Optical tracking model

General question about the A-10C module and optical sensors (TGP, Mavericks): how accurately is the optical tracking functionality of the sensors modeled in the sim? I mean, did ED actually write an algorithm that actively tracks objects based on camera pixels contrast? Or is this a simulated, simulated behavior?

DCS is a simulation. So it is simulated. There is no contrast. There also are no real objects nor sensors. It's all colorful pixels on your monitor.

The TGP and maverick a modeled in a way that you will enjoy a perception of operating real sensors though.

This simulation has its limitations. IR-imaging is somewhat odd...

DCS is a simulation. So it is simulated. There is no contrast. There also are no real objects nor sensors. It's all colorful pixels on your monitor.

 

The TGP and maverick a modeled in a way that you will enjoy a perception of operating real sensors though.

 

This simulation has its limitations. IR-imaging is somewhat odd...

Well, there IS contrast and those colorful pixels are all that would be necessary to make the simulation work the same way the real thing does it.

But in DCS it is implemented in a simplified way. DCS "knows" where the targets are and therefore can easily decide if you can aquire a lock or not - it has not to analyze the actual image.

The real thing can (for the good or bad) lock onto anything that provides enough contrast, but in DCS you can really only get a lock on an "object". This way you can even lock onto an object that is (partially?) hidden behind vegitation (remember: trees are "see-through" for everyone ... except us human players ;o). But afaik there is some simulation of how lighting affects the sensor's capabilities. DCS calculates a lock-on-probability based on the light conditions (when appropriate). I.e. you will have a hard time to get a lock with a CCD MAV during a pitch black night.

Flagrum, as you will know, the contrast you refer to is also nothing but colorful pixels. All imaging in DCS is a processing of graphics, not of simulated sensor inputs. Same with gain.

Just wanted to point this out very clear, since the OP's is looking for the simulation of sensing.;)

Flagrum, as you will know, the contrast you refer to is also nothing but colorful pixels. All imaging in DCS is a processing of graphics, not of simulated sensor inputs. Same with gain.

Just wanted to point this out very clear, since the OP's is looking for the simulation of sensing.;)

Perhaps we just misunderstand eachother here ... dunno if we disagree on something here or just agree ... ;o)

A ccd sensor is just a two dimensional array of photo diodes. Each diode can be considered as a "pixel". The electronics and/or software then just compares the output ("brightness") of each of these diodes and compares it to their neighbouring diodes. If the difference is big enough, then there is "contrast".

This could be almost directly replicated within DCS - pixels are just pixels, no matter how they were generated or where they come from.

But this determination of contrast (and some other criteria) is not done in DCS. DCS does not look for contrast to identify a (potential) object that a weapon can lock onto. Instead of this image processing, DCS just knows if the sensor is actually looking at an lockable object - as the sim knows the position of each and every thing already.

DCS _could_ simulate the image processing part like the RL counterparts and I think, it would not even cost too much processing power. But the results would probably not really be satisfying. Whatever a DCS sensor can look at, it will not have the exact same characteristics as in the real world. DCS does just not look 100% photorealistic and algorithms that work (more or less) well in RL, probably would not work as good (or too good?) in the sim.

Thanks guys.

This could be almost directly replicated within DCS - pixels are just pixels, no matter how they were generated or where they come from.

You totally got my point here.

Oh, and:

DCS is a simulation. So it is simulated. There is no contrast. There also are no real objects nor sensors. It's all colorful pixels on your monitor.

No?! Really?! Aaawwwwww, shizzles, I really thought every time I fired DCS there were actual military planes flying awkwardly, missing every other target and crashing in Ukraine... :lol:

Its a rather crude simulation of the weapon compared to the very exciting methods used in simulating the flight models. Sadly when you have a system this simplified it is inherently quirky.

So rather than simulating the environment and conditions and the capability it simulates the aggregate probability of capability and in a manner which is sadly very transparent for the end user after long enough. Its the reason why for instance the Force Correlate mode is useful as a sniper rifle when in real life its far less accurate than standard centroid tracking and so also capable of hitting at a far greater range than centroid mode in either the sim or in real life, by a margin so considerable as to be grotesquely unrealistic.

It is therefore one of those weapons where you need to make a conscious effort to use it realistically and self limitation is the only method for ensuring you aren't exploiting the weaknesses of the simulation.

