Still no destroyed models

[Second 0]

Okay penetreation discussion in it's full glory. I agree with SexyWombat that penetration model it self, if made simple, probably is simple to code.

Now.

Let's (again) presume we have lone M1A1 rolling down road, crew is getting drunk so they aren't most consentrated to keep up scanning their around. Jolly songs and stuff involved.

+

There is curve in road. In that curve we have experienced insurgents (having few RPG-7s) hiding behind bushes.

=

We have good ambush scenarion, where M1A2 probably will be reduced to pile of dust.

...

Not really. Infact we have scenario where insurgents opens fire from maximum distance at tank's FRONTAL armor. Crew is getting drunk so i takes time for them to react. But it doesn't matter as insurgents puts ALL their PG-7s into FRONTAL armor of tank! Then tank starts doing it's job with main cannon and machineguns. Imagine this in every skrimish or major battle. Not that pretty or realistic really.

Got it? It could be simple to add penetration model but it also requires AI to understand when it's not a good idea to fire at frontal armor and when it's not. I dont' know how hard process this is to put in to game. I tend to think that anything related to AI is pretty hard. True many games has this, but those are mostly mostly wargames and shooters which are designed for MP.

Reason why BIS dropped penetration models from most objects probably is because it was resource hungry. ArmA HAD trees and houses which were penetrable by at least 7.62mm long cartridges, some house's structures even with smaller. Then by patches ArmA started to run smoother and smoother (and people praised), and soon i discovered that i can't anymore penetrate house walls or trees with small arms fire. Penetration of structure even changed projectile's trajectory slightly. Interesting eh?  

[icebreakr 3157]

I'm glad my topic had so many replies so far, and hopefully BIS will incorporate the wishes of hardcore gamers into their latest masterpiece.

Not necessarily connected with destroyed models, but I will get very mad if I'll see bouncing models of burning choppers and aircraft in A2 Authors (MapFact) of AH64 addon had a great script that changed AH64 into a burning wreck model when it crashed. Hopefully BIS can pull off something similiar to that?

p.s. and BIS, please visit Watkins island and make a steep dive with chopper and try to gain height back... why does the terrain height still affect flight physics same as in old OFP? You should fix that, too

[vilas 477]

concerning this what you said Second:

http://www.flashpoint1985.com/cgi-bin....=damage

over year ago i was talking how to make penetrative/realistic damage engine using new LOD for "armor" and etc.

[Second 0]

concerning this what you said Second:

http://www.flashpoint1985.com/cgi-bin....=damage

over year ago i was talking how to make penetrative/realistic damage engine using new LOD for "armor" and etc.

Seems interesting, but i don't see part where improvements of AI to aim flanks of MBTs, and wait for flanks to be exposed, would be discussed. Would be nice to know how hard task it is to put such routine into game.

[Michael_Wittman 0]

concerning this what you said Second:

http://www.flashpoint1985.com/cgi-bin....=damage

over year ago i was talking how to make penetrative/realistic damage engine using new LOD for "armor" and etc.

Nice solution but you still too simple. I also did a efford some time ago on the Red Orchestra Forum. First you just need to set a general material thickness say its 5 for a UAZ at any side and 50.000 for M1 front hull armor..Then you set the equation to calculate the penetration of the round wich is:

- penetrated_thickness = f (ammo_type, gun_caliber, distance, angle)

Then you pick steel penetration capability from the most reliable tables you can get and calcule the function by Nonlinear Regression

Then you set some basic child meshes inside the vehicle representing vital parts. Depending the mesh the particle hits that will trigger the correspondant effect over the vehicle....from a mere stop to the amazement of the flying turret.

In adition you can deal with the shnarpel by giving the particles a penetration value according to caliber and distance and void the calculation according to vehicle armor (there is no point in making the calculation of the shnarpel from 40mm grenades over a tank)

So once we have calculated the penetration value and if the particle hits the child mesh representing the room for the crew/passangers then it can see different consecuences

penetrated_thickness > armor_value = full penetration (everyone dies)

penetrated_thickness < armor_value and >armor_value*0,9 = internal armor flanking (random kill/injure people inside)

Its evident that if the particle didnt reach this point it didnt penetrate and therefore ricochet off.

Thats my 2cents for the ballistics.  

[max power 21]

Okay penetreation discussion in it's full glory. I agree with SexyWombat that penetration model it self, if made simple, probably is simple to code.

Now.

Let's (again) presume we have lone M1A1 rolling down road, crew is getting drunk so they aren't most consentrated to keep up scanning their around. Jolly songs and stuff involved.

