Experimenting with Task Editor Generated Missions

[Captain_Trips]

I was playing around the last couple of days and thought I'd share.
Has anyone played with doing this?

[video removed , better workflow below...]

Edited September 3 by Captain_Trips

[KillerBee]

Used to do this a lot with Vander’s EMG.  Create a mission, then mod/update into something more realistic.  Worked great.

[Captain_Trips]

Next step is I'm trying to teach AI to cull it further.
I'd like to create a library of clean templates of mission types with all the Fast Travel and escorts and scoring logic setup that I can the customize for variation.

20mb he just laughed at me.  Getting trimmed down I have a chance.

 

[Captain_Trips]

I think I figured out the flag that hides generated missions from the Single Mission list.

Open the mission file in text editor and change 
MissionType = 1203;  (or what ever number it is...)

to

MissionType = 0;  (I assume this denotes user made.)

This makes the mission show up in Single Mission list.  Therefore you don't need to export to a group file and import back in as I did in the video.

***IMPORTANT***  For this to work, delete the .msnbin so it gets regenerated or else the flag is still stuck in there.

That saves a time consuming step.

I had suspected that before but couldn't get it to work so I thought it must be something else.  But I hadn't deleted the .msnbin so I was still getting a stale flag.

[Edit]

So further testing the name change is also necessary (understandable) but is not sufficient alone.
So my current test workflow looks like:

1.  Generate a desired mission in Task Editor and start at least up the the Briefing page to generate the files on disk. (Generates _gen.mission, et al)
2.  Use Notepad to edit MissionType = 0;  (In the Options block near the top.)
3. Open "_gen.mission" in mission editor and make whatever edits you wish.  Maybe add something to the description to recognize.
4. Save-as a different mission name like "Test.mission".  (That generates a new .msnbin file.)

The new named mission generated from Task Editor now shows in the Single Mission list.


 

Edited September 3 by Captain_Trips

[Jaegermeister]

2 hours ago, Captain_Trips said:

I think I figured out the flag that hides generated missions from the Single Mission list.

Open the mission file in text editor and change 
MissionType = 1203;  (or what ever number it is...)

to

MissionType = 0;  (I assume this denotes user made.)

This makes the mission show up in Single Mission list.  Therefore you don't need to export to a group file and import back in as I did in the video.

***IMPORTANT***  For this to work, delete the .msnbin so it gets regenerated or else the flag is still stuck in there.

The new named mission generated from Task Editor now shows in the Single Mission list.


 

This is correct, it has to be saved as “single mission” to show up in the missions file and play, and have a new .msnbin file generated.

I have generated many missions to break out certain parts of the logic from the template and save them as group files, but be aware that the mission logic the devs use for the templates is not very efficient for single missions because it is designed to accommodate many possible scenarios instead of just one. Also there is no sub grouping, so if you are planning on editing the missions, it is challenging to sort out what is what.

If you are planning on using the missions as they are, you can just edit the text files and go from there.

[Captain_Trips]

13 minutes ago, Jaegermeister said:

This is correct, it has to be saved as “single mission” to show up in the missions file and play, and have a new .msnbin file generated.

I have generated many missions to break out certain parts of the logic from the template and save them as group files, but be aware that the mission logic the devs use for the templates is not very efficient for single missions because it is designed to accommodate many possible scenarios instead of just one. Also there is no sub grouping, so if you are planning on editing the missions, it is challenging to sort out what is what.

If you are planning on using the missions as they are, you can just edit the text files and go from there.

Yeah, just for farts I though I'd used them as is with some minimal manual culling.
I might make a quick prototype campaign just to demonstrate and get an easy one under my belt.

Next step is to work with AI to get him to understand the dependency graph of just what is needed for the minimal mission and cull everything else.  I was getting close.  That will be step two.  🤪

Step 3 is further customization of the generate scenario.

Eating an elephant.  Cut it into small squares and never stop chewing. 🙃

AS an old programmer I can't help myself.  Figuring stuff out is almost as fun as flying.

Edited September 3 by Captain_Trips

[Jaegermeister]

1 minute ago, Captain_Trips said:

Yeah, just for farts I though I'd used them as is with some minimal culling.
I might make a quick prototype campaign just to demonstrate and get an easy one under my belt.

