IFF implementation

nuff said

More things to set up during EGI alignment ... :thumbup:

Would be nice get some response from the AI in case of IFF not set, or wrong code ect :smilewink:

I'm all all for this, as I already switch my IFF on at startup! :P I'm sure this will get implemented in DCS when the next US fighter module is released though, as the A-10 will need to respond to other aircrafts' IFF queries, or ED implements some kind of transponder communication with ATC, possibly if the radios are out and trying to emer. land.

From what I have read, the IFF system would be fairly complex to impliment. It is also an area where the line between classified and not is very thin making an effective and approved model that much more difficult.

Yeah, making an IFF system that "works" is easy, but making one that works correctly including all steps and the errors inherent to the system is a whole different problem. Then again, this does also go for radar systems with false contacts, jamming and so on.

IFF is weird.

From an AI perspective it's very hard to do, as currently it 'knows' everything is hostile or friendly. Once you introduce IFF, everything would have to be neutral until proven otherwise (or however you set your mission ROE). It would be awesome, but a major headache to the AI programmer who would now have to consider what to do when targets are considered as having invalid IFF (but aren't necessarily hostile). Last thing we want are forces shooting each other down.

The classification of it gets really stupid. The NSA gets fidgety when you mention COMSEC, and even though the only truly classified aspect of IFF is Mode IV (and, specifically, the daily keys for it, for obvious reasons), no one actually knows how much they'd be able to model.

The way Mode IV works isn't classified... what the signal is isn't classified. Hell, I don't even think the way the pulses send the signal (and thusly communicate to other systems which data is Mode 1 / 2 / 3/C / 4) is classified.

But once you have them all together, suddenly it's classified. It's kind of stupid, since the fundamentals of IFF are well-known by almost everyone who cares at this point. Honestly, as a COMSEC guy and F-15E avionics guy who works AAI/IFF - I can't tell you why modeling IFF would be a secret, because I really can't figure it out myself.

The only real answer I can give you is that the NSA is a bit retarded. I won't go really into it, but the way they demand COMSEC handled is really antiquated...

IFF implementation :thumbup:

Nail on the head Frostiken. And it's not just the NSA, it's the same it every country. The same applies to many other systems on modern aircraft as well.

Most if the stuff that is actually classified is totally irrelevant to developing a sim, but it still causes issues. Sadly it's not a situation that's likely to change.

Well, since the next DCS is, regardless of airframe, going to be something with an AAI system, maybe ED will surprise us and got a workable solution :]

Here's hoping.

To ask you guys who obviously know what you are talking about. What if it remains like now, with only radar ID to go on, no IFF? Would it mean a big hit on realism, or isn't IFF really that important? (Questions on the edge... :P)

Modeling IFF isn't a problem.

Even if you want to model IFF problems like multipath issues or a friendly responding between you and a bandit, it isn't an issue.

Classification isn't an issue either.

ECM/ECCM? Not an issue.

All of the above is strictly time and effort to model an instrument.

The /biggest/ problem, as described by everyone above, is the AI's decision making ability with respect to gaining IFF and ROE.

Do you just shoot at everything past some point on the map if it doesn't respond friendly (or fails to be ID'ed as friendly over the whole ID matrix)?

Do you have to VID everyone?

Maybe you need to just intercept a plane with a particular tail number?

What if your friendly is coming back with a busted IFF set - how do you set up safe corridors and ID procedures (an intercept aircraft etc)?

This is all ... not easy, and very tedious, but it can be done. Now, given that for example the ATC is relatively simple in DCS right now, I wouldn't expect any semblance of a complex IFF implementation.

IFF is weird.

 

From an AI perspective it's very hard to do, as currently it 'knows' everything is hostile or friendly. Once you introduce IFF, everything would have to be neutral until proven otherwise (or however you set your mission ROE). It would be awesome, but a major headache to the AI programmer who would now have to consider what to do when targets are considered as having invalid IFF (but aren't necessarily hostile). Last thing we want are forces shooting each other down.

