Do you want Laaagdoll Physics in ArmA 3?

[neokika 62]

Hi Fireball

How many players in Left For Dead?

Kind Regards walker

Hello,

From 4 to 8 but during periods of 20-30 seconds there are 30+ full rag-doll animated and synchronized entities with no impact on fps/bandwidth.

Not sure what you trying to say?

_neo_

[walker 0]

Really? Also when shot with lets say...a 105 mm howitzer. Or when hit by a 30 mm canon round from a APC? Even when shot with the powerfull 50 caliber sniper round a limb could fling off and fly a couple of yards. Modern warfare is all about physics mate

Hi CyclonicTuna

There is a video in that quoted post if you follow it to source you will see addional videos with a 50 calibre shooting a pig; I suggest you watch it. It shows the effect on human analoge, in this case a pig, of 50 caliber sniper; then rexamine your last statment. By the way 30 mm canon and large HE explosions already make you fly around in the RV engine and have done since BIS's OFP.

To neokika

The 30 AI and 8 clients of left for dead is meagre. 30 entities is just an average number of clients for ArmA never mind AI.

Entity counts in ArmA run to many hundreds even thousands. And player client counts run to 100 plus. Client load expands netcode load exponentialy. It always amazes me that people Dis BIS's net code when they can not show a single other game that deals with a 10th of the entities and clients that ArmA does.

Kind Regards walker

Edited May 25, 2011 by walker

[dmarkwick 261]

It is that I understand the computational and programming cost of Ragdoll.

My impression from reading your posts on ragdoll is that you are only considering a 100% sync implementation of ragdoll, down to fingers and hands. Ragdoll is not a single solution, rather it is a generic term for freeform jointed animation of entities that have no other animation. It can be dialled up or down, and implemented in many different ways, in the case of ArmA3 I suggest that only the body world-space position is important.

The vast majority of ingame ragdoll animations will be the simple falling down of a unit where it is. As such a ragdoll system will generate a unique pose on each client machine, and there will be no practical repercussions. People trying to use bodies as cover will be subject to the vagaries of the hit system on multiple poses, however they'll also be subject to the vagaries of the stopping benifits of corpses, which will be similarly ambiguous. No pragmatic difference between the two, I suggest it's a non-issue.

For cases where the unit does NOT fall down where it stands, lets say down an incline or a stair, there may be an amount of dead body warping as the server sends out a final resting position that may be different (although not too different) from what the client just witnessed. I believe such an update to be worth the cost, seeing as it will be by far in the tiny minority of ragdoll cases.

I believe the above approach to be well within the abilities of the current BIS engine limits, with PhysX applied.

---------- Post added at 12:49 PM ---------- Previous post was at 12:44 PM ----------

Entity counts in ArmA run to many hundreds even thousands. And player client counts run to 100 plus. Client load expands netcode load exponentialy. It always amazes me that people Dis BIS's net code when they can not show a single other game that deals with a 10th of the entities and clients that ArmA does.

I will concede there will be an overhead for remembering each corpses' pose. But I think it would be minimal and well within the bounds of increasing the engine's abilities with increasing PC abilities. This info does not need to be transmitted across networks, it's all done clientside. Only the body positions need to be synced, and that already happens.

Ragdoll outside the player's regard does not need to be done in real time either - no-one will care if it takes a body 15 seconds to fall down 10 km away.

[Leon86 13]

oh come on, ragdoll deaths are in rainbow six raven shield. Calculating the things clientside will not cause problems. Sure there might be a minute difference in some floating point calculation once in a while but that wont cause an outcome to be noticably different, most of the time the outcome will be exactly the same. It's not like floating point calculations are random or something.

[walker 0]

...It's not like floating point calculations are random or something.

Hi Leon86

They are.

Please Read

http://gafferongames.com/networking-for-game-programmers/floating-point-determinism/

Read

http://www.gamedev.net/topic/567626-deterministic-physics-replay/

Kind Regards walker

Edited May 25, 2011 by walker

[Leon86 13]

Floating point operations are accurate enough. There is no way a 3 second death animation would have a significantly different ending on 2 different machines. You're seeing problems that just arent there for something as small-scale as this.

Edited May 25, 2011 by Leon86

[dmarkwick 261]

Floating point vagaries are a non-issue. Only the body position is important.

[AlexVestin 24]

Walker, "Laaagdoll Physics" is an improvement. If the engine can't handle it somehow, I'm sure BIS wont include it.

They know what they're doing. Don't you think so too? =)

[NeMeSiS 11]

Walker, "Laaagdoll Physics" is an improvement. If the engine can't handle it somehow, I'm sure BIS wont include it.

They know what they're doing. Don't you think so too? =)

Well, all we have for now is a picture with a dead guy in it. That doesnt really tell is wether it is in or not, could just as well be an animation.

[AlexVestin 24]

That doesnt really tell is wether it is in or not, could just as well be an animation.

I'm fully aware of it not being confirmed. But it's not out of the question yet.

[NeMeSiS 11]

I'm fully aware of it not being confirmed. But it's not out of the question yet.

I understand, i just wanted to point it out in case its not in and everyone will go like "BUT BI PROMIIIIIIIISED!" again. :p

[Mr. Charles 22]

That animation seems kinda strange for an animated death. But i will keep my mouth shut til E3 :)

[lethal 0]

Hi Leon86

They are.

