Ragdoll Physics: A Powerful Tool in the Future of the Gaming World

Ever seen the short animation that appears all over the Internet of a woman falling through an endless supply of bubbles? Sometimes she gets caught some bubbles and her limbs flail about frantically, her back arcs backwards realistically as she hits the bubble, her arms lay aside lifelessly, and then she slowly slides off the impeding object and continues to plummet below?

These are an example of ragdoll physics - a concept that