+

There is curve in road. In that curve we have experienced insurgents (having few RPG-7s) hiding behind bushes.

=

We have good ambush scenarion, where M1A2 probably will be reduced to pile of dust.

...

Not really. Infact we have scenario where insurgents opens fire from maximum distance at tank's FRONTAL armor. Crew is getting drunk so i takes time for them to react. But it doesn't matter as insurgents puts ALL their PG-7s into FRONTAL armor of tank! Then tank starts doing it's job with main cannon and machineguns. Imagine this in every skrimish or major battle. Not that pretty or realistic really.

Got it? It could be simple to add penetration model but it also requires AI to understand when it's not a good idea to fire at frontal armor and when it's not. I dont' know how hard process this is to put in to game. I tend to think that anything related to AI is pretty hard. True many games has this, but those are mostly mostly wargames and shooters which are designed for MP.

Reason why BIS dropped penetration models from most objects probably is because it was resource hungry. ArmA HAD trees and houses which were penetrable by at least 7.62mm long cartridges, some house's structures even with smaller. Then by patches ArmA started to run smoother and smoother (and people praised), and soon i discovered that i can't anymore penetrate house walls or trees with small arms fire. Penetration of structure even changed projectile's trajectory slightly. Interesting eh?  

Good post!

[vilas 477]

i think that making for next game such structure in something that will be "reorganised" fire geometry will help

now fire geometry has objects

good , if there would be 2 types :

- shield not causing vehicle dammage, but stopping bullet of hit lower than XX

- strategic for armour - like now present in f.g. lod

it will give more realistic dammage structure

[Commando84 0]

Okay penetreation discussion in it's full glory. I agree with SexyWombat that penetration model it self, if made simple, probably is simple to code.

Now.

Let's (again) presume we have lone M1A1 rolling down road, crew is getting drunk so they aren't most consentrated to keep up scanning their around. Jolly songs and stuff involved.

+

There is curve in road. In that curve we have experienced insurgents (having few RPG-7s) hiding behind bushes.

=

We have good ambush scenarion, where M1A2 probably will be reduced to pile of dust.

...

Not really. Infact we have scenario where insurgents opens fire from maximum distance at tank's FRONTAL armor. Crew is getting drunk so i takes time for them to react. But it doesn't matter as insurgents puts ALL their PG-7s into FRONTAL armor of tank! Then tank starts doing it's job with main cannon and machineguns. Imagine this in every skrimish or major battle. Not that pretty or realistic really.

Got it? It could be simple to add penetration model but it also requires AI to understand when it's not a good idea to fire at frontal armor and when it's not. I dont' know how hard process this is to put in to game. I tend to think that anything related to AI is pretty hard. True many games has this, but those are mostly mostly wargames and shooters which are designed for MP.

Reason why BIS dropped penetration models from most objects probably is because it was resource hungry. ArmA HAD trees and houses which were penetrable by at least 7.62mm long cartridges, some house's structures even with smaller. Then by patches ArmA started to run smoother and smoother (and people praised), and soon i discovered that i can't anymore penetrate house walls or trees with small arms fire. Penetration of structure even changed projectile's trajectory slightly. Interesting eh?  

Damn! Didn't know that at all, I knew that they done some fixes to tree's and foliage so there would be less lag when playing on the north side in one of the patches but I guess it could make people mad if they got shot inside house through walls all the time, I know i would even if i knew why..

About tanks blowing up vs not always blowing up I think it has to do with giving players the feedback so they know if they hit the tank or not

Its like you cocking some food in the microwave oven and you check visually if the food has changed its colours and shape enough so it looks like its hot

Player expect that tanks that get destroyed are shown off in a special way. Imo when flying a chopper or anti tank aircraft thats especially important sometimes when it goes so fast you check either the radar for red or white icon or you try to check for smoke, fire or explosion visuals or sound fx.

I guess Bi could put in some smoke for a knocked out tank or something cause as a pilot you're not always ensured you got the proper time to see if you took out a target, especially in a mp enviroment such as warefare where shit hits the fan from random directions.

Warefare in a even more updated version in Arma 2 will be really great to see how its changed, If it had some big navy ships in it it would have like everything a player could ask for.

[Törni 0]

It would be also nice if a hit could kill some crew members depending on where the AFV was actually hit. So a frontal hull penetration might kill the driver while gunner and/or commander can escape via hatches.

[lugiahua 26]

Hit points system can still work well, but depends on how detail they work on it

Look at F/A-18E/F addon, it detailzed different damage effect on airplane...

[Flanker15 0]

Getting the AI to not shoot at armor facings that are too strong for a weapon is quite doable.

