My concern about the hind....

[Fri13]

On 4/6/2021 at 10:15 AM, 3WA said:

Well, the Mi-24 is more of a troop transport than a tank killer.  It's weapons should really be for defense.  At most, it should be used offensively in COIN style situations, where you aren't facing major anti-air threats.  That's more for the Ka-50 and the Su-25.

 

Just the opposite. Mi-24 was made as "Flying IFV" to be multipurpose helicopter for various tasks. And it does it very well. It has the Anti-Tank capabilities better than smaller dedicated ones. It has AT rockets and in P variant the 30 mm cannon to take out most of the tanks, why it will have ATGM to destroy better protected MBT's and long range targets where it can't reach with those two.

 

 

The weapons it has are for offence purposes, not for defense. It is a primary attack helicopter for army aviation to get troops on the ground in front line and continue supporting them and other landing troops from Mi-8 as common air assault mission is. And directly operating in a Anti-Air threaten environments with MANPADS, AAA and short range SAM systems, hence it received flares, IR jammer and even RWR. It is not a SEAD or DEAD helicopter, and not meant for a scouting either like KA-50 is made to. Need to point that while it can transport infantry effectively, it was not its primary tasking as it is more useful as attack helicopter than "Taxi to combat" that Mi-8 does under protection of the Mi-24.

[Fri13]

On 4/6/2021 at 7:37 PM, Mad_Shell said:

mmmh, in my experience trees block AI los. I believe it may be explained by 2 things:

1) you gave the task "attack unit" to the Bradley, and in this case the AI is omniscient about the unit it has to attack, and always knows exactly where it is

 

The LOS calculation should be performed even then. The problem is that AI knows all the time where every unit is moving even if there are trees or buildings between, only the terrain will block the LOS checking and AI will revert back to its previous state. This is problematic as when you perform a pop-up behind a ridge to hover above trees, the AI has already reacted to you on the moment only trees were between you and AI, so on split moment you peek over the trees, the AI will fire at you and you see shells coming on your way in moment you lay eyes on the target. 

 

The Group logic as well applies, if any unit in group detects you, all units will detect you. There is small factor in difficulty level where there are couple second random delay between units but all know exactly where to shoot. 

 

On 4/6/2021 at 7:37 PM, Mad_Shell said:

2) the helo is on the ground, so it's atually not masked by the foliage

 

That is one of the problems in DCS, we don't have ground foliage (to hide ground units, to hide low level flying helicopters). The LOS is required to be straight there as ground vehicles fight that way. You have units inside forest of opposite side and unit will shoot each others through the thick forest if the trunks don't block the view. 

 

We need bushes, we need dozen of different kind trees, small and thin and not just wide and tall. We need more bushes than just couple kinds. We need variation around the forests etc. Inside a forest they can be so that there are just few bushes or none. But outter edges should be heavily different kinds. 

On 4/7/2021 at 10:08 AM, Lurker said:

I think the invisible "hitboxes", for lack of a better word, on trees need to be a made a little larger so that they block lines of sight more consistently. LOS in DCS is not the same as in Real Life. In real life LOS would make the enemy units unaware you were there, in DCS this simply means that the enemy AI knows you are there it just won't fire at you until it has a clear sight. Please correct me if I'm wrong, but the AI In DCS seem to be pretty omniscient of enemy units in a certain radius. 

 

The hit boxes are pretty good around the trees. But the AI spotting capability needs to be changed. If the vehicle crew is not buttoned up then they will have wide view zones to look around. And if they don't have engine running then they can hear you from long distance. But when they are buttoned up and engine is running, they should be pretty much blind to outsides. Have a few view zones and simulate the radar scan zone behavior here each unit needs to scan their corresponding areas to find an enemy. 

If one unit spots an enemy, there should be a delay depending how that unit can transmit that information to platoon leader and from there to everyone. It would require still that everyone participating the task would try to find the location where to look first. That takes time. 

And if someone gets suddenly blown up, it means panic. No information of the enemy and friend just blowed up. It means quickly to get in the cover etc. That is again something that would require first knowing something where the fire is coming from so what direction the cover or concealment needs to be. If that can't be done, it is panic to just get out of there to somewhere that gives concealment everywhere. 

 

These would mean that AI units would need to scan the treelines, scan the expected enemy direction, listen the environment and even have someone looking for the sky once the air warning has been issued from air defense network or someone has spotted air threats. 

All these would slow down the AI units capability to spot and react to helicopters and airplanes and different ways open opportunities to fly and move and engage them without being constantly tracked once inside the unit maximum detection range. 

And this process doesn't need to be done than on units that are close to the enemy units, to save CPU cycles, as not all the time is there hundreds or even tens of units engaging such a way.

