Military AI Works

I'm new so I don't know if this has been done to death, but I've managed to get the FSX/P3D RAF Typhoons to show up - and move - at MSFS Lossiemouth:



I'd be delighted to share the method/process with anyone here who is as interested in this as I am!

Hi tiger1962 and welcome to the forum.

Well done for getting the result you have. It is not an easy exercise. We have long been experimenting with converting FSX native models for use in MSFS as you will see from the various threads on the forum. Indeed Pete's recent posts using the latest techniques, particularly in PBR texturing, are just fantastic:

https://militaryaiworks.com/forums/topi ... 4&start=20

The downside, at the moment, is that since Asobo broke the AI engine, the aircraft do not perform well to flight plans. It is, though, possible to overcome that by using the AIG AI front end tool. We are all waiting to see how the AI engine behaves and what we can do with it, when Flight Simulator 2024 is released on the 19th November.

John

We put the brakes on this due to the inconsistances of the AI engine in MSFS.
However, it may be that you hit on something that overcomes those.

So it would be interesting to see your process in case there is something that we can learn from.

Thanks for the reply Steve, I'm at work this morning but I'll be in touch this afternoon.

OK here goes:
I'm using MSFS Premium Deluxe on Windows 11 Home and the tools I use for AI Traffic creation are ADE2020_Alpha_21 and AIFP34315, I also do a bit of manual aircraft.cfg editing with Notepad.
I use Airport Design Editor (ADE) for AFCAD editing because all of the default MSFS airports, airfields and airstrips have been created with a program or app which made all parking spaces as ramps, with a parking radius of 5m or 7m for GA Small up to 23m for GA Medium. It also created all taxiways as apron links and placed the nodes far too close together on curves, so there are broken taxi links everywhere and AI traffic simply stops or times out and disappears.
I use Don Grovestine's AIFP34315 because it's MSFS compatible and has plenty of built-in checks to warn you of any faults before you try to create a traffic.bgl. The only fault it found with the Typhoon package was that the aircraft have no parking radius defined, but it compiled the .bgl's perfectly well, after I'd made a few Notepad edits to the aircraft.cfg file:
MSFS needs some specific details to be present in the .cfg files for the AI engine to control the traffic correctly, the edits I made are as follows:

Each [fltsim.xx] section needs the two lines below to be added:

isAirTraffic = 1 ; airtraffic flag for variations
isUserSelectable = 0 ; flag off for non selectable planes

I also changed the parking type from MIL_COMBAT for testing purposes:

atc_parking_types=RAMP

The [general] section needs to be in default MSFS format like this:

[General]
atc_type = "$$:Eurofighter" <-- the $$: prefix forces the ATC engine to say Eurofighter
atc_model = "$$:Typhoon" <-- the $$: prefix forces the ATC engine to say Typhoon
Category = "airplane"
performance = ""
editable = 0
wip_indicator = 1
icao_type_designator = "EUFI" <-- all icao info MUST be correct for multiplayer model-matching
icao_manufacturer = "Eurofighter"
icao_model = "Typhoon"
icao_engine_type = "Jet"
icao_engine_count = 2
icao_WTC = "M"
icao_generic = 0 <-- icao generic value MUST be 0 for all AI traffic

I get the ICAO info from their website: https://www.icao.int/publications/DOC86 ... earch.aspx

That's sufficient to get the models to show up in MSFS at least, but they would fly much better if the .cfg files were separated out in default MSFS format:

ai.cfg <-- contains important autopilot control PID's etc
aircraft.cfg
engines.cfg
flight_model.cfg
systems.cfg
target_performance.cfg <-- contains important reference climb, cruise, descent and approach speeds etc. for the AI engine

The only other thing is the MSFS folder and file format, along with the required layout.json and manifest.json files - without these nothing will show up in MSFS.
I use the freeware layout generator which is drag and drop, very quick and easy to use.
The typhoon package I created is in a main folder titled MAIW Typhoon.
The traffic bgl's are in folders named scenery/world/traffic.
The aircraft folders and files are in folders named SimObjects/Airplanes/Typhoon.

Sorry if you already knew all this, I'll be more than happy to give you any other info and help that I can.

It is great to see in MSFS, again one of the reasons I continue to use P3D, my MAIW set up. Following the progress over the years and the challenges even with the AIG team in MSFS. Had hoped for more in the coming 2024 version, looks like AI again is way down the list.

Thanks for the explanation, which is mostly familiar. 4 questions though:

- Which Typhoon model are you using and have you done any conversion to gltf format with MCX?

