Tank AI and Armor

There are couple of big issues with tank AI and tank armor on some units (like M1A2 and T-72/80/90)

I made a video going into detail, demonstrating all issues, but it turned out 52 minutes long and I am guessing no one will watch it, so I will try to explain the main issues in text format.

Arguably, the M1A2 is the most advanced tank we have in DCS, as all other MBTs don't have a specific model and we can assume they are the base ones. Saying that, it would be normal to think that it will perform the best.

The thing is, it performs on a different level. It dominates all other MBTs. The russian MBTs should have a trick up their sleeve, ATGM. Here comes a problem : AI won't use the ATGM unless they are "out" of AP range, which is around 3600 meters and a consistent number across all MBTs in the game (don't know why the AI is forbidden of engaging targets with AP beyond that range and the player (and the AI) cannot lock onto targets further away with AP, but can with HEAT). Additionally for some reason it takes them 13 seconds to aim a shot for an ATGM, but it takes them 4 seconds to aim an AP shot.

So combine the facts that the Russian MBT AI won't use its most accurate weapon, unless at extreme ranges, is highly inaccurate with AP and the armor on the ABRAMS is on another level from all other MBTs in the game and you have a pretty frustrating experience.

If the the Ru MBTs are outside of their AP range they will shoot ATGMs, hit (most of the time), do enough damage to disable the enemy tanks (when the Abrams doesn't absorb the ATGM) and be able to easily finish off the M1A2s. Unless, of course, the M1A2s don't snipe them from 4KMs with a HEAT shell from their first shot.

The ATGM from the T-90 does the same damage hitting from the front and the side. HEAT shot to the side of an M1A2 one shots it, while an ATGM doesn't.

Here comes a twist, T-72 AI on average does better than T-90 AI against M1A2. T-72 HEAT shells can penetrate more than T-90 shells can. T-72 HEAT shell can pen the lower part of the lower plate, while T-90 cannot.

Ah, armor.... well, now that I mentioned lower plate.... for some reason, the hull armor (frontally) on the M1A2 is stronger than the Turret armor. Isn't the whole point of modern MBTs to be able to engage hull down ? Meaning, the cheeks on their turret are usually the strongest armor on the tank.

Right now, in DCS, the cheeks are the easiest part of the M1A2 to penetrate. AP shots in the lower plate, upper plate, driver hatch, trap (turret ring) shot don't do anything, even from point blank. The breach of the gun sits somewhere between the hull armor and the cheeks in terms of how much it resists (which should be the weakest part of the armor) and the cheeks, more often than not are a one shot kill with AP, unless long range. T-90 HEAT struggles to penetrate M1A2 side most of the time, meanwhile the T-72 HEAT can do damage to the lower plate. Russian MBTs have the worst armor in game, they are one shots 90% of the time, no matter range or angle or which part you shoot.

I am guessing AI aims for center mass, so you can guess where most shells hit... the hull (if they even connect a shot).

I know that this is a deep rabbit hole and we cannot really have a high fidelity armor simulation, as most of it is classified, but there are some core issues, which have to be addressed and game play will be more than good enough for DCS. You can see now why the video came out to be 50 minutes long :music_whistling::lol:. It is kinda of a repeat of the M4 soldiers vs. Russia AK soldier. You can't really have a real battle, because one side clearly outperforms the other one all the time, because of artificial handicaps (AK solider being multiple times more inaccurate and russian tanks not using their ATGMs)

Issues #1: AI weapon choice has to be adjusted, if they can take in consideration aspect of hostile unit, even better, but this AP range border has to be adjusted, ASAP. ATGM should be primary weapon of choice against MBTs frontally up to 500-1000 meters and the aim time should be half that of a normal shot.

Issue #2: Some of the armor values have to be checked. I am not saying that the M1A2 has to be nerfed, modern tanks have excellent protection, hull armor and turret armor has to be checked and other MBTs have to be brought up to standard.

Issue #3: If AI cannot fire under a certain health percent, so should the player. Increase of reload time with damage would also be nice (I believe it used to be a think, but not present now)

Ground units although on first hand simple, have a lot of potential and actually a lot of small detail to them, they just need some polishing. If the technology used with AI aircraft, to scan around them and react to threats they have only scanned (meaning, no all seeing AI), can be applied to the ground units, it will be awesome.

Most of the issues can have simple fixes for now, by just changing value numbers, like at what range to shoot what and armor values

Upload the video. Maybe the developers will have a look. Not so many players are experienced with the Combined Arms module because it's not what most players are after in DCS. So, please share the video :thumbup:

Well, here it is... pretty informal and loose, but...

pqDjv26pnOs

In general, tanks are suffering from extreme low-fidelity, the list is quite extensive...