[Lurker]

What you are asking for Fri is almost impossible to simulate within the DCS framework. It would bring our PCs to their knees. However some small changes could be made to simulate at least "believable" ground AI to a certain extent. 

[Silver_Dragon]

2 hours ago, Fri13 said:

 

The KA-50 was selected as successor to Mi-24 in first place. It was meant to be built in hundreds, each flight to be lead by a single KA-52. And then transport cargo, troops and vehicles with the Mi-8 as already done almost exclusively. So the Mi-24 cargo space was to be abandoned and idea of "Flying IFV" to be left behind. And, Mi-28 did that all. 

By every means the KA-50 is superior to Mi-24 except acting as "Flying IFV". That is where the primary difference should come that we can carry own squad inside and drop them somewhere and support them, or have them support us with MANPADS. Or we can carry extra load of rockets and missiles in the cargo for one (or two) rearming. 

Like think about flying to safe spot and rearm all 8 ATGM once or twice and get back to fight. But does ED model this is a another question (doubt it).

 

So many ways the Mi-24 is more of a experience and just fitting helicopter for many places. 

 

 

Sorry, but not. Ka-50 first and Ka-52 after has been design as "Heavy Armed Scout helicopter". The "real" succesor of the Mi-24 has the Mi-28 Havoc, as AntiArmour Attack helicopter. The Mi-24 (and variants) has maintain as "Assault Helicopter" on Mi-35 versions.

[Fri13]

11 minutes ago, Silver_Dragon said:

 

Sorry, but not. Ka-50 first and Ka-52 after has been design as "Heavy Armed Scout helicopter". The "real" succesor of the Mi-24 has the Mi-28 Havoc, as AntiArmour Attack helicopter. The Mi-24 (and variants) has maintain as "Assault Helicopter" on Mi-35 versions.

 

The KA-50 was in the early phases planned as replacing the aging Mi-24 fleet. For various reasons it was not driven further for that, hence as well why Mi-28 won competition.

It was surprise to me as well before reading about its history from Mil. 

[tusler]

Fri13 Thanks for the info, I never cease to learn on these forums. Makes sense why the KA50 is built like it is. I am wondering if the Hind is going to be a underpowered acft like the HH-3, or the SH-3 or the MI-8 is, I know it is fast once it gets moving but power wise I really wish someone would make a module of the H-53 series or the CH-47

[Karon]

On 4/6/2021 at 5:30 AM, tusler said:

From what I've read the Hind and the KA50 have the same top speed. I would think the KA50 is more maneuverable. I am really trying to find a reason to buy the Hind as a solo pilot but it seems the biggest reason maybe multi-crew in multiplayer might be the Hind's niche market. I could see it being a bunch of fun if the AI pilot is good, I would like being the gunner...Time will tell.

I will probably have to buy it just to support the module builders anyway!

 

They do different things, as others already said. Also, if you are flying in an early Cold War scenario, you won't have the Ka-50 in the first place. Not to mention the number of Shark/Werewolves produced and that only Russia used it. Depends on what type of experience you are looking for

[Mad_Shell]

1 hour ago, Fri13 said:

 

The LOS calculation should be performed even then. The problem is that AI knows all the time where every unit is moving even if there are trees or buildings between, only the terrain will block the LOS checking and AI will revert back to its previous state. This is problematic as when you perform a pop-up behind a ridge to hover above trees, the AI has already reacted to you on the moment only trees were between you and AI, so on split moment you peek over the trees, the AI will fire at you and you see shells coming on your way in moment you lay eyes on the target. 

In my experience this is simply not true, and there is absolutely no difference between terrain masking and tree masking. AI won't react at all if there is foliage between you and them, and if they spot you but you hide behind trees, the AI will revert to its normal state after a few moments.

[Fri13]

46 minutes ago, Lurker said:

What you are asking for Fri is almost impossible to simulate within the DCS framework. It would bring our PCs to their knees.

It would not. As it is not processed at all by all units all the time. 

That is why we need a one AI that is cheater. We need dozens/hundreds of different AI's to work standalone form, without information of each others. And key component for all is a "Cheater AI" that is responsible to control what AI units gets activated and what processing they can do. 

 

Example you have a one MBT and one helicopter.

The MBT sits in a edge of forest having 180 degree view to south. It has engine turned off and crew is outside observing scene/waiting. 

The Helicopter is approaching from the south flying NOE behind tree lines. 

 

It would be stupid to run all checks once a second (as it does now) for MBT or helicopter that do they spot something, until it is really possible. 

That is where the "cheater AI" comes by, as all units are stupid/simple until the "Cheater AI" activates each unit corresponding AI functions.

