AI_ATC Nellis AFB

[Gary]

This is right up my street too! - thank you for your efforts, for those of us who enjoy the added immersion ATC procedures bring this is very much appreciated.

On the basis I suspect users might need to tinker with VA profiles, voice training and alike, plus the self training you rightly point out required to use this properly (and that others might want to add Voxfox compatability etc) I think the sooner you can a version released the sooner all the above can start.

I'm frankly quite desperate to get my hands on this...lol

Regards and thanks again,

Gary

[Abburo]

On 8/8/2024 at 9:29 PM, USSInchon said:

THIS IS AWESOME!!!! great work.  I have a few notes (for reference I'm a 23 year tower and radar controller at both civilian and military airfields in the US). 

->The ATIS information we don't say QFE we just say "Altimeter Two Niner Niner Two"  All altitudes are read as single numbers ie One Four Thousand for 14,000.

->Runway taxi instruction phraseology is "Runway ## taxi via taxiway X X X [hold short or cross runway if needed]."  the pilot will hold short of the runway themselves.  also only when giving a progressive taxi instruction would you use turn left Charlie right alpha or something similar.  it takes a lot of time to say those words, and in our world seconds count, so the less we have to say the better. 

->Single pilot fighter aircraft takeoff clearance is: "c/s runway XX, wind XXX/XX, change to departure, cleared for takeoff" (personally I say change to departure after the takeoff clearance but that's what the book says).

-> In the US Flight Levels don't begin until FL180, below that its One Two Thousand One Seven Thousand, Two Thousand Five Hundred, etc.

-> When giving a radar vector and not using turn left/ turn right heading, say "fly heading"

It looks like some of these things were corrected later in the video when you were inbound, such as fly heading and removing the QFE and QNH phraseology.

-> The overhead sequence from tower should be "report initial (3-5 nm from runway), then give break instructions ie left/right break, at the approach end, mid-field, departure end, etc.  for sequencing in the pattern.

-> The runway exit instructions (since it seems to know where you're parking) are "Turn left/right next taxiway, at the end, or [taxiway name].

 

I wrote these notes as I watched the video.  again this is great work, just some things that would make it a bit more polished.

 

Ref:  7710.65AA Air Traffic Control Handbook (used by both US Military and Civilian ATC).  The tower section is Chapter 3, Radar is under Chapters 4 and 5.

 

Great comments and maybe some of your recommendations can be implemented.

However you might need to know the Moose framework is using Text To Speech APIs to transform written text in sounds. Hence, the digital voices used for real time translations don't have the "pilots" preparation to read one seven thousands, or two niner niner :),  but will just read the number dynamically provided which in this case will be seventeen thousands or 299. So, these are some technical limitations of the framework used.

Also entire system is based on various FSM (finite state machine) to know if the plane is moving, stationary, takeoff, landed etc. This system however is not aware about turning the plane or various positional characteristics unless is querying for status periodically, which translates to a high CPU overhead.

Having these aspects in mind, and with your experience, maybe you can suggest some workarounds for tuning this system?

[Avalanche110]

9 hours ago, Abburo said:

Great comments and maybe some of your recommendations can be implemented.

However you might need to know the Moose framework is using Text To Speech APIs to transform written text in sounds. Hence, the digital voices used for real time translations don't have the "pilots" preparation to read one seven thousands, or two niner niner :),  but will just read the number dynamically provided which in this case will be seventeen thousands or 299. So, these are some technical limitations of the framework used.

Also entire system is based on various FSM (finite state machine) to know if the plane is moving, stationary, takeoff, landed etc. This system however is not aware about turning the plane or various positional characteristics unless is querying for status periodically, which translates to a high CPU overhead.

Having these aspects in mind, and with your experience, maybe you can suggest some workarounds for tuning this system?

AI_ATC uses 2000+ pre recorded sound files and all processing is done locally.

I looked into using the MOOSE text to speech API, Its an excellent system and would have made things a lot easier. But it comes with some trade-off's.

* It's complicated for users to set up and would exclude 99% of players.

* It costs $$$ each time a message is generated.

For this reason I decided to stick with pre-recorded sound files.

[Spetz]

Looking great! One thing I didn’t see, or perhaps missed.

Are you changing actual frequencies in game or does the F10 menu override all frequencies when using VA? Would be great if you had to change freq’s as requested such as with SRS.

[HILOK]

2 hours ago, Spetz said:

Are you changing actual frequencies in game or does the F10 menu override all frequencies when using VA?

you have to change frq (manually or switch presets) 

[Abburo]

10 hours ago, Avalanche110 said:

AI_ATC uses 2000+ pre recorded sound files and all processing is done locally.

I looked into using the MOOSE text to speech API, Its an excellent system and would have made things a lot easier. But it comes with some trade-off's.

* It's complicated for users to set up and would exclude 99% of players.

* It costs $$$ each time a message is generated.

For this reason I decided to stick with pre-recorded sound files.

Yes and No.

It indeed can use the pre-recorded files and those are pre-built by Moose devs.

For Text 2 Speech API there are the following considerations:

1. Microsoft Desktop Voice - 100% free - is using your Windows installed voices. Depending on the backend use (srs.exe or gRPC) it can handle from 5 to 10 different voices

2. Google T2S API - a higher number of voices and more human modulated. This is free up to a monthly number of characters processed, which usually is good enough for single player. 

3. Other T2S (Azure, Amazon, etc) .. never used so cannot say anything, but i assume is pretty much similar to the Google API.

Conclusion: T2S can be used completely free with no limitation with MS Desktop Voice. Else the scripts can be configured to quickly change the provider if the case, so that each week let's say you may use a different provider to avoid getting into paid plan.

