and then it works in isolation then what? There is no guarantee that this micro version displays the behaviour or that if the "fix" will work back in the full sized scenario.

 

I understand it saves you the 30 mins of playing time to get to that point, but this is a small portion of a whole.

 

This seems to be just a bucket of extra work for me to hopefully fix one small issue, when maybe its caused by unit density or some other factor that a bare bones test bed doesn't represent.

 

Not to mention the few other places where similar things happen / don't happen on a seemingly random basis.

 

AARs will not help, really. The fact is: if there is something wrong with the logic, then it would be apparent in a simple stripped down scenario. I was saying to reproduce the situation shown in a series of simple routes and see what happens.

 

If it works there, then it should work elsewhere. If it doesn't work with 500+ routes and waypoints, then it implies that its a problem with the relationship between the routes and waypoints in the area. Otherwise what is expected here, that we go in with a hammer and bang on the code based on assumptions? 

 

So, I am simply stating what would be needed to give it a serious investigation, that is all.  

[Gibsonm]

47 minutes ago, Volcano said:

 

AARs will not help, really. The fact is: if there is something wrong with the logic, then it would be apparent in a simple stripped down scenario. I was saying to reproduce the situation shown in a series of simple routes and see what happens.

 

If it works there, then it should work elsewhere. If it doesn't work with 500+ routes and waypoints, then it implies that its a problem with the relationship between the routes and waypoints in the area. Otherwise what is expected here, that we go in with a hammer and bang on the code based on assumptions? 

 

So, I am simply stating what would be needed to give it a serious investigation, that is all.  

 

Yep and that's why I posted a copy down in the relevant forum section complete with @Ssnake and @Volcano to highlight it for you both.

[Volcano]

I am not tracking. Out of necessity, I sort of swoop in and read bits and pieces of posts and I often reply without the full context because of it. If there is something somewhere that I am supposed to look at, then PM me the details please.

[Gibsonm]

19 hours ago, Volcano said:

I am not tracking. Out of necessity, I sort of swoop in and read bits and pieces of posts and I often reply without the full context because of it. If there is something somewhere that I am supposed to look at, then PM me the details please.

 

PM sent.

 

Sorry not to deliberately flag it earlier.

[Gibsonm]

 After about 40 runs of "test and adjust" (40 x 90mins [in real time] + modifications) it now seems happy.

 

Relevant details extracted from the other thread:

 

19 hours ago, Gibsonm said:

Another 6 runs and now the three Company sized units move as required.

 

Wherever possible I've abandoned this choice:

 

 

As it appears problematic / unreliable.

 

Unsure if its worth looking into?