So not until the helicopter reach a pre-set "audible" range to another unit, neither AI does anything. When the audible range is reached, is the MBT AI commanded to check that can it hear the helicopter. Check includes "Engine Off", "Crew member outside" and some other values that needs to become TRUE to be able detect the helicopter by sound. If that doesn't happen, no checks are performed further until Cheater AI tells to do it again for some reason. If the detection is made, the MBT unit gets activated some modes. Like crew becomes alerted and focus visual search on the direction of the sound. 

On that moment the MBT AI is just scanning the general direction as a IR seeker or Radar scans in the fighters.

If the spotting happens, again MBT status is changed to combat mode where crew buttons up and now it is commander and gunner only that will try to find and track the helicopter visually. Each time the LOS is broken, search is committed to general direction (like a missile guidance mode) for X period and then returned to general direction etc. 

 

The units on the ground are not so many that would each require perform it that would stress a CPU. We can have even a hundred units to scan a given area as it is about spotting something. And if someone does spot, it is assisting others by various rules to spot the one. Like it is expected that once a one soldier in a squad spots something, he can alarm others and point the threat direction. Similar thing is in the vehicle commanding, one spots something and alarms others and others need to concentrate for search in the general direction to find it. 

 

Even if you have 5000 units, lets say 5 km from the helicopter, they don't do anything because they do not know it is there. Unless the MBT would alarm someone on that area to their direction, like a SAM system or a MANPADS patrol. Again they need to perform visual search or activate Radar and try to search target. Now it is two units instead 5000 that is performing the search, where one is visually tracking target. 

The "Cheater AI" makes the check that can the MBT radio a another party. If it can't, no processing will be initialized. 

 

Majority of the units on the ground are stationary, dummies. They don't move, they don't react or check anything unless they would find something or they are commanded to do something. Their job is to sit somewhere and wait. Even a recon units can be moving across the land without performing any single LOS check or such. Because the "Cheatin AI" is not activating those modes as there is no one near by. 

Once the "Cheating AI" knows that opposite units are closing each other that various spotting methods can be done, it activates them.

Friendly units do not generate such tasks as it would mean that all are running it all the time. And assumption would be that units know how to ID each other visually. There are rules that could be done for that as well but there are many other rules like assumptions that same unit will stick together and in time they know who belongs where. 

 

It is all about cheating. 

In a dynamic campaign the player doesn't care that what kind a clash happens 150 km from player plane. It is totally irrelevant. It is literally just rolling some dice for generating the possib combinations in a real time process, meaning it is simulated by long period of time the gun shot exchanges and wounding, destruction etc. Instead running all in 1-2 second period, it is done in 5-15 minute period. All that extra time is spent for other tasks. Combat is not over in seconds but minutes or hours depending are artificial units made to retreat and reposition themselves by various rules. Like if group A is commanded to defend a bridge, then they are not going to charge after a enemy recon car that is driving 2 km away already and leave the bridge undefended. There is no need to simulate any complex things in majority of the places, that is eased with the inaccuracies in units exact positions, conditions etc.

A lot of things needs to be cheated and kept vague, but it needs to happen in proper manner where one AI is not controlling to units to engage each others, as it is similar case as playing chess alone and purposely try to make other side win. 

 

It is like a RPG where the game host controls what other players can do, change their statuses based their actions and simply limit their options and capabilities how to engage each other but let them to do the play with rules. 

 

The whole gaming industry is about cheating. But it needs to be done clever way so it doesn't become obvious and it requires building rules what everyone can do.

 As dynamic campaign is not one that is over in 1-2 hours. We sit in a cockpit flying in real time, there is not happening much in real time in a large scale conflict. The hard part is to run a simulation in high speed, like 8x or 60x accelerated speed to jump in specific future time. As that is the moment when lots of things needs to be simulated and roll a dice. But this is nothing new in the RTS game side where thousands of units are run this way 20 years ago, each turn to make major changes. Far more stressing than running things in real-time where such simple things as moving unit 100-500 meters can take literally a minute. And one minute for CPU is infinite time to be used for other things to check out. 

 

But right now in DCS all AI units are performing checks constantly. Just having units active on the map is consuming CPU cycles for worthless checks and polling time schedule to do nothing. Why such modes are required as "deactivate group/unit" and so on are required because otherwise you can't run anything if you don't cheat and tell game not to care about such units at all. 

 

 

[Fri13]

27 minutes ago, tusler said:

Fri13 Thanks for the info, I never cease to learn on these forums. Makes sense why the KA50 is built like it is. I am wondering if the Hind is going to be a underpowered acft like the HH-3, or the SH-3 or the MI-8 is, I know it is fast once it gets moving but power wise I really wish someone would make a module of the H-53 series or the CH-47

 

I wouldn't say that Mi-8 is underpowered aircraft, but as anything if you load up them to maximum weight you will have penalties. That was example the Mi-4 problem that when it was in assault configuration you couldn't load any infantry on it. So it was either troop transport or assault helicopter. And Mi-24 was wanted to be different where it can have the infantry squad in and some weapons to perform the assault and support for them.  You can example see here the transport configuration:

 

https://youtu.be/JZ5je96v8H8?t=1748

 

Where at 29:30 it unloads squad and carries 4x rocket pods. It is out of the frame does it have AT missiles with it even. As the ~2500 kg max loading capacity is limiting factor.