- How do the lights and afterburners perform?

- Do those changes to the aircraft.cfg file enable the aircraft to stick to the flight plans, ie, the intended aircraft departs each time, on schedule and reaches the destination, or returns to base, on time?

- Have you also managed PBR texturing? We found that without it, the aircraft appear too dark and gloomy in MSFS.

John

I'm using a typhoon package that was included in the msfs lossiemouth scenery by nickb007 on flightsim.to. I've seen the lights working but not the afterburners yet. The msfs format .cfg files do improve the flight characteristics, two typhoons did go-arounds this afternoon while I was still on the runway. Perhaps I could send you the package for further testing?

Aha, then that's my Typhoon that Nick converted to a static object in MSFS format, with my permission.

Thanks for the offer of the package, but I don't really want to test further in MSFS now. I'm waiting to see what FS2024 has to offer. In any case, I think Pete Beeby would need to be involved here as he has taken the work we did earlier to a new level, as he has illustrated in the forum topics here.

John

PS, I've checked your image against my files this morning and I'm pretty certain that is indeed my FSX Typhoon. The question is, who converted it to work in MSFS as AI and with what authority?

OK, so I have found and downloaded Nick's RAF Lossimouth package for MSFS and confirmed that my FSX RAF Typhoons (and Poseidons) have been converted for AI use in the package without my knowledge. I haven't gone as far as installing the package, but I assume from the first line of the thread here:" I've managed to get the FSX/P3D RAF Typhoons to show up - and move - at MSFS Lossiemouth", that the included AI doesn't actually work in any case. It certainly doesn't include PBR texturing, which is really a "must". It's not surprising given the complexity of the conversion process, that involves extensive MCX manipulation, xml coding and multi-component graphical design.

I have e-mailed Nick for an explanation.

John

Update,

Nick returned my message with an apology for the confusion. His intention was not to enable the Typhoons and Posiedons as AI in his Lossimouth scenery, but to make them appear as static objects. That's why he disabled the links in the AFCAD so they wouldn't move. We all know that MSFS doesn't handle custom AI well.

He is going to remove the AI aircraft components and replace them as placed static scenery objects in the package.

John

.....Glad to hear that you got some clarity on this John and that it will be resolved.

I just wanted to share my thoughts on what my current stance is, regarding MAIW and MSFS2020/2024.

We know, that we can get previous generation AI Aircraft models ( FSX and some P3D ) pretty much as they are, only a bit of config file work needed to get them to show up in MSFS. We soon found out however, that some things were never going to work on them, Afterburners for example, because these effects are done totally different in MSFS than in FSX.

I also noticed at that time, because of the look of the new MSFS2020, the models didn't appear to " cut it " any more and looked " visually impaired " in comparison with the other elements placed into the Scenery. In more simple terms, they stuck out like a sore thumb !

This was also the case when we first converted one of John's JYAI models into the new ( MSFS ) glTF format models, but did not go down the route of using PBR Materials, which were of course supported by the MSFS SDK.

It was at this point, that John (JYAI) , Steve ( Firebird here on the Forums ) and I ( The 3 MAIW Members that were initially involved in the FSX to MSFS conversion process, publicly here on the forums ) decided to pursue the conversions using elements of the SDK, which included the use of PBR Materials and modified XML custom coding etc.
This was a steep learning curve and at times very frustrating for all concerned.
On several occasions we had to " backtrack " on what we had already implemented, due to changes made to the SDK and so on.

Despite all of this, John was pretty much able to churn out a model to the new MSFS format, albeit using some " Legacy " elements.
Although the inclusion of PBR was going to be difficult and time consuming, he could clearly see the benefits to the visuals on his model when seen in MSFS.

John compiled a Tutorial supported by myself, for the FSX to MSFS model conversion process. Whilst some bits of it might be out of date, fundamentally it is still valid.

It is well documented here on the forums as to why MAIW ceased all further MSFS AI Aircraft developments collectively, I really don't want to elaborate on that any further.

What I wanted to say is that, with the knowledge that we now have, we are best placed for what comes next in MS2024.
With the steps already taken, to move our model conversion ( this refers to how the XML Coding is handled ) away as much as possible from the " Legacy " to " Modern " format of the MSFS model standard, will mean that hopefully whatever is required going forwards to get our models into MSFS2024, will be marginal.

Lets just hope we have an Improved AI engine in MSFS2024, to make all of this worth the while.


Regards,

Pete.