Cool Game Breakable Technology - Force Unleashed

[Ishtvan]

This looks pretty cool (aside from the lame buzz word name, "digital molecular materials"). I wonder how they do it, since calculating real stress fields within each moveable rigid body seems really expensive. Maybe it has some binary "will it crack" test and then some sort of optimized crack propagation based on mechanical properties?

 

 

EDIT: Also this

[Komag]

Do do a LOT of that seems like a good use of a physics card like Ageia

[STiFU]

Yeah the Euphoria physics-engine. Saw some awsome stuff of that one. You should also check out the indy jones videos. I saw a man on a hanging bridge trying to balance it. AWSOME!!

[demagogue]

I started playing with Euphoria through the Endorphin editor before all of this started getting announced ... before it was cool. (Well, there are differences ... Euphoria is for in-game behavior-physics, Endorphin is for canned animations, but you get the gist of it).

 

BTW, I recommend to everybody to play with Endorphin, too. There's a completely free public version which allows you to do everything the Pro version can do, you just can't import textures (not legitimately anyway).

 

I guess you guys probably know this, but the DMM tech is different from Euphoria, BTW, one does materials the other behavior.

 

As for how that DMM works, my guess is it handles it at the material level, and it's all statistical, yeah, I guess now reading what you (Ish) said more carefully ... basically what you said.

Edited May 14, 2008 by demagogue

[Baddcog]

Looks pretty cool. Much better than standard breakables.

 

Funny how they use a ball model on the cheap wood and an R2D2 on the 'good wood'.

 

Deffinately alot more realistic but still has a ways to go. Really plywood is layed out in opposing layers. vertical/horizontal/vert... I only saw vert splinters there

Still cool how the wood could be tougher with a different setting.

 

This could be a modelers nightmare. How do you tex the inside of the wood properly. Especially with properly laminated plywood.

 

The castle was cool but less convincing. It was breaking more like a solid piece of clay, not like a castle made of stone bricks.

[Komag]

Yeah, it's like the first baby steps into the future. I want to play the games in 15 years... I mean 15 years ago all we had was SNES physics

[Domarius]

Shit Can't wait to get Force Unleashed.

[STiFU]

Oh yeah, I just had a quick look at it, so I must have overseen that Euphoria and DMM are for different purposes. Anyway, Euphoria is awsome as well...

[Ishtvan]

Euphoria is cool, but it seems like the logical progression of ragdoll physics, rather than a radically new idea. With the ragdolls that we've had for some time now, you've got joints and constraints simulating the human joints, and you can apply forces to each if you want (for example in Doom3 or Halflife, you could pick up a ragdoll by the head and swing it around).

 

It seems like the logical next step to simulate the human musculature at that point, applying forces that keep the ragdoll upright, balancing, and other things a human might want to do. Everyone who worked with ragdolls was probably thinking of it. I'm sure they put a lot of work into actually implementing it though. It's probably not trivial to write an algorithm controlling the muscles to stay standing upright in all situations, even though we humans do it unconsciously.

[STiFU]

Yeah, and I guess it takes a whole lot of tweaking and testing to make it look right. Sure, this feature had to come eventually but I am still amazed that those guys managed to implement it already. I mean, what else do we need now regarding ingame-physics? The only thing I can think of is proper liquid simulation (like big oceans and authentic watersplaters/-falls and stuff), but maybe DMM will also be capable of that in the end...

 

That being said, I am assuming there woun't be very many new developments in the physics-department anymore, but instead ragdolls will be apllied to many more objects and more forces will be introduced, like wind and leaves of a tree. What do you think?

Edited May 16, 2008 by STiFU

[Crispy]

Yeah. Basically the tools will get better and the algorithms and computers will get faster, so we'll have utterly stupid levels of interactivity.

 

I predict that the AI will have trouble keeping up with the sheer level of world interactivity. Forget moving a chair to block a door... how about, say, collapsing the ceiling to make a pile of rubble in its way.

 

What happens when an earthquake destroys part of the castle? You've got partly-ragdolled-but-not-dead AIs in the rubble... do the other AIs try to dig them out? Just imagine how much coordination and physics smarts are required to dig someone out of a pile of rubble without hurting them. Even humans don't always do it properly (don't want to remove the rubble that's supporting the concrete slab above the survivor!).

[demagogue]

One reason I liked Euphoria so much when I heard about it was that it was developed by people in the cognitive science department (I also did a lot of cogsci in undergrad), studying reflexes and muscle coordination, which can get pretty complex, not just the biomechanics, but the actual cognition of it.

 

They developed Euphoria for their own research simulations, and only later did the idea come to apply it to animations and games. So I don't get the impression they were originally tweaking it to look good for games, but they kept back-testing it against experimental data until it accurately modeled how humans really react, so it looks realistic. But that said, I'm sure once they started marketing it as a game app then they started doing some tweaking for what the market wanted.

 

As for what Crispy is talking about, figuring out these sorts of physical puzzles, it sounds like those experiments where they put AI (Shrldu-type) into FPSs with simple physics puzzles and see how much they can figure out. Very quicly they run into the AI-complete problem (I think that's what it's called) where an AI will have to have the full knowledge and common sense of a normal human to do even simple tasks. Anything that appeals to common sense is a problem.

I guess my point is, for those kinds of problems, it's a sort of all-or-nothing thing; either you have AI that can fully capture human common sense, or they'll have problems figuring out even simple cause-effect puzzles.

Edited May 16, 2008 by demagogue