But you are not so effective attack helicopter if you are required to land to unload infantry and then get up in air to support them. Why it is better have a more capable Mi-8 (more space etc) do it while Mi-24 defends it or attacks the enemy. 

(btw, person who calls me wrong to say Mi-24 is a multirole attack helicopter, can check the 29:38) 

 

The KA-50 was to be more than a recon helicopter. The one pilot had smaller workload than a multirole single-pilotted fighter (like F/A-18) has. But it is easier when you have a flight commander to designate you the targets (tell what found targets you engage) sitting in a KA-52 next to its pilot. Together you could get a flight like 3x KA-50 and 1x KA-52 where you can attack effectively using all datalinks and digital navigation and targeting systems. Like compare it to Mi-24 where you need to fly in formation, stay in contact with radio and coordinate with the flight as trained to do things. Where KA-50 was to be able separate and position themselves more freely and share their targets to each others. Literally like a swarm of sharks does. The Mi-24 is just old limited design compared to KA-50 or Mi-28. Comparing it to even SA342 Gazelle you don't get anything like that in it. So what you get your or target GPS coordinates, but when you don't see on a digital moving map where others move or where others are aiming, it is limiting factor. Many player knows that Su-27S in DCS is great in single player as you have the datalink for flight, but it doesn't work in multiplayer that you would see here others fly and what they are targeting at. It makes coordination far more difficult, compared to now F/A-18, A-10C and F-16CM where you get the datalink working. 

 

The Mi-24 will be interesting to fly as it doesn't have all that "mumbo jumbo". Like good old paper map with cross showing where you are, and then just visually fly to engage targets.

It is like Su-25A but as in helicopter form and many other capabilities. 

[Fri13]

30 minutes ago, Mad_Shell said:

In my experience this is simply not true, and there is absolutely no difference between terrain masking and tree masking. AI won't react at all if there is foliage between you and them, and if they spot you but you hide behind trees, the AI will revert to its normal state after a few moments.

 

What I mean is that when you have LOS below the trees (as is), the AI that knows all the time where you are (just not allowed to engage you) the moment you slip through that underside of trees, the AI starts to react to you, and if you manage to come up behind trees on that moment, they will engage you. 

You can not fly behind a ground unit to surprise it, as on moment you enter its engagement circle, it will spot you unless you have something between you to brake it. This means that you can not perform a proper scouting (as to see you need to reveal yourself) and pop-up as AI will react to you on moment you become visible. So you can't start a attack run from 4 km by firing couple missiles as AI knows your existence. They are just currently so stupid to do anything if you are outside of their engagement ring even if you would hover just 1 meter further than it. 

 

I should have somewhere a video where I am in a KA-50 hovering behind a tall office building (like 10 meters from it). I know that AH-64 patrol is flying about 3 km from my location and waiting them to appear on any moment the edge of the building to shoot at their flank. Nothing happens. Then I decide that I perform a small pop-up to check where they are. On the moment I come up slowly the building roof I get 30 mm shells on face and dead. The AI Apache pair (only other unit in mission) knew where I was even when I have been all the time behind a building, and not just did it do that, it did know my velocity to time the shells hit me from about 1 km distance before I could even see it. I only saw the Apache in the replay video before shells hit, that it is just visible above the roof. 

 

The only thing I could get was that the hitbox for building LOS is shorter than the 3D model is. And that was reason why it could fire by anticipation couple seconds earlier to kill me on moment I popped up visible (or it saw my top rotor in time).

 

 

[Hiob]

1 hour ago, tusler said:

Fri13 Thanks for the info, I never cease to learn on these forums. Makes sense why the KA50 is built like it is. I am wondering if the Hind is going to be a underpowered acft like the HH-3, or the SH-3 or the MI-8 is, I know it is fast once it gets moving but power wise I really wish someone would make a module of the H-53 series or the CH-47

How, by any stretch of imagination, is the Mi-8 an underpowered aircraft?

[malcheus]

1 hour ago, Hiob said:

How, by any stretch of imagination, is the Mi-8 an underpowered aircraft?

 

I guess it depends on how much cargo you take along.

It's prone to VRS, perhaps that's what he means?

 

As for the AI;

First off, AI is a big word for an object that runs a few if-statements; but in the end it doesn't really matter how the AI comes to a certain action, so long as it's believable from the player perspective. 

The AI being omniscient (which in this case means, it can use all available data for it's if-statements) isn't a problem, so long as it still acts believable.