The normal way is to assign every facing on a unit (front, back, side, top, turret front etc...) a value then assign every weapon a value.  

If the weapon has a lower value than the armor the AI won't use it against that facing.  The more complex you make it (more values, more weapon factors like range and armor type) the better the AI works.  

You can also attach experience modifiers to the system allowing situations where inexperienced units will fire on units that they can't hurt, also you can attach the morale system to it so paniked units will do the same.

The normal way to model the actual armor thickness of unit parts is with the RGB or greyscale system.  You have your unit model with all its parts and you paint on different colours/shades which represent different thickness, this lets you make a very detailed armor system with small parts having their own thickness.

[max power 21]

much of what you describe is in ArmA already, and was in OFP, just not on a per hit location basis.

[Michael_Wittman 0]

Getting the AI to not shoot at armor facings that are too strong for a weapon is quite doable.

The normal way is to assign every facing on a unit (front, back, side, top, turret front etc...) a value then assign every weapon a value.  

If the weapon has a lower value than the armor the AI won't use it against that facing.  The more complex you make it (more values, more weapon factors like range and armor type) the better the AI works.  

You can also attach experience modifiers to the system allowing situations where inexperienced units will fire on units that they can't hurt, also you can attach the morale system to it so paniked units will do the same.

The normal way to model the actual armor thickness of unit parts is with the RGB or greyscale system.  You have your unit model with all its parts and you paint on different colours/shades which represent different thickness, this lets you make a very detailed armor system with small parts having their own thickness.

mmmm...you are talking about weightmaps....and as thats only a extension of the hitpoint system gone mad...I mean if you want a detailed weightmap for every single point it can run wild...try to imagine if the mesh (tank or whatsoever) got hit by 4 different calibers from 4 different angles..then it becomes a moving matrix with quite a lot of numbers (position x.y.z.color) EACH ONE, EACH VEHICLE...even those left behind semi-destroyed....in best of the cases you´ll find the need to destroy the object as soon as its disabled, burn....etc

I think its not a bad idea but out of the capability of current hardware to render it live....also you can go further making the mesh more dense and use the weightmap to actually deform the mesh but again to render it live is science fiction today.

I think the aproach I gave some post ago is a good starting point to approach the problem...although its high level languaje (under C++ .... maybe).

The solution "on-the-run" to make AI recognize wether they are facing front or side or rear is make the mesh to carry another 4 invisible grids ..then you make the AI to make a local YX infinite proyection calculating how much of the invisible mesh is trim by proyection...then you compare trimmed grid with actual grid and set conditions as follows:

if distance center to center<400m

frontmesh=mesh1 // proyection trim AI on the vehcile

leftsidemesh=mesh2

rightmesh=mesh3

rearmesh=mesh4

if mesh1<=1.00 // this case is for not shooting front in any case

      dont shoot

elseif mesh2=>0.7

      random shoot * 0.6 // highly chance of shooting

elseif mesh3=>0.7

      random shoot * 0.6

else mesh4=> 0.5

      shoot

end

This is very very rough aproach but I think is comprehensive for making a single AI recognize if its facing front, side or rear...

Cheers

[Yoma 0]

Won't that end us up in a silly game of "point front armour to ai so they can't shoot at us"?

I mean won't that introduce another exploitable ai weakness?

[mr.g-c 6]

Won't that end us up in a silly game of "point front armour to ai so they can't shoot at us"?

I mean won't that introduce another exploitable ai weakness?

Why, its like that in RL too, i would also use it against human players....

Forget lame balancing, make it like in RL... as Michael_Wittman already explained, it should be possible to "program" AI so they would use RL-Tactics and attack a tank at its weaker points - just like "the enemy" would do in RL.

[Pulverizer 1]

Quote[/b] ]The solution "on-the-run" to make AI recognize wether they are facing front or side or rear is make the mesh to carry another 4 invisible grids ..then you make the AI to make a local YX infinite proyection calculating how much of the invisible mesh is trim by proyection...then you compare trimmed grid with actual grid and set conditions

Wha... Just compare "angle from soldier to tank" to "angle of tank". Difference less than 135, shoot. Very simple.

[mr.g-c 6]

Quote[/b] ]The solution "on-the-run" to make AI recognize wether they are facing front or side or rear is make the mesh to carry another 4 invisible grids ..then you make the AI to make a local YX infinite proyection calculating how much of the invisible mesh is trim by proyection...then you compare trimmed grid with actual grid and set conditions

Wha... Just compare "angle from soldier to tank" to "angle of tank". Difference less than 135, shoot. Very simple.

Do you think this is already doable by script?

Are there command for angle query?