Next step is to work with AI to get him to understand the dependency graph of just what is needed for the minimal mission and cull everything else.  I was getting close.  That will be step two.  🤪

Step 3 is further custimization.

 

Sounds like an interesting plan. There have been quite a few comments on the forum that the AI behavior in template generated missions such as career mode is less than ideal, so I don’t know if that is something you have considered.

It is easy enough to manage if you make the missions from scratch, but if it is built in behavior, you would have to either isolate the AI attack commands and modify them, or add in overrides that will do things like make the enemy AI drop tanks at engagement, break off and go home when attacks end, get wingmen to stay on task, etc.

It can be done, but I am not sure exactly how in your case.

[Captain_Trips]

Well, I'm not doing anything serious, just playing around.  
I might just post the link here in this thread for testing rather than sharing it officially.

 

Edited September 3 by Captain_Trips

[Captain_Trips]

OK.  So I've been doing some experimenting.

As you warned @Jaegermeister, the Task Editor generated mission are a mixed bag and far as a starting point.

Here is what I am finding from my experiments:

1.  20mb mission files is just filled with objects all over the map I will not not need. I suspect they are doing some kind of fancy content streaming so I hope that whole thing isn’t being loaded at mission start, but I don’t know that. In any case, it offends my OCD so I wanted to try to cull a lot of that without breaking things.

2. As Jaegermeister warned, the generated success conditions often seemed overly complex or excessively tight. Sometimes they seemed over scaled for the number of planes I chose. So I wanted to review the success criteria and often greatly simplifying.

My general philosophy is, I am a big fan of simple. Get to the target area, kill a couple of things, and that is mission success. There will be more things to kill if you want, but of course each pass is additional risk. But kill as much as you please. Bon Appetite.

3. I am not an expert mission builder, but I would add a few touches here or there manually.

4. I am using Jaegermeister's awesome campaign template as a structure to stitch the generated missions together into a storyline. Not historical. Not meticulously hand crafted, but playable.

So I am using AI to assist in this experiment.

I created a custom GPT Project, and uploaded all the available IL2 Mission editing manuals into his “Knowledge” files, so he is reasonably good at editing the mission files.

If anyone wanted to try out what I have so far (NOTE: WIP):

Campaign (WIP): Lead Sleds Campaign
Source missions: Mission source

 

It’s only 5 mission so far. The time wasn’t spent on the missions. The time so far was in nailing down a repeatable process with the AI. I probably re-created the first mission 30 times trying to work out the proper approach. Now I think I have a fairly repeatable procedure.

Here is the current procedure I am using with the AI. If you were making a custom GPT project, you could drop this into the Knowledge files along with the Mission Editor manuals so he doesn’t have to re-invent the wheel.
IL2_Korea_Template_Mission_Procedure

I also and having him translate my English description into the other localization files. I only speak English, so I hope those are reasonable. Best I could do. Here is his procedure for that:
IL2_Korea_Description_Translation_Procedure

The basic idea is I upload the large raw mission file and AI first analyzes the success conditions and gives me an executive summary of how it’s setup. We then discuss the options and I usually have his simplify it based on our discussion.

At that point he updates the success logic and then begins culling.

I was to scared to just start deleting stuff by hand, so he analyzes the mission as a series of dependency graphs to determine what was safe to delete. The guidline was to keep any object required by the success logic and any objects within a certain distance of the waypoints. A 20mb mission file usually ends up about 4mb and the msnbin about 1.5mb. Much better.


Of course, if you’ve done work with AI it can often be an exercise in frustration. I have found the best results in using this prompt. It lets him review the procedure first and THEN I upload the file. I explicitly force a Go-NoGo decision before he just runs off in the weeds. Like:
 

Quote

“I am going to provide you generated mission. Convert this generated mission into a template using our current procedure. Follow all approval gates: stop after reviewing the success conditions and wait for my approval before doing any culling. After approval, complete the conversion and verify every required step against the stored procedure before giving me the finished file. Are you ready?”

That is how I’ve gotten the smoothest round trip. YMMV

I was planning around 10 mission eventually.  Please remember all this is still just tech demo prototyping WIP state. I’m just experimenting for the fun of it, and I understand these won’t be near the quality of a hand-crafted, historically researched campaign.

Constructive feedback is welcome.

Edited September 9 by Captain_Trips