 

 

The classification of it gets really stupid. The NSA gets fidgety when you mention COMSEC, and even though the only truly classified aspect of IFF is Mode IV (and, specifically, the daily keys for it, for obvious reasons), no one actually knows how much they'd be able to model.

 

The way Mode IV works isn't classified... what the signal is isn't classified. Hell, I don't even think the way the pulses send the signal (and thusly communicate to other systems which data is Mode 1 / 2 / 3/C / 4) is classified.

 

But once you have them all together, suddenly it's classified. It's kind of stupid, since the fundamentals of IFF are well-known by almost everyone who cares at this point. Honestly, as a COMSEC guy and F-15E avionics guy who works AAI/IFF - I can't tell you why modeling IFF would be a secret, because I really can't figure it out myself.

 

The only real answer I can give you is that the NSA is a bit retarded. I won't go really into it, but the way they demand COMSEC handled is really antiquated...

Modeling IFF isn't a problem.

Even if you want to model IFF problems like multipath issues or a friendly responding between you and a bandit, it isn't an issue.

Classification isn't an issue either.

ECM/ECCM? Not an issue.

 

All of the above is strictly time and effort to model an instrument.

 

The /biggest/ problem, as described by everyone above, is the AI's decision making ability with respect to gaining IFF and ROE.

 

Do you just shoot at everything past some point on the map if it doesn't respond friendly (or fails to be ID'ed as friendly over the whole ID matrix)?

Do you have to VID everyone?

Maybe you need to just intercept a plane with a particular tail number?

What if your friendly is coming back with a busted IFF set - how do you set up safe corridors and ID procedures (an intercept aircraft etc)?

 

This is all ... not easy, and very tedious, but it can be done. Now, given that for example the ATC is relatively simple in DCS right now, I wouldn't expect any semblance of a complex IFF implementation.

The answer to all these is real war zone orders. Not quite sure but I suspect that the paradigm from real life is that if something doesn't respond to interrogation and refuses to comply on radio gets shot.

But I really don't think ED has to do such a hardcore implementation. For me it would be good enough for AI to know if its friend of foe and AI that belongs to same flight as user can wait for player to make the assessment. That with the addition of a small possibility of false interrogation would model the real world paradigm close enough without requiring the amount of resources that hardcore full modeling requires.

Anyways its a simulation at an arbitrary level, even with systems fully modeled one can argue that its not full simulation because ED models systems and not circuits that respond to environment.

Not quite sure but I suspect that the paradigm from real life is that if something doesn't respond to interrogation and refuses to comply on radio gets shot.

I doubt that decision is so immediately clear, and the why is because of what GGTharos says, what if the transponder/IFF is broken and the radios are out, real life pilots would have to make a rational decition based on the facts he is given. Has the aircraft in question possibly been shot to pieces? Is he coming from the vicinity of enemy airspace? Are there other factors like a combat patrol/flight plan scheduled for this aircraft that is likely to make it friendly? Like what Frostiken and GGTharos agrees upon, I can imagine that making a good AI taking all these factors into consideration makes for a difficult task.

I'm certainly no Expert, nor i will tend too be one :P

But somehow i get a feeling it shouldn't be too hard too implement at player level eventhough i can understand its quite hard too implement it for the AI..

But isnt there a golden middleway...

I'm mean for a player IFF shoudn't be too hard too implement with all it quirds and oddities of IFF, and in doubt i think like in RL a player can eyeball a target too see if its friend of foe like i think they have too in RL..

But i understand its the AI that causes troubles for implementing it..

Isnt it possible too "cheat" this into the AI (i hate the word cheating)..

Lemme try too explain what i have in my head..

AI has a reference table of Neutral/Friendly/Hostile units, when it notice a target it looks up in the table if its neutral/friend/foe ignores Neutral and Friend..