[Pulverizer 1]

If it was scripted, it'd be a bit ugly to keep the AI from not firing by itself. But sure, there's getPos, atan2, getDir, doFire, removeMagazine and addMagazine.

Maybe you could make a modified AT weapon that had such priorities that it would only be used by AI when scripted to doFire.

[Maddmatt 1]

Quote[/b] ]...

Wha... Just compare "angle from soldier to tank" to "angle of tank". Difference less than 135, shoot. Very simple.

A simplistic solution like that would not be good enough.

[NoRailgunner 0]

There is much more to do and take into development if you dont want that ridiculous tank combat like in "Codename: Panzers".

Have a look into specifications of AT missiles and their usage. Its a difference if you shoot with TOW2 system, AT-14 Spriggan or with SMAW, RPG-29. What about different guidance and max combat ranges? Attacking (moving) tanks on all sides, in all positions and under every condition shouldnt be that limited for AI. If player can shoot at tanks, AI should be able to do the same. Imho adding extra AI weapon skill could improve AI and mission making too. Next thing is that vehicles ArmA2 could have different damage levels according to warhead, place of impact and protection systems. Would be nice to see that crews are not ejecting if the damage isnt critical. Think that AI should be able to drive (fast) backwards to get into cover or simply avoid beeing hit.

[Pulverizer 1]

Quote[/b] ]...

Wha... Just compare "angle from soldier to tank" to "angle of tank". Difference less than 135, shoot. Very simple.

A simplistic solution like that would not be good enough.

Obviously, it would be an addition to other stuff that goes on before AI decides to take a shot, such as line of fire, target speed and distance. It would be good enough to keep the AI from shooting at tanks from the front in a very simple and effective fashion, but not to make them good enough tank hunters of course.

[sanctuary 19]

Maybe the "useless" AI front AT shot problem could be avoided if damages were not impacting only the tank hit points or armor.

This way a direct front shot, while not destroying the tank, could seriously damage the optics, the main gun , the machinegun,... etc...

And would then become a source of worry, as even if nearly undamaged from the front shot, a then blind tank or having faulty weapons could then be a death trap for its crew until repairs.

[Michael_Wittman 0]

Quote[/b] ]The solution "on-the-run" to make AI recognize wether they are facing front or side or rear is make the mesh to carry another 4 invisible grids ..then you make the AI to make a local YX infinite proyection calculating how much of the invisible mesh is trim by proyection...then you compare trimmed grid with actual grid and set conditions

Wha... Just compare "angle from soldier to tank" to "angle of tank". Difference less than 135, shoot. Very simple.

Looks clever but:

- an angle is expressed 2D....so what angle do you pick? zx, xy, yz?

- you also didnt explained how to calculate the relative angle of the surface facing.

- neither you explained how can the AI distingue front, side and rear

Respecting the scripting although I havent yet dirt my hands on it most of what I´ve seen is low level languaje wich is slower to process...enough a dramatic drop on fps if...say 100 AI are executing.

Other thing that should be modelled is the area target for AI...have any of the present people ever saw a AI controlled tank shoot HE in a area target fashion?

Hitpoint for me in no-way-near a solution for armor combat and is 2nd thing to address after they fix the ultra-accurate aiming of the AI and dumb pathfinding.

One correction to my rouch approach...according to Israel-Lebanon conflict records a full penetration doesnt mean that all the crew are doomed..but instead a highly chance of at least 1 of the crewmembers to survive.... Im talking of 60tns tanks not 45 tns...

[Maddmatt 1]

...It would be good enough to keep the AI from shooting at tanks from the front in a very simple and effective fashion...

But in some cases that wouldn't be best. A simple angle check simply isn't a proper solution, not for a 3D game. You need to check what parts of the tank the AI can actually hit. You wouldn't want the AI to sit and do nothing when he is close enough to get an accurate shot and the tracks are visible...

And some weapons would still have a chance against the front of the tank.

Michael_Wittman's post covers the rest.

[max power 21]

And, of course, the angle checks and such would have to be combined with some kind of probability of the unit just firing anyway.  This would be especially important if you were designing damage effects for the equipment outside of the hull, like the prisms, gun sights, external weapons, and so forth.  Also, depending on what kind of mission you had in mind for the unit.  If you were in an open, pitched battle against tanks and apcs, you might want to hang on to those things.  If you are a unit like an insurgent, say, and you mission is just to fire a rocket into a tank and run away on the off chance that it will start a fire or something (as happens), then you'd have to take that into account.

I mean, having a really complicated system like per-point of impact damage solving would almost be wasted on something as simple as front/side/back.