...it simulates the aggregate probability of capability and in a manner which is sadly very transparent for the end user after long enough.

It hasn't become transparent to me yet. Could you explain a little more? How do these weaknesses become most apparent (Maverickwise I mean)?

I hope that in future DCS versions we will get real locking algorithms which are just fed with simulated input. Maybe ED can get permission to embed some older versions of real locking algorithms similarly he gets for other airplane systems.

It hasn't become transparent to me yet. Could you explain a little more? How do these weaknesses become most apparent (Maverickwise I mean)?

DCS is working digital in this case. Either the object can be tracked or it can't.

There is no computing of sensed data, which distinguishes between weak and not so weak signals and attempts to calculate best tracking marks and profiles out of this. Real life E/O trackers compare whole scenes and evaluate these for trackable signatures.

DCS only jumps on an object if it is defined to be trackable and if no environmental factors prevent that from happening (sensor masking, weather)

In real life things like background scatter (especially IR) or diffuse contours play important roles. Not so in DCS.

And that is why

to embed some older versions of real locking algorithms

DCS doesn't know temperatures.

Thanks Rongor. I always felt tracking was a way more complex thing than just directing the gate on an object and BOOM you get a lock. Now I begin to see HOW this is actually more complex.

Very interesting IMO.

And that is why

won't help here, as this is not how it works in DCS.

 

DCS doesn't know temperatures.

Well, I hope at some point DCSW objects will get temperatures too. Then, I don't see any obstacles to implement real locking algorithm.

Maybe new graphic engine will bring better IR handling.

They would have to create a layer of textures for IR, for all the models there are ingame...

Edit: read it as "for all surfaces there are"

Don't get me wrong, of course it would have been cool if they had tried that. But I can also understand that they didn't and probably won't. Probably it is not worth the required work. DCS features so much more than E/O.

Well honestly I feel that it would be worth it since this is more than just a flight simulator but also a combat flight simulator and combat is one part aircraft one part weapons.

Frankly the weapons in DCS are very unpolished. There are very few weapons that you could say aren't without gaping holes in their representation for one reason or another so tidying all that up, even with some concerted effort to focus on that and only that for a spell, would be one of those things that makes for a quantum leap forward because as a simmer its those extra little details that make this into more than just playing at realism but actually playing a game thats about realism.

Not much point in having a brilliant digital replica of an aircraft only to often find yourself not quite flying it like real pilots do because if you did some of your weapons wouldn't work quite how they're meant to.

I some how doubt it'll happen but... one can only hope.

It's a question of adding a IR color component to each vertex. Something like that must exist already to render current IR images, I guess.

But then, the scene that is shown on the MFD should be sent to a module as a raw image, and that module would execute the image filters needed to find a target or targets in that image. And yet some kind of stochastic algorithm should be needed to keep identification of targets in time from frame to frame. And that or similar would apply to every sensor of every aircraft and weapon using some kind of target lock through image processing, be it IR, visible or radar (microwave).

I don't see a simulator doing target lock that way. It's just too expensive. The simulation "knows" the targets. It's far better to do all kind of emulations of the real behaviour than using the actual algorithms, which, of course, surely they are top secret anyway.

Yeah true.

Probably, the physics benefit the most accurate modelling in the simulation. After all, mankind's been flying for some time now, and there's nothing secret anymore about the mechanics of subsonic flight. As to weapons systems... I believe the level of modelling is satisfying as it is now, considering the A-10 is a combat plane still in service today. Maybe in 20 years we'll have a simulator which depicts all these details with perfection, as the Hog will be quite obsolete by then, and home computers will be all that more powerful.

And to digress: I always saw big military stuff as classified, and I'm amazed one can get already so close to the real thing just sitting at home. I mean, really, it looks like anyone who can fire this mean machine's engines in the game could do it in RL! (I'm NOT implying that one could fly the plane and manage targets and all that stuff; there's so much more to real aircraft operation than we get in DCS. Even if the E/O or anything else was better modelled, you'd still not qualify to even enter a real army simulator I guess. But I don't care, this simulation is incredible, I love it! :joystick:)

Classification isn't really the crux of the issue. The physics of free fall munitions is no more difficult to comprehend for us than that of the lifting body principles on a modern fighter jet. That therefore gives us an unfortunate nagging sensation on the backs of our necks when we realize that the Mk82 AIR ballute is clearly failing to do its job and instead making the use of those weapons pointless. Much the same with the dispersion of the GAU8 which is greater than it should be or the serial under-achievement of the CBUs the 87 in particular.

