Slashdot Mirror
The State of Video Game Physics
The Guardian's games blog convened a panel of engineers and other experts to talk about the current state of video game physics. A great deal of research is currently going on to make better use of multiple cores so that advanced physics tools and engines can take advantage of all the processing power available in modern computers. Many of those tools are being put to work these days to find more realistic ways of breaking things, and game developers are trying to wrap their heads around destructible environments. Mike Enoch, lead coder at Ruffian Games, said, "This idea of simulating interactions and constructing the game world similar to how you would construct the real world generates more emergent gameplay, where the game plays out in a unique way for each player, and the player can come up with solutions to problems that the designer might not have thought of." Another area that still sees a lot of attention is making game characters more human, in terms of moving and looking as realistic as possible, as well as how a game's AI perceives what's happening. "The problem is not necessarily in having the most advanced path-finding technique with large-scale awareness; we need to have more micro behaviors, with a proper physics awareness of the environment," said software engineer George Torres.
This generation of students is just damned lucky to have access to such computing power. In the old days, the most readily accessible computing power was an 8080 hobbyist board. Simulating the universe on that is impossible. The students of that era were stuck with just manipulating integrals and derivatives.
Life is unfair. I hate it.
Maybe I'm just old-fashioned, but I preferred the simpler games, the ones that didn't have as rigid physics and things of the nature. Compare modern first-person/third-person shooters and compare them to the classics like Perfect Dark, The Legend of Zelda or Goldeneye. They were so much fun because handling was so easy, you could move, you could strafe, etc. It was so much better! And yet, as games become more realistic, all that happens is that your character becomes more sluggish and less powerful, harder to manipulate. All for the sake of reality, and graphics which will always get old. But the gameplay never gets old. That's why classics are what they are - they're acceptable graphically and a hell of a lot of fun to play.
Want proof? They still have Street Fighter tournaments, Melee tournaments, etc. if you look around in the right places. On the other hand, who cares anymore about Metal Gear Solid 4? Man, even playing Super Mario World is much more fun than the New Super Mario Bros. on the DS, simply by virtue of the fact that the older one is simpler, freer, gives you more control, more imagination, more room to enjoy it.
Seriously? It's gameplay that makes you come back, not reality. I wish we'd drop the reality of things and just make games fun. But I guess now I'm old enough to just make my own games. Sigh. It had to come down to this, didn't it?
I have left slashdot and am now on Soylent News. FUCK YOU DICE.
Indeed, some of the most physics-accurate games I've played, have been some of the most generic and dull in memory. Greater physics can add to a game, but /designed/ physics, is what makes a game /fun/.
Realism won't work any more than it works in Hollywood movies. They need a "Hollywood Physics Engine", with a bit of ACME cartoon logic tossed in. Examples:
1. Fruit stands are magnetic: every thing comes toward them.
2. Things fly strait up and spin end-to-end when they are blasted or exploded in any way. (see also #9)
3. Cars hitting a bail of hay or lump of garbage fly 300 feet. Good guys always land upright while bad-guys always land top first.
4. Sexy breasts jiggle slow and long
5. In space, everyone can hear you scream.
6. Sparks are the most common element in the universe. Every nick and prink causes vast amounts of sparks.
7. Space explosions are usually poofy despite no atmosphere. If it's really big, then an expanding bluish saturn-like ring spreads out from the center.
8. If slow-motion is used, then the bullets are 500 times slower for every 1x speed reduction in human movement.
9. People fly almost strait up in the air if within 200 feet of any explosion. The exception is if they are near a metal hand-rail, in which case they rotate around the rail during the explosion, until facing downward.
10. Poor tire traction, AKA "skidding", actually makes cars go faster. Heroes never win unless they skid a lot. The more smoke from the skid, the faster the car.
11. When jumping between buildings or platforms, nobody ever has a good margin: they always barely make it. Physical laws expand the width to be barely below the maximum of the hero.
Table-ized A.I.
Space ships bank when turning because it stop your Earl Grey from spilling all over the console.
You'd think people complaining in a physics thread would know some.
No sig today...
A dozen years ago I developed and demoed the first ragdoll physics system that worked. Among other things, I'm responsible for the "ragdoll falling downstairs" cliche; that started with a demo I did in 1997. I looked at ragdolls as a first step. I was expecting game development to go in the direction of physically-based characters driven by active control of character muscles. That hasn't happened.
The problem is partly technical and partly dramatic. The dramatic part I encountered in dealing with Hollywood types. What directors want is to specify the start and end conditions; the job of the system is to realistically get the character to the desired ending mark. In real-world stunt work, there are wires, guides, and rails that make things go the way the director wants, even when that's not physically realistic. When that's not enough, cuts are used to conceal the lack of realism.
Physics systems are inherently unidirectional - you keep working forward from the current state. This is fundamentally incompatible with directorial control. As a result, the trend in character animation has been to get enough motion capture data to cover the things you want the character to do, and use a motion splicing engine to patch the pieces together. (This, incidentally, was first used in Godzilla, the movie, for the baby 'zillas). That's become more or less the standard approach for games.
Using a character control AI to drive the character's muscles realistically has been attempted, but with modest success. Motion Factory tried this in the 1990s; their system was only kinematic, and not too successful. Havok is trying it now. For this to work, you need computerized muscle control good enough to drive a real-world robot, like Big Dog. And then it has to look good from an aesthetic perspective. It's really a hard robotics problem, which is why I was interested in it in the first place.
From a gameplay perspective, if you take the physics seriously, you lose the "superhero" capabilities of game characters. Jump off a balcony, and don't expect to land on your feet. Jumping up to a balcony? Forget it. Hand-to-hand combat works about as well as it does at the dojo. ("Your left foot was too far forward for that throw. Again!" "Yes, sensi.") Trying to control a physically realistic character via a joystick is nearly hopeless. You can't even drive a real car very well through a remote joystick, let alone a game pad. (I've actually done that; using a remote steering wheel is a huge improvement over a joystick.) In driving games for consoles, the physics is tweaked to make the car incredibly stable. (Lowering the center of gravity to below ground is a common trick.)
So what do we have? Ragdolls. "Infinitely destructible environments." Some skin deformation. Cloth. Plus rain, snow, water, and explosions that don't feed into the game play at all. (That's mostly what the "physics cards" do.)