47 replies to this topic

[Koniving]

The performance gain of turning an LRM 20 into a 4 missile system...would be incredibly null and void. The difference is pretty insignificant and there isn't much draw on the system for that because the missiles don't actually have an intelligence just a destination that's either stationary or moving.

The issue with the infantry, is the human models are nearly as many to more polygons than a number of the mechs themselves. The gun, while definitely not anything I've seen in the BT universe, is also very excessively detailed which is ideal for something you can see up close but not ideal for something you want to mass produce on screen.

Now to field enough infantry to be remotely convincing.... and then, to have an AI just as "complex" as that for the mechs for each individual person and it is no wonder why they had performance issues.

They need a cheap way to render numerous people, and a cheap way to do the AI.


Fraction of the polygons actually used on the people models (in fact it's less than 100). Can actually be reduced further I'm sure. A decent texture that's repeated and a hidden or obscure face can easily allow the same "person" to be put on the field many times without additional resources for unique textures.

While the individual movement is nice, robbing them of being independent units and control by making several of them move as though a squad but as a single as a single actor (or as Unreal Engine 4 calls them, "Decorators") would significantly cut on the resources they would use in terms of AI.

The smallest group I suggest, 2, could act like these guys from the third Matrix movie.

Now the two will always have a reason to stick together, one has the Heavy SRM launcher and the other is the loader. Likewise they can die together.

A team of 5 can have a squad leader and a squad, one of which would have a weapon more meant for use on vehicles while the others can spam auto-rifle or MG fire. Another setup of 5 could be carrying the parts for a portable support MG (1,200 meter range against other infantry, 90 meter range against mechs) or portable support SPL (1,500 meter range against other infantry, 90 meter range against mechs) or portable SL (also 1,500 meter range against other infantry, 90 meter range against mechs).


Despite the fact that you see 7 insects, this is one monster in FF7. A single entity represented by a bunch of fake individuals. Its the same concept. One entity, appearing as a fake "5 soldiers."

They can also be mechanized infantry, in small vehicles such as SRM-equipped motorcycles or rocket trucks, which travel in small groups (and are actually a single unit and not multiple units). This not only simplifies the CPU load by only having one AI per "group," but the 3D load is reduced significantly too because the whole group should never take more polygons than to render a single 'detailed" person or combat vehicle allowing 50+ "people" to be rendered for the polygonal price of 5 Manticores (probably for significantly less depending on how many polygons are in MW5's Manticores, I'm fairly certain it's more than 500).

The challenge this presents is that PGI would have to actually have a decent animator, as the resulting abomination would actually have 5 skeletons to deal with and animate. That's quite an undertaking and the ultimate reason I hold little faith in PGI being able to pull it off, given that their animations are lackluster at best and there'd be no way to "cheap" their way out of doing them from scratch. Even more difficult to hide the fact that they're just tied together.

Now, a simpler method is they could still have individuals, but use something like an RTS setup, where the independent decorators simply have the most basic face location, face target, aim/attack AI with animation calls, and then another AI can be given control of them, giving movement and attack orders. But this, in turn, would require them to create multiple layers to the AI that can handle using their environment without pre-scripted markers on the levels. This puts the actual challenge on the person in charge of the AI, which is another area that tells me that regardless of what that specific Dev that Psuedo mentioned would love to see done... I don't think PGI will be able to do it.

Edited by Koniving, 14 December 2019 - 12:16 AM.

[Karl Streiger]

Koniving, on 14 December 2019 - 12:15 AM, said:

The performance gain of turning an LRM 20 into a 4 missile system...would be incredibly null and void. The difference is pretty insignificant and there isn't much draw on the system for that because the missiles don't actually have an intelligence just a destination that's either stationary or moving.

Depends on the hit-reg mechanic you use. Think of the average number of LURMs flying through the air in mwo, now reduce tge number of necessary calculations by factor 5, that is something.
For PvE and Client Hit Reg, you are correct however.