 

An example of an if statement would be "if you have line of sight: shoot"

Shoot being: Calculate the trajectory of the observed object, and put five rounds there.

 

The AI is made a bit more believable by adding a "wait 2-5 seconds", and "spread the shot a bit".

 

But currently, I don't believe it goes much further than that. 

There's also talk of a spotting mechanic, but I don't know if the AI ever not spots you when in range and in LOS...?

 

In an ideal world, each AI has a viewing angle, and is scanning it's surroundings, had a chance of spotting objects within the viewing angle (and a chance of not spotting things in it's viewing angle based on distance and other variables), etc.

But this is way too complicated to simulate for possibly hundreds of objects at a time.

 

Also the player is completely unaware of such mechanics, so it can be simplified with a simple "spotting chance per second when LOS exists".

You can calculate this from the player instead of from the AI's as well (LOS goes both ways), so you only have to do the check for AIs within a certain radius and with LOS.

Computationally, this should be doable. 

 

It would also be a bit more believable (and enjoyable) if the first shot wouldn't be on target every time; so why not make the AI more accurate with consecutive shots until LOS is broken? 

These should all be relatively simple tasks for an AI (add a diminishing random number to the angle and elevation of each shot), but it would make flying helos  much more enjoyable. 

 

 

 

 

 

 

[CrazyGman]

6 minutes ago, malcheus said:

 

I guess it depends on how much cargo you take along.

That's pretty much the case for any helicopter.

 

I think people are going to be surprised the limitations the AH-64D is going to have with the cheese wheel, eight hellfires, 2 rocket pods and full fuel and ammo. (I'll be surprised if it's not beyond max take-off wieght with all that)

 

The Huey though compared to the Mi-8 is way underpowered.

Edited April 8, 2021 by CrazyGman

[Mars Exulte]

17 minutes ago, CrazyGman said:

That's pretty much the case for any helicopter.

 

I think people are going to be surprised the limitations the AH-64D is going to have with the cheese wheel, eight hellfires, 2 rocket pods and full fuel and ammo. (I'll be surprised if it's not beyond max take-off wieght with all that)

 

The Huey though compared to the Mi-8 is way underpowered.

 

 

 

  Yeah, everytime I see somebody complaining about something being underpowered they invariably fly around like this :

 

actually I can't find the picture now, it's a photoshopped F-18 with TERs and Mk82s on every hardpoint.

 

  Most people seem to not understand MTOW literally means ''more than this and it can't fly''. They load to MTOW (or a little beyond) then seem shocked they have trouble @@

[malcheus]

The biggest problem for me as a player is the tendency to go into VRS, which feels a lot like not having enough power. 

The huey doesn't have this tendency.

 

I think it's so obvious in the mi8 because it has to land with cargo. It won't be as much of a problem with attack helos, because you will very rarely have to land with all the weight you took off with. 

[SparxOne]

8 hours ago, Fri13 said:

Last time I read about it, the plan was WW2 -> Ground Vehicles -> Modern aircraft (inc. helicopters). 

 

Now that you mention this it reminds me of a little subtle thing, havn't done any research on this so can't be 100% sure, but, from what i vaguely remember, that was meant only for the WWII ground vehicles as that seamed to be a pretty much logical step to have a complete experience in the WWII aspect of the game. 

So i believe that their plan was to develop the damage model for WWII planes and ground vehicles than modern aircrafts and lastly modern (everything after WWII) ground vehciles.

 

If i'm right on this, it would mean that my initial statement still stands, we are on a long wait before it happens.

[Fri13]

5 minutes ago, SparxOne said:

 

Now that you mention this it reminds me of a little subtle thing, havn't done any research on this so can't be 100% sure, but, from what i vaguely remember, that was meant only for the WWII ground vehicles as that seamed to be a pretty much logical step to have a complete experience in the WWII aspect of the game. 

So i believe that their plan was to develop the damage model for WWII planes and ground vehicles than modern aircrafts and lastly modern (everything after WWII) ground vehciles.

 

If i'm right on this, it would mean that my initial statement still stands, we are on a long wait before it happens.

 

It was for all ground vehicles, not just WW2 ones. As ED is developing as well a new fragmentation system. Chicz just mentioned about it hour ago.

 

 

 

That includes as well all missiles and everything that explodes and cause fragmentations (so WW2 flak cannons as well etc).

Just adding a damage values and effects for various parts of the vehicles is not enough if the cause for damage can't be calculated (fragmentations). 

 

This is as well about all bombs, rockets, artillery shells etc. As it is about A2A missiles. 

 

 

[Dragon1-1]

Ground vehicles today have the same basic components as WWII ones (engine, transmission, fuel tank, gun, gunsight, wheels), so a DM for WWII vehicles would also translate to modern ones. 

