Physics and AI
- Thread starter Sonic
- Start date
And how much flexibility will programmers have with that 1TFlops of power ? Split in 4-8 processors each with 4-8 128 bits simd units doing non-dependant FMADD vops ? That's a LOT of power for sure, but I wonder how much of all that power could realistically be used and sustained in a game ? That said, no matter what, it's still an incredible amongh of crunching power 
This is a nice topic.
Imo this ties in nicely with the recent GDC speech of Jason Rubin.
The following is all IMO, so feel free to disagree.
As somewhere in the future graphical improvements will stop to be the most important factor on deciding which game to buy, as most people will probably not notice a difference between a 100.000 or a 150.000 polygon car (and that's a 50% improvement) in a GT3 like racing game, developers will need to improve other areas (Like physics and AI) even more than they are doing now, to differentiate their games.
This and the great amount of processing power that will be available in the next generations of game-consoles opens IMO a huge field of possible gameplay innovation based on physics and AI.
As an example take this small game: porrasturvat
This is just an example for a game that would not be possible without ragdoll physics. Even if you don't like the game (I think it's hilarious but that's just me ) take it as a primitive example of a gameplay idea based on physics with graphics as a byproduct.
Live for speed would be another example where accurate simulation is a top priority and graphics, while still important, are not the deciding factor for playing this game.
IMO there are tons of gameplay possibilities based on physics.
I played around for hours with the Actor demos and I'm sure I'm not alone.
Creative people could certainly come up with some neat mini-games or will find ways to integrate physics in new gameplay aspects of old genres.
For me this will be more exciting than the graphical improvement that naturally comes with the next generations of consoles.
I know from discussions around here that as "good" perceived physics need a lot of tweaking and balancing (there was the example of Driving Emotion Type S which IIRC featured an impressively detailed physics model, but was simply frustrating to drive), so this will certainly place another burden on the developers but IMO it's worth it.
Imo this ties in nicely with the recent GDC speech of Jason Rubin.
The following is all IMO, so feel free to disagree.
As somewhere in the future graphical improvements will stop to be the most important factor on deciding which game to buy, as most people will probably not notice a difference between a 100.000 or a 150.000 polygon car (and that's a 50% improvement) in a GT3 like racing game, developers will need to improve other areas (Like physics and AI) even more than they are doing now, to differentiate their games.
This and the great amount of processing power that will be available in the next generations of game-consoles opens IMO a huge field of possible gameplay innovation based on physics and AI.
As an example take this small game: porrasturvat
This is just an example for a game that would not be possible without ragdoll physics. Even if you don't like the game (I think it's hilarious but that's just me
) take it as a primitive example of a gameplay idea based on physics with graphics as a byproduct.Live for speed would be another example where accurate simulation is a top priority and graphics, while still important, are not the deciding factor for playing this game.
IMO there are tons of gameplay possibilities based on physics.
I played around for hours with the Actor demos and I'm sure I'm not alone.
Creative people could certainly come up with some neat mini-games or will find ways to integrate physics in new gameplay aspects of old genres.
For me this will be more exciting than the graphical improvement that naturally comes with the next generations of consoles.
I know from discussions around here that as "good" perceived physics need a lot of tweaking and balancing (there was the example of Driving Emotion Type S which IIRC featured an impressively detailed physics model, but was simply frustrating to drive), so this will certainly place another burden on the developers but IMO it's worth it.
Saem said:
We don't usually need that much numerical analysis for games, but with decent physics you do need to understand error-control and decent methods of solving matrix systems.
A good uni course in numerical analysis should give you the basics, things like Runge-Kutta, A-Stable regions and (the all important) sparse matrix solving.
randycat99
Veteran
Ballistics and "pool-ball-like" collision behavior are what I would tend to lump into "physics" simulation. I think these things are important to maintain realistic portrayal of on-screen action. Of course, most people probably don't realize unnatural motion and reactions, and it stands out for me because I have a mechanical engineering background. So take that how you will.
So something struck me strange while watching a cut scene of a game I happened to be playing today. Potentially, game developers and artistic directors don't particularly care about the utmost in physics behavior (given the computing power or not). I would think that given the luxury of offline rendering/computing when making a cutscene, extra care would be directed toward the realistic depiction of physics behavior. However, it seems whoever is in charge simply opts for what "looks about right". So what we often get is somewhat cartoony behavior when it comes to situations depicting ballistics and dynamic motion of bodies with a mass acted upon by external forces. This is offline processing, so not bothering to implement the necessary computations for real physics is somewhat boggling to me. Even if real physics simulation isn't supported in the particular development platform, you could still set up the dynamic system and get some data out of an outboard dynamics simulator (they do exist in engineering/scientific circles). Then again, the cost of this extra development software may not be justified to the game production supervisor if he reasons "someone" can just "eye it" and "fudge" the behavior into the animation. Doing so isn't cheap, but it isn't exhorbitant, either. A $10k piece of software working in concert with an entry-level solid model CAD workstation ought to do it. Nevertheless, developers probably wouldn't go that far just for a cutscene animation, so here we are...
This is a bit different than designing a realtime, in-game scenario. If the computing resources aren't available or the appropriate physics engine doesn't exist in the game engine, then you have to "eye it" and pre-designate an action that "looks about right". So getting what you get is somewhat understandable. Now take the situation where the resources are available via some future uber console hardware, and I anticipate that realistic physics would still not get the attention it deserves (per the offline rendering example). I just think that game developers are simply content to "eye it to look about right" than to bother with real physics. If they cannot get it in pre-rendered scenes today, I don't hold out much hope that it will get done in-game in future consoles even with vast computing resources (unless a concerted outcry is organized to generate greater awareness of the issue amongst game developers).
So something struck me strange while watching a cut scene of a game I happened to be playing today. Potentially, game developers and artistic directors don't particularly care about the utmost in physics behavior (given the computing power or not). I would think that given the luxury of offline rendering/computing when making a cutscene, extra care would be directed toward the realistic depiction of physics behavior. However, it seems whoever is in charge simply opts for what "looks about right". So what we often get is somewhat cartoony behavior when it comes to situations depicting ballistics and dynamic motion of bodies with a mass acted upon by external forces. This is offline processing, so not bothering to implement the necessary computations for real physics is somewhat boggling to me. Even if real physics simulation isn't supported in the particular development platform, you could still set up the dynamic system and get some data out of an outboard dynamics simulator (they do exist in engineering/scientific circles). Then again, the cost of this extra development software may not be justified to the game production supervisor if he reasons "someone" can just "eye it" and "fudge" the behavior into the animation. Doing so isn't cheap, but it isn't exhorbitant, either. A $10k piece of software working in concert with an entry-level solid model CAD workstation ought to do it. Nevertheless, developers probably wouldn't go that far just for a cutscene animation, so here we are...
This is a bit different than designing a realtime, in-game scenario. If the computing resources aren't available or the appropriate physics engine doesn't exist in the game engine, then you have to "eye it" and pre-designate an action that "looks about right". So getting what you get is somewhat understandable. Now take the situation where the resources are available via some future uber console hardware, and I anticipate that realistic physics would still not get the attention it deserves (per the offline rendering example). I just think that game developers are simply content to "eye it to look about right" than to bother with real physics. If they cannot get it in pre-rendered scenes today, I don't hold out much hope that it will get done in-game in future consoles even with vast computing resources (unless a concerted outcry is organized to generate greater awareness of the issue amongst game developers).