Please Read

http://gafferongames.com/networking-for-game-programmers/floating-point-determinism/

Read

http://www.gamedev.net/topic/567626-deterministic-physics-replay/

Kind Regards walker

That's only half true!

With a bit of extra work you can have deterministic fp operations (and thus physics) across different platforms as long as they adhere to the IEEE754 standard as Elijah kindly explained in his posts here, here, here and here provided you - as he puts it - go ninja on the code :cool:.

he does indeed mention that none of the physics apis satisfy this strict definition of determinism which is why they (GPG) went head and did their own implementation; so i'm curious as to how BIs implementation of the physX api works provided they even rely on determinism as a minute difference in calculation on the physics won't have a noticable impact imo and shouldn't propagate and escalate within the simulation because of its limited lifetime...

other games come to mind as well that seem to rely on determinism because of the number of states that would need to be synchronized... sins of a solar empire, the total war series to name two.

Edited May 25, 2011 by lethal

[walker 0]

Hi all

The problem with the non deterministic nature of floating point calculations is something called sensetive dependence on initial conditions, this is what can cause wildly different answers off a small variation in one floating point calculation.

You also have the problem of irational numbers, where the number is not a fraction and where it cannot be expressed accurately with floating point notation. In fact for most irational numbers, and there are an infinity of them, we just apoximate.

√2 is an interesting example try it on your windows calculator.

I get the answer: 1.4142135623730950488016887242097

Which I then multiply by itself eg I square it and

I get the answer: 1.999999999999999999999999999999

Once again I got the square root then squared it; logically the result should be 2 but because √2 is an irrational number and cannot be calculated exactly it will always be an aproximation.

And such irrational numbers are as I said scattered throughout maths meaning that.

ALL PHYSICS CALCULATIONS ARE INHERENTLY APROXIMATIONS!

When you accept and understand this then you can start moving on and understanding what is doable in a simulation.

Kind Regards walker

[dmarkwick 261]

None of that matters. It's all non-sequitur.

[froggyluv 2130]

I'm worried aboout you Walker. What will you do if laaagdoll is indeed in the next version of Arma?

Will that be it for you? Will 3+3 suddenly equal 9? Will dogs and cats start living together, ya know, total chaos.....armageddon :eek:

[lethal 0]

true but that has nothing to do with determinism but is basic computer science ;)

as long as you approximate the same way each time with the same initial input the result is always the same and thats the only thing that matters.

[dmarkwick 261]

I'm worried aboout you Walker. What will you do if laaagdoll is indeed in the next version of Arma?

Will that be it for you? Will 3+3 suddenly equal 9? Will dogs and cats start living together, ya know, total chaos.....armageddon :eek:

He has a year to indoctrinate us all in how he managed to solve this problem.

[walker 0]

true but that has nothing to do with determinism but is basic computer science ;)

as long as you approximate the same way each time with the same initial input the result is always the same and thats the only thing that matters.

Hi lethal

You are forgetting time.

And assuming that all the hardware and software in every PC is the same, is nonsence.

Kind Regards walker

Edited May 25, 2011 by walker

[hellfire257 3]

[dmarkwick 261]

[lethal 0]

of course not all hardware and software is the same. but it is true that there are certain standards that are guaranteed to apply. thats what they are for and thats why they exist.

system requirements for any kind of software are there to guarantee that certain functionality exists in a system and that those standards are met - and if by time you mean the speed of the calculations then the same thing can be said there.

its a moot point really. deterministic networked simulations that rely on massive amount of floating point calculations exist.

and i have to agree with others here as well - a minute error shouldn't have a significant impact on the whole simulation as you seem to think. we are not simulating a whole world of physical objects chaotically interacting over the lifetime of a whole game but simply looking at a small window in time from where an actor looses the animation he was playing and switching over to a physical object being dynamically animated till he comes to rest.

Edited May 25, 2011 by lethal

[ryguy 10]

And assuming that all the hardware and software in every PC is the same, is nonsence.

Calculations of ragdoll physics are math. No matter what software the end result is the same. Does my Arma 2 look any different from yours? Nope... same chernarus... same bullet ballistics...

[gammadust 12]

Risking giving support to the OP's opinion which I do not hold...

Calculations of ragdoll physics are math. No matter what software the end result is the same. Does my Arma 2 look any different from yours? Nope... same chernarus... same bullet ballistics...

...mediated by hardware and software. (that is what you're missing, disallowing your conclusion)

it has been mentioned the ocasional errors of floating point calculations.

For further illustration, surprise yourself with rockets falling after 5 seconds take off and friendly fire incidents of a certain Patriot Missile (battery failing to hit a Iraqi SCUD by 600meters)...

and for a more down to earth example

and an explanation

so it all depends on programming and engeneering. Of course ragdools is certainly not a life or death situation (sorry about the drama).

Edited May 25, 2011 by gammadust
for accuracy

[Iroquois Pliskin 0]

Calculations of ragdoll physics are math. No matter what software the end result is the same. Does my Arma 2 look any different from yours? Nope... same chernarus... same bullet ballistics...

Shush. I'm playing on a Pentium III and it can't do projectile ballistics on this scale, so you'll have to get used to my laser beam shots online. :(

P.S. Haven't read so much BS in a long time, the guy (topic starter) is living in the year 1999.

Edited May 25, 2011 by Iroquois Pliskin