[Fri13]

2 hours ago, malcheus said:

The AI is made a bit more believable by adding a "wait 2-5 seconds", and "spread the shot a bit".

 

If the "spread a bit" would be "a bit" and not "Shoot 10 degree after the target where it was".

When the first WW2 damage modelings were shown, it was mentioned by ED that they needed to improve that shooting accuracy as the AI always aimed at the pilot and killed the pilot on each time. So they made the WW2 pilots to shoot and then correct their aim based their shells trajectory until they started to hit the target. 

 

2 hours ago, malcheus said:

But currently, I don't believe it goes much further than that. 

There's also talk of a spotting mechanic, but I don't know if the AI ever not spots you when in range and in LOS...?

 

AI does always spot. The only factor there is the random timer from entering to LOS and reaction. That is tied to the difficulty level. There was a one forum member that made a good test of the AI spotting timings and it had like 10 units in line on group and basically even at lowest skill level the AI reacted right away but some were delayed by a second or two. It was funny to see how the turrets started to point when he popped up behind from hill at their flank. 

 

2 hours ago, malcheus said:

In an ideal world, each AI has a viewing angle, and is scanning it's surroundings, had a chance of spotting objects within the viewing angle (and a chance of not spotting things in it's viewing angle based on distance and other variables), etc.

But this is way too complicated to simulate for possibly hundreds of objects at a time.

 

It is not too complicated as it requires to be done by cheating. If we put 100 units on same side on same small area, there is no reason what so ever to run any single LOS checking or scanning surroundings when there is no enemies. It is waste of processing time to even try. Now if we suddenly spawn a enemy unit middle of the group, that is very unrealistic behavior to happen, but it would require just a few units to really react to it and then alarm the others. So just few searches and then all to know what is going on as it would be so obvious thing. 

 

And we are talking about real-time from player perspective, there is no requirement to calculate everything in real-time as computing perspective. Meaning that 100 units to perform a LOS check is not required to happen in X milliseconds, but it can be done in 5-10 second period. We need to add to AI as well the human factor that it does not constantly scan the places but gets lazy. Why the ALARM STATE is important factor that unit can stay as such about 30 minutes and then the focus will drop. 

 

2 hours ago, malcheus said:

Also the player is completely unaware of such mechanics, so it can be simplified with a simple "spotting chance per second when LOS exists".

 

It is currently once per second. So every second the AI rolls a dice that is a unit X inside unit Y spotting range. If that happens, then DO.  Every second wasting processing for tasks that AI shouldn't be doing. Place a 1000 units on the island 500 km away from player and CPU dies for the processing. Make them shoot at point on ground and your CPU is melted.  

 

2 hours ago, malcheus said:

You can calculate this from the player instead of from the AI's as well (LOS goes both ways), so you only have to do the check for AIs within a certain radius and with LOS.

Computationally, this should be doable. 

 

Totally doable. It was already a couple decades ago, with single CPU. Now we have 6-16 cores (or 64 cores with Threadripper) and we would very well have plenty of sharing by allocating different parts of simulator to different cores.

 

1. Player aircraft systems and flight modeling.

2. Blue side AI aircraft flight modeling.

3. RED side AI aircraft flight modeling.

4. AI pathfinding and LOS

5. Missile/Ammunition trajectory calculation.

6. Fragmentation / damage modeling calculations.

 

We could very well combine multiple calculations for own sub-simulations that doesn't need to work in sync with others.

 

2 hours ago, malcheus said:

It would also be a bit more believable (and enjoyable) if the first shot wouldn't be on target every time; so why not make the AI more accurate with consecutive shots until LOS is broken?

 

That might happen if it is up to the AI that is being developed for Mi-24. As it has interesting spotting calculations. And such should be done for AI trying to shoot at you. 

Like if you have a Mi-24 flying toward you from about 1000 meters distance. Why would you shoot 5 degree away from it? You would shoot little above it or straight at it, but not below it. 

If it is flying around you, you wouldn't be aiming behind it or directly at it. You would be aiming to predicted flight path front of it. Trying to estimate a lead and get shots at it.

 

 

2 hours ago, malcheus said:

These should all be relatively simple tasks for an AI (add a diminishing random number to the angle and elevation of each shot), but it would make flying helos  much more enjoyable. 

 

Totally would. Just simple current AI adjustments with the aiming time, the aiming point, randomizing aiming around that aiming point and make the AI actually shoot more proper bursts than full auto for 4 seconds each time etc. 

 

Simply having a AI to have blind zones (smaller scan sectors) would make amazing results. 

[Fri13]

21 minutes ago, Dragon1-1 said:

Ground vehicles today have the same basic components as WWII ones (engine, transmission, fuel tank, gun, gunsight, wheels), so a DM for WWII vehicles would also translate to modern ones. 

 

A BTR-60 is not so much different from a German half-track SdKfz 251. Little bit more lights and so on, but otherwise as complex. 