And yes, RTS detail level of infantry using clever textures and maps instead of polygons should not be that expensive either.
Minimum squad size of 6-7 simple for the reason that each squad has one or two weapons that can damage a mech. You also really should not consider individual movement and hitboxes, a gauss slug in the direction of a squad should be sufficient to maim one or two soldier's

[Koniving]

Karl Streiger, on 14 December 2019 - 12:08 PM, said:

Depends on the hit-reg mechanic you use. Think of the average number of LURMs flying through the air in mwo, now reduce tge number of necessary calculations by factor 5, that is something.
For PvE and Client Hit Reg, you are correct however.

And yes, RTS detail level of infantry using clever textures and maps instead of polygons should not be that expensive either.
Minimum squad size of 6-7 simple for the reason that each squad has one or two weapons that can damage a mech. You also really should not consider individual movement and hitboxes, a gauss slug in the direction of a squad should be sufficient to maim one or two soldier's

Won't matter how well pixel art is done, or what quality pixel art has (and lets be honest it's PGI so it would look like pixel art), 2D Doom-style infantry would stand out like a sore thumb when you consider that beyond actually using the mech people can CTRL Shift V and fly around with a camera to take pretty pictures.

Everyone's gonna notice the "hi-res doom marines."


And we won't be saving that much on performance by having sprites, and well designed maps... we're talking PGI and a map generator..

And if you go RTS style..

It's still gonna be pretty noticeable. Perhaps more so.

[Koniving]

MW5's mainly PVE, so the benefits are pretty much non-existent unless you're playing with friends. And it's not the multiplayer performance issue that comes up, it's the single player.

The bigger issue is the AI than the polygons. It's significantly less effort to simply use a 100-300 polygon mesh and animate it.
Each AI combat AI in MW5 right now uses a minimum of 7 trees.
Mechs use 39 trees. (Edit: Ally mechs use 39 trees. Enemy mechs use 24.)

Humans currently have 12 trees related to movement and the like, 11 of which are tenable to combat though several are worker-related in combat conditions. if they can set up support weapons then you have about 3 more. If they can swarm the player or try to climb inside, then you have perhaps 4 or 5 more..
Then you add this by the individual level and put 50 people on the screen running these trees.
You now have 950 individual trees floating around bouncing between black boards, checking on circumstances, finding locations to go to, targets to acquire, barks to be called... That's just the infantry.

Currently, I've never seen more than 6 enemy mechs on the screen at once, that's 144 individual trees. Combine with some 16-odd vehicles for an additional 112 for a total of 256 behavior trees. Then you add in your 3 teammates whom actually have 117 and now you have 373 trees.

Now granted, Unreal trees are event driven so they only check when events occur, but these trees are open for each AI, waiting for events to occur, jumping between blackboard references and tree calls to decide what tree should be active and when for every individual soldier.

It's said online that Unreal 4 begins to bog down when there's about 400-500 behavior trees on the field at the same time.
On maps as large as MW5's that is probably reduced even further compared to normal first person shooter maps.

With the fake individual setup I described, we could have 50 soldiers on the screen with only 95 trees on the field.
With the RTS style setup, we could have 50 on the field, with at most 150 individual trees, and I'm thinking no more than 50 command trees as a max for the most insanely complex command AI imaginable. So 200 trees tops (but likely significantly less, like in the 174 range).

Combine the fake individual soldiers (squad units) with the predicted trees and you have 468 trees.
Combine the RTS style two-tier AI and you have 547 to 573 trees.

Significantly more tenable, but dangerously in the bogged down performance range regardless.

Edited by Koniving, 14 December 2019 - 12:51 PM.

[Koniving]

Update 12/14/19
Began listing completed (as in completely finished, needs only artwork if even that) BT weapon concepts (straight list). Filled out Autocannons and missile ammo types.