Not very straightforward configuration for sure, but perfectly doable.

 

Edited January 18 by Abburo

[Abburo]

4 hours ago, Spetz said:

Looking great! One thing I didn’t see, or perhaps missed.

Are you changing actual frequencies in game or does the F10 menu override all frequencies when using VA? Would be great if you had to change freq’s as requested such as with SRS.

The ATC frequencies are defined via moose configuration. While default they cannot be dynamically changed during gameplay, with a bit of scripting a menu to change the ATC frequencies during gameplay can be possible. However this is not realistic, as tower frequencies are always a given.

[FupDuck]

Wow this sounds really cool! Am I right in understanding though that this is only available for NTTR?

[t_hedlund]

49 minutes ago, FupDuck said:

Wow this sounds really cool! Am I right in understanding though that this is only available for NTTR?

Yes, that is my understanding, I would imagine that once this gets released for NTTR, there will be much demand for it on other maps. It would be great if ED would implement this into the core sim. 

[Abburo]

@FupDuck @t_hedlund 

This Moose functionality applies to all maps and has nothing to do with ED. Time to time might be broken due to maps updates done by ED, however the moose dev response time is extremely quick and from user perspective it requires only to update the moose.lua file within your mission.

Be advice, this functionality requires a minimum scripting understanding from end users end!

[Avalanche110]

@FupDuck @t_hedlund @Abburo

Correct, its just Nellis AFB. 

I would like to do more air bases on other maps in the future. I've tried to make the code as reusable as possible, so future projects "Might" not take as long to develop.

Just to clarify, AI_ATC uses some of the MOOSE classes but It is not part of MOOSE. I am not affiliated with the MOOSE Dev team.

[Purzel]

Excellent work! Outstanding!
That's exactly, what we need....
Thank you and your team for all that efforts you put in this!!!
Must be a lot of work...

As far as I understand, you build a dedicated voice-setup for only NTTR, but
wouldn't it be easier to generate just the names of all known airfields on all maps and just do the rest of the voice-lines in a generic way
(giving runways, directions, heights, etc.), so you can mix them depending on what airfield you are?
Correct me, if I'm wrong - I don't have any clue how your approach to this is or how it has to be programmed...!

Do you consider a version without using "Voice Attack"...?
We already have so many external tools running (as SRS, any-joystick-software, TrackIR, DCS BIOS, TacView, etc.) so another software would use resources again and I would like to have running as few as possible software in the background...
Apart from this the "export.lua" needs a certain order of it's command-lines to work properly, so that could be a problem at the end...
(I don't know if AI ATC will go into the export.lua, but you know what I mean... )

And if you need someone to test it, here I am...
Greets Purzel

[Avalanche110]

 

Great questions @Purzel

It was designed to be modular and portable to other maps. It can read out the name of any airbase or runway heading.

However it does require a lot of data points in order for it to work. I think each airbase will be its own project. How much work is required will depend on the airbase.

It’s not essential to use Voice attack, but I highly recommend it. It allows you to keep your hands on the HOTAS rather than fumble with key commands.

its also a great tool for learning and practising realistic Radio Calls. Performance over head is minimal and I don’t notice any impact on DCS with it running.

[Spetz]

On 1/18/2025 at 2:45 AM, HILOK said:

you have to change frq (manually or switch presets) 

Perfect. I hate using easy comms and changing freq’s to talk to tower, ground, etc, will be great.

[t_hedlund]

@Avalanche110 thank you for elaborating on how it works, look forward to it being released. 

[Avalanche110]

Part two of the Radio call series covering VFR recoveries back to Nellis.

If you like the video, be sure to like, share and subscribe.

[gunterlund]

is this avail or work in progress?

[Avalanche110]

Part 3 of Cabi's tutorial series. 

This video covers lFR Recoveries back to Nellis. A HUGE thankyou to Cabi and Garf for the work they have put into this Tutorial series.

[Gary]

This has been teased for a while now and I'm really looking forward to trying it out. Do you have any release roadmap - even if subject to change of course?

Regards,

Gary

[Avalanche110]

@Gary Thanks for the question.

The good news is that AI_ATC is "Feature complete" for release.

We are working on trying to ensure that users get a trouble free experience when setting up AI_ATC and Voice Attack.

Its a very complicated subject matter, combined with complicated software. We don't want the barrier to entry to be insurmountable for new players.

Right now were working on documentation and testing, understanding where players have difficulty and finding ways of mitigating it.

I'm confident it will be released within a few weeks. (subject to change) 

 

[Gary]

Thank you for the update. Appreciated.

Gary

[PeneCruz]

Watching with interest 

[MirageEnjoyer]

Eagerly awaiting this! An instant addition to my NTTR servers when this goes live. Thank you in advance for all of your effort.

[USA_Recon]

On 3/2/2025 at 5:57 PM, Avalanche110 said:

@Gary Thanks for the question.

The good news is that AI_ATC is "Feature complete" for release.

We are working on trying to ensure that users get a trouble free experience when setting up AI_ATC and Voice Attack.

Its a very complicated subject matter, combined with complicated software. We don't want the barrier to entry to be insurmountable for new players.

Right now were working on documentation and testing, understanding where players have difficulty and finding ways of mitigating it.

I'm confident it will be released within a few weeks. (subject to change) 

 

Great news - hopefully you’ll have some different options - small groups etc - would want a “free option” if possible.

[Avalanche110]

A huge shout out to @PeneCruz for integrating VAICOMPRO with AI_ATC.