As far as armour goes, the damage modelling is very crude (health bar), with no subsystem damage (external or internal) modelled. I'm talking about things like individual armour elements, tracks, engine, transmission, horizontal/elevation drives, ammunition, breech, barrel, crew members, optics/sensors or radios.

Penetration mechanics do seem to be simulated somewhat, but seems largely approximated based on aspect, distance and coarse location. For instance testing the T-90 against M1A2 at 1000m, firing frontal aspect, 30° and 90°.

I found that the M1A2 is invulnerable to both the AP and HE rounds frontally, some shots damage the tank insignificantly on rare occasions and kill the tank in even rarer occasions (usually when the round hits either the UFP or driver's optics/turret ring.

At 30° and 90° OHKs are near impossible not to get, with both AP and HE (even if you fire at the turret cheeks, which AFAIK would present the greatest LOS thickness possible, and even if it penetrated, would only take out the breech/barrel/elevation drive at best, so there is *some* degree of armour element modelling, but it definitely doesn't calculate penetration based on LOS thickness or anything.

It's just, from the front it's largely immune, and from everywhere else, it isn't.

Here's a guesstimated idea of what the M1A2's protection is like, based off of the only source I can find (Swedish trials), take with giant mountain ranges of salt. The real thing is classified.

I mentioned that individual armour elements/arrays aren't modelled, as evident with the tests against the Abrams, and ERA isn't implemented at all.

But tank protection is just one thing, there's a lot missing, or that could do with some serious overhauls, but I won't get into it here, they're low-priority/not planned at all, and probably won't be until we get a full-fidelity tank module, which I would welcome.

Well, the health bar system is not necessarily bad, with health lowering, the top speed is decreased and reaction time of the crew is increases (reload time used to be increases, as well, I believe) until it reaches a certain amount of health, where AI will no longer be able to engage with the main gun.

I think these are reasonably good levels of damage modeling for what we have right now (if the reload times are increased). Is it perfect ? No, but for now it is good enough. Problem occurs when a player takes control of that low health tank, which supposedly has a disabled gun and it able to fire it and reload with its usual speed.

Vehicles do have different armor values at different places and angles do affect them, but some values just need adjustment/fixing. The values on projectiles as well.

I think this would require ED to start a whole new chapter if they'd start dedicating more time and effort for simulating tanks and other ground units. No doubt, it would be incredible! Imagine operating a tank with other players online at full fidelity! WOW!

Who knows. ED might be even bigger in years to come and a number of developers might eventually dedicate their work to ground units.

Give them time... Other issues are more demanding as of now.

Nice data there, Shadow! Got it from a Chinese user or something?

I think this would require ED to start a whole new chapter if they'd start dedicating more time and effort for simulating tanks and other ground units. No doubt, it would be incredible! Imagine operating a tank with other players online at full fidelity! WOW!

Who knows. ED might be even bigger in years to come and a number of developers might eventually dedicate their work to ground units.

Give them time... Other issues are more demanding as of now.

 

 

Nice data there, Shadow! Got it from a Chinese user or something?

Found here https://forums.eagle.ru/showthread.php?t=264115, another thread I had started long time ago. It was referenced from another thread, posted by a Deka developer I believe.

Full fidelity tanks are a far fetched goal, in my opinion. A short term priority should be the polishing of armor and projectile values, as well as AI improvements, followed by more damage levels (like gun not working and increases reload times)an improved and extended Combined Arms User Interface, including the ability to operate units more like in a strategy game, getting rid of the group mentality, just my two cents.

It makes me sad to see 'hitpoints'. More so to see the fanboys fight for "full realism" for things like weapons loadouts, rivet counting, water droplets, avionics colors, but ignore, deny or simply shun that everything outside of the cockpits in DCS World Core, is very unrealistic, shallow and simplified.

It makes me sad to see 'hitpoints'. More so to see the fanboys fight for "full realism" for things like weapons loadouts, rivet counting, water droplets, avionics colors, but ignore, deny or simply shun that everything outside of the cockpits in DCS World Core, is very unrealistic, shallow and simplified.

You cannot not have hitpoints.... Don't understand what the big drama about hit points is ? You think that if vehicles have module damage or have highly detailed damage models, there won't be hit points ? Instead of having one general pool of hit points, you will have hit points for tiny individual parts of the vehicle....

Do you think that real time simulation of complex materials on different ground and air vehicles on a large scale can happen with today's technology ? No it can't...

Shallow ? I would rather say that you are uninformed.

...

Full fidelity tanks are a far fetched goal, in my opinion. A short term priority should be the polishing of armor and projectile values, as well as AI improvements, followed by more damage levels (like gun not working and increases reload times)an improved and extended Combined Arms User Interface, including the ability to operate units more like in a strategy game, getting rid of the group mentality, just my two cents.

In that case I agree with you. BIGNEWY just confirmed working on the issue. Obviously it will take time.