Within a certain area of operations all non friendly targets are fair game, i doubt a Civ with a Cessna 172 will fly happely around in the middle of a air/ground war without beeing shot by both sides (i doubt in RL in a combat zone where fighting is going on, neutrals and civians arent supposed to be there, and if they do, i recon they are on there own, and smart thing for them is too stay the hell out of sight at least that i would do caught in the middle of a raging battlezone)

In Case its a Friendly with IFF problems, the AI still gets a signal from the damaged plane when its in (certain) visual range, so it knows that its a friendly with no working IFF, cause the plane type is in its friendly reference guide

I know its not a perfect sollution and it nowhere i suppose how it works in reality but i think with some clever programming most traps described here above can be ruled out by this..

Again i'm no expert, but it sounds to me that this would gimmick the working of IFF to a playable model for a AI and maybe you smart programmers can even make this better as i can think of.

That would be great if that's /all/ you had to do. Maybe those smart programmers have already thought of this and a lot more ;)

Maybe maybe not, i'n my expierence sometimes in my line of work sometimes people overcomplicated sollutions wich tend too end in endless discussions, while some idiot came up with a newbie idear wich no-one had thought of and was a perfect sollution, cheap and simple..

But them we are truckers not programmers ;)

Hehe, nothing like a little rumor thrown out there to stir things up hey GGTharos ;)

I see what you mean. I am currently on the fifth year of my master's in informatics, and have enough experience to understand that something is usually more complex and time consuming to implement in practise than what people usually think. Having said that, it has occurred that 'new' guys have come up with brilliant and simple solutions, although I believe the definition of a good programmer is how good he/she is at making a simple solution to a seemingly hard and complex problem, and that usually takes a lot of experience.

The problem is really that people make a lot of assumptions inbetween making decisions that they do not state.

To try and explain how this works ...

First and foremost, to implement a working IFF, the AI has to be programmed with the ability understand and ROE's.

In a simple case, where you have 'Fire on Identified Bandits' vs. 'Do not fire on unkowns' the AI now needs to be programmed with the following:

- An ID 'confidence' status for their target.

- A 'track' for the target, which can be lost. The idea is that the AI aircraft should be able to reasonably guess that the target that reappeared 30 sec after dropping from radar is the one they had been tracking.

- An ID matrix. GCI/AWACS, EID (IFF/NCTR), VID, location, target behaviour/flight profile.

- Techniques to gain required ID. Ie. they now need to 'learn' how to intercept a target without being subject to the target suddenly turning and blasting both of them out of the sky.

That's just the basics, and they don't even go into how the AI is going to end up being aware of locations, how they can autmatically set up a CAP to intercept and check out 'non responding unknowns', etc.

If you can't program things to be like this, there's pretty much no point in doing anything regarding IFF since it won't add squat to immersion other than setting a couple dials, which you can already do.

The problem is really that people make a lot of assumptions inbetween making decisions that they do not state.

 

To try and explain how this works ...

 

First and foremost, to implement a working IFF, the AI has to be programmed with the ability understand and ROE's.

 

In a simple case, where you have 'Fire on Identified Bandits' vs. 'Do not fire on unkowns' the AI now needs to be programmed with the following:

 

- An ID 'confidence' status for their target.

- A 'track' for the target, which can be lost. The idea is that the AI aircraft should be able to reasonably guess that the target that reappeared 30 sec after dropping from radar is the one they had been tracking.

- An ID matrix. GCI/AWACS, EID (IFF/NCTR), VID, location, target behaviour/flight profile.

- Techniques to gain required ID. Ie. they now need to 'learn' how to intercept a target without being subject to the target suddenly turning and blasting both of them out of the sky.

 

That's just the basics, and they don't even go into how the AI is going to end up being aware of locations, how they can autmatically set up a CAP to intercept and check out 'non responding unknowns', etc.

 

If you can't program things to be like this, there's pretty much no point in doing anything regarding IFF since it won't add squat to immersion other than setting a couple dials, which you can already do.

Cool! You did the ground work, now start programming , 1.11 is not far :smartass:

Just kidding of course :smilewink: (no, really)

I work in the game developer business myself, so I know very well the time, energy and hard work needed to implement new features.

But if you ever have the opportunity, this would be a nice addition. :thumbup:

I don't program for ED, I just volunteer my time for some stuff. But I don't mind submitting my ground work. As you no doubt know though, it's all time/effort, and one picks some features over others ;)