The list will not include weapons still being worked on, so there are STILL yet more autocannons being made. In a later update I'll also include notes like rarity, or year introduced/extinct if they do not yet exist or will soon cease to exist (once that year passes, you will never find another one again, most likely.)

They are above the shared fluff versions and are found here.

I was lazy with listing missile ammo types a second time, so under ammo types I threw in a link to another post that lists
all the ammo types that are within the timeline in whatever possible capacity along with what they are, what they do, and if they don't have a reason to not include them they will be brought in at some point. (There's like a dozen ammo types in addition to generic the missile type differences).
Same post also lists all the planned HUD and weapon mechanic changes. (Don't get your hopes up I'm not redesigning it; leave that to someone else, just tweaking how it works so that cool down bars are now weapon heat-buildup bars.)

Edited by Koniving, 14 December 2019 - 06:12 PM.

[Azhrael]

I dont know if it was suggested, but plz get rid of the SIZE on the weapons slots. Leave it as MW4:


- 1 Missile Hardpoint 4slots big.

Instead of the ****** :

- 1 Large MIssile Slot.


I cant possibly understand not being able to rig 2xLRM5 instead of a single LRM10 in a hardpoint.

PD : Cant believe we are pooling our hopes on MODS the very first day after launch....depressing.

PD2 : Tnx for whatever you mod guys can do for us in advance.

[Koniving]

Azhrael, on 14 December 2019 - 10:21 PM, said:

I dont know if it was suggested, but plz get rid of the SIZE on the weapons slots. Leave it as MW4:


- 1 Missile Hardpoint 4slots big.

Instead of the ****** :

- 1 Large MIssile Slot.


I cant possibly understand not being able to rig 2xLRM5 instead of a single LRM10 in a hardpoint.

PD : Cant believe we are pooling our hopes on MODS the very first day after launch....depressing.

PD2 : Tnx for whatever you mod guys can do for us in advance.

Will have to see exactly how it works once they unlock it.
Had a breakdown in 2015 for splitting hardpoints, so for a specific size slot you could squeeze in 2 smaller weapons. or combine two slots for a bigger weapon. The problem is that the slot system is likely coded in such a way that it'd take genuine scripting skills to redo it.

It's likely going to be easier to remove the "hard" part of the hardpoint system, i.e. "Only ballistics here" and go "Welp it's a ballistic now, but feel like using missiles or energy today?" and then tie something to keep the immersion/realism for changing it.
And for mechs where there's basically no variation between the variants because PGI didn't implement actual different weapon variants to the mechs, well two-fold there's the weapon variants I'll have, and I can also do things like put in additional hardpoints to work with or put in additional variants where instead of a large missile hardpoint it might have 4 smalls, or 2 mediums, etc.

Somehow I can't help but imagine the Catapult when you mentioned the large hardpoint for a missile. Catapult A1 is definitely unfairly punished, it doesn't even have the different sized ears..
On left, A1, on right, C4.
But PGI is cruel to us...
We don't even get to have doors.
(Edit: Grabbed the "reshoot" after PGI screwed us rather than the original. Got the original up now.

If it's coded simply, I could do something simple like 1 large slot could fit 2 mediums or 3 / 4 smalls.
The issue with fititng 4 LRM-5s for example, is you've replaced the extra LRM-20 for a faster firing, lighter LRM-20 made out of 4 separate launchers when thinking of vanilla launchers.

Given the 60+ LRM launchers I've got on a list to make it wouldn't be that straight forward, but considering an LRM-20 is 10 tons, generates decent heat, fires at a certain rate and then 4 LRM-5s combined is 8 tons, generates considerably less heat per volley but also fires more frequently, you'd eliminate the very reason for LRM-20s to exist with a straight 4 smalls for a large.

Edited by Koniving, 14 December 2019 - 10:48 PM.

[Azhrael]

I just noticed these guys on my game :




You see them? Walking around the mechs? The animations are pretty basic, but they DO look like complete " People " models, moving back and forth. Maybe these could be useful for the "Infantry" concept?

Just a half cent.