Then 2K22 Tunguska.... It has a lot more wires and hydraulics and coolant systems than either one. Same as with M1A1 Abrams. 

But how detailed we really require it to be on ground units?

 

I see that we could forgive a lot about minimizing the damage zones and parts inside to dozen or two, instead hundreds like in a WW2 aircraft. 

Like how much does it matter if a 30 mm API round penetrates inside BTR-60 engine compartment that is it a fuel pipe leak or is it a coolant leak or oil leak? Those could be randomized really.

"30 mm API hit the engine" Roll a die of 6 eye for:

 

1. Nothing

2. Coolant / Oil Leak

3. Fuel Leak

4. Set engine on Fire 

5. Kill the engine

6. Kill the engine

 

Wouldn't that be enough?

Make the 14.5 mm HMG turret as one element, hit at it and render the weapon unusable and kill one squad member inside the vehicle who was manning the turret. 

Sure it would be nice to have "Optics has been destroyed" but it is just one statics more about some bombs or other fragments. Where 30 mm API hitting that turret will cause serious spalling and injure or kill multiple person from the compartment. So throw a dice that how many dies from a 8+2 soldiers inside.

 

A S-8KO rocket hit the center of vehicle = kill 6 and injure 2.... 

A 12. 7 KORD gunner shot three tires from the right side = immobilize the vehicle with 3/4 blown tires. 

 

WARNING ABOUT AUDIO!

 

 

 

Edit: Seriously anyone should question a moment that what does anyone see outside from a military vehicles? We have a infantry squad leader perfectly tracking a fast mover above him with the turret sight (the infantry squad leader was leader of the vehicle, the vehicle commander took control of the vehicle when the infantry squad leader unmounted from it) and as well that if a helicopter pop-up suddenly any direction (than from front) the squad leader knows exactly when and from what direction and from what range... 

 

We have a ground units like LAV-25 that shoots more accurately than a AAA units. And that variant doesn't even have a laser range finder (installed in LAV-25A2 upgrade to be used against air threats like hovering helicopter). 

 

Because all the ground units capabilities to engage helicopters way too quickly, accurately and so on. It makes all helicopters just very unsuitable for DCS ground combat inside those units engagement ranges. 

Edited April 8, 2021 by Fri13

[Dragon1-1]

1 hour ago, Fri13 said:

Then 2K22 Tunguska.... It has a lot more wires and hydraulics and coolant systems than either one. Same as with M1A1 Abrams. 

Not really. All this can be reduced to generic "turret actuators". Which, BTW, WWII vehicles also have. Tunguska has a radar, which is an important and easily damaged part, but that's it. It would, for example, be very unlikely that a hit which would take out Tunguska's firing computers wouldn't also wound/kill the gunner. It would make sense to put in more detail in something like ArmA, but not here, where the gunner is only another part of the damage model, not a separate entity.

[Fri13]

36 minutes ago, Dragon1-1 said:

Not really. All this can be reduced to generic "turret actuators". Which, BTW, WWII vehicles also have.

 

Sure there are (hydraulic or electronic, some even mechanic) but those would more more of a parts of bigger elements (batteries, engine, hydraulic system overall).

 

36 minutes ago, Dragon1-1 said:

Tunguska has a radar, which is an important and easily damaged part, but that's it. It would, for example, be very unlikely that a hit which would take out Tunguska's firing computers wouldn't also wound/kill the gunner. It would make sense to put in more detail in something like ArmA, but not here, where the gunner is only another part of the damage model, not a separate entity.

 

IMHO we should have crew separated from the vehicle as damage model. Same way as a transported infantry in helicopter is separate element. Like carry a infantry squad inside Mi-24 and get under fire where bullets hits the cargo compartment. Injure/kill the soldiers in the path, but don't the pilot or co-pilot. 

Same thing is with the APC units, you can have infantry squad survive from the attack but damage the vehicle. Or you can kill the personnel and leave vehicle operational.

 

As the human being is critical part of the damage modeling and the AI logic. The human being is fearful, trying to survive. If a APC suddenly is under fire and none of the hits penetrate armor or cause any spalling, the APC will totally be driving out of fire as at any moment there can come something bigger.  If it is a attack process where troops needs to get transported over open area and they are under artillery fire etc, then better push forward than stop there or try to get around and retreat. 

 

This is what makes a Mi-24 more effective as you could threaten or control the ground units with rockets or cannon without destroying or damaging the vehicle itself.

Like if a APC is near edge of forest, couple rockets next to it and it will reverse back to cover deeper inside forest, opening a possibility for own units to attack.

If a APC is moving on the road, short burst front of it should get it to drive in the forest for cover even if you don't hit it and damage it.