None of these things are particularly opaque since what people who are struck by the knowlege we have of what they believe ought to be secret military stuff fail to comprehend is that its all public domain for the most part. Its called the laws of physics. Even something as complicated as a Maverick missile, once you know how it does it in principle, can be modeled even without full access because the laws of physics give us enough information that we can fudge it into existence.

Since a simulator is not reality but instead a closed environment that we can manipulate its therefore possible to tweak this thing we've created into performing how it would in real life. Its therefore just a question of effort in most cases and unfortunately it seems that you don't get the dollars by developing weapons nearly as well as when you develop aircraft. So it will be that DCS will swell with excellent aircraft and even A2G radars but will still have some nagging issue with the weapons that are carried by these platforms.

It seems then that mostly people are tolerant of these issues for the same reason people were tolerant of the issues with flight models 10 years ago. Its no better anywhere else so "it'll do" is your only choice.

the dispersion of the GAU8 which is greater than it should be

It is?!

the serial under-achievement of the CBUs the 87 in particular

You're basically telling me I might not be as lame as I thought at this.

Seriously I didn't know about the weapons flaws. I'm certainly not as hardcore as most here, but I've watched almost all YouTube has on DCS, read the manual completely maybe three times, read quite a few other PDFs, webpages and forum threads, and I haven't yet heard the weapons were that off.

EDIT: BTW, I like your analysis.

It is?!

Yes. There is an in game way to observe it. The BATA predictive round impacts (little donuts) are always in an apparently tighter grouping than the actual round impacts.

Someone has released a mod though that modifies the gun impacts to be tighter.

You're basically telling me I might not be as lame as I thought at this.

 

Seriously I didn't know about the weapons flaws. I'm certainly not as hardcore as most here, but I've watched almost all YouTube has on DCS, read the manual completely maybe three times, read quite a few other PDFs, webpages and forum threads, and I haven't yet heard the weapons were that off.

 

EDIT: BTW, I like your analysis.

Well there is plenty of debate to be had over the degree of the inaccuracies but its undeniable that for instance the JDAM is probably the least well modeled weapon since it has none of its real life counterparts main features beyond the basic ability to hit the SPI within a certain radius of certainty. The actual flight model for it is probably one of the least fleshed out (I don't want to sound like I'm denigrating the developers) and reaches its terminal velocity at release or something like that taking forever to hit compared to more well modeled weapons. The Paveways for instance apparently have an advanced flight model while the JDAM doesn't.

If you poke around enough you'll find the issues discussed. The CBUs in particular seem to have the most frequent discussions since they're the ones that have actually had some developer attention but still aren't quite there yet.

Its natural though that you'd not notice since you have no frame of reference. Most people even fans of military aircraft don't know much about the weapons in question and since even tactics about how to use them can be scarce even in these forums and are nonexistent in the manuals there's really no way for anyone to compare the modeling except when someone with a bit more knowledge comes along and informs us. Though when you read A-10s over Kosovo, hear about the effects of the CBU-87s, then compare that to the sim you can't help but furrow your brow.

At this point in time I'd rather see the A-10's weapons get totally upgraded to be as well modeled at hte A-10 itself rather than buy one of a dozen new modules that'll carry some of the same weapons. Its all a question of user desire though. Most users don't use the DCS modules to the level that they'd even notice the difference or care.

It is?!

 

 

You're basically telling me I might not be as lame as I thought at this.

 

Seriously I didn't know about the weapons flaws. I'm certainly not as hardcore as most here, but I've watched almost all YouTube has on DCS, read the manual completely maybe three times, read quite a few other PDFs, webpages and forum threads, and I haven't yet heard the weapons were that off.

 

EDIT: BTW, I like your analysis.

Here is few interesting threads about weapons:

http://forums.eagle.ru/showthread.php?t=135598

http://forums.eagle.ru/showthread.php?t=143556

Very interesting indeed, thanks Marluk.