 

In a 2K22 Tunguska you have computers on one side of compartment and you can penetrate armor or cause spalling that will destroy the systems, without killing the operator. Or you can penetrate it cleanly and kill operator but leave the systems intact (even if unlikely, it is still possibility), that would leave operations like recover unit and recrew it by transporting new crew (like with many vehicles, after cleaning the mess). 

This doesn't mean that we need to have every single element modeled inside a vehicle, but some important ones would be good, like main hydraulic or electronic but not everything. 

 

It is important in these to maintain the crew and vehicle as separate entities as it goes through everything. You can scare the infantry, you can render them unconscious, cause them get in shock etc. Various weapons and elements do that. A close high explosion will not be easy to recover when it came just out of blue. Why we need the simple soldiers that will unmount/mount vehicles by AI logic. 

 

In a single seated aircraft pilot is critical element, kill it and aircraft comes down. Injure it, he will likely eject. Get the pilot eject and plane is trashed.

There it is easy 1/0 element. But when it becomes a two pilot vehicle like Mi-24, where you can kill the pilot but co-pilot survives and flies back to base or lands on near by safe location.

Similar thing is with the infantry squad inside, injure many of them and helicopter will abort their attack (bur will continue back to base) and land on near by safe location to help wounded, depending threat and area. 

 

A MBT can have driver killed so it doesn't anymore move, but it can still shoot back and defend itself by popping smoke etc. Commander or gunner can switch places and in short time get vehicle moving to safe but lose capability shoot or spot targets. 

 

These things sounds "too complex" but they really ain't. There are old games utilizing these features like Theatre of War or Men of War.

The Combined Arms would be excellent to provide the RTS elements to the DCS World. As there are more RTS players than there are virtual pilots. 

And having capability to have players concentrated to small area (lets say 20 x 20 km) combat where they can command the ground forces and generate support calls for the aircraft, that can be either AI - or the players. DCS World could very well handle a 20-30 players at the time, commanding small groups of units across various places without turning them exactly so deep as mentioned above ones. 

We do not never get any FPS elements in the game other than maybe a MANPADS and ATGM operator outside the cockpit. That is it. All the ideas of the ARMA is just out of the scope of the DCS World. So closest really there will be is sitting in a cockpit and seeing that infantry loads and unloads from the helicopter/vehicle and that is it. 

But such element would require such minor details as separate vehicles and infantry, and their effects. 

 

The DCS World already has the maps. It has the theaters. It has the themes and most important units. It just needs reconstruction. 

Like it would be amazing experience to fly a Mi-24 while a friend is in Combined Arms and playing the ground units. It just isn't great yet because these AI limitations, but there is huge potential. When the ground combat turns to more realistic and slow paced action, there is more time to fly around the combat areas and actually participate to it.

As right now we can have one pilot take-off from base, launch few missiles and come back with a 4-6 MBT's destroyed. Not a real challenge.

 

And Mi-24 is more about gun and rockets than it is about missiles. It is about first finding the enemy, then find a way to get to them and so on. 

One can always pretend great success in Mi-24 by placing group of M16 guys on the open and few APC next to them and just slaughter them on the spot. 

 

[Dragon1-1]

There definitely is such a thing as "too complex", but not with regards to gameplay, but performance. Every single thing you simulate will take CPU cycles. This is something that DCS already uses a lot of, and it also doesn't increase exponentially anymore, indeed, single core performance has lately stagnated. Given the scale, DCS missions involve a lot of units, far more than you'd usually see in Theater of War. This means everything that you do, for a single unit, multiplies in impact the more units are on the field. On top of that, in multiplayer, aggressive optimization is often not possible.

 

Above all, the new DM shouldn't bog down the sim much more than the current one does. Same goes for ground vehicle AI. Complex crew AI is nice, but if you've got three "brains" per tank, as opposed to one, you've got three times more impact on the CPU. As such, I'd prefer to treat the crewmembers as subsystems of the vehicle. Every update to AI that is made needs to be coded with an eye on performance, because when you've got enough area to potentially support brigade level operations, chances are there will be scenarios involving precisely that. If you want Combined Arms to be a good large-scale RTS, you can't get bogged down in ArmA-level individual soldier simulation.

[tusler]

Fri13, I imagined (incorrectly) that since the Mi-8 frequently looses rotor speed to the point of dropping generators when you grab collective that it was a function of reduced power compared to other heli's built around the same era but after viewing the spec's I see it is more of a difference in how the acft are constructed and the mechanics operate than a power issue. U.S . puts generators and pumps on accessory gearboxes instead of main trans on large cargo birds, I'm just use to H-53's where unless extremely loaded heavy there was always an abundance of power. I see the MI-8 for it's time back in 1967 had 2,400 shaft horse power engine's which was far more powerful than HH3 or Sh3. I stand corrected.

Edited April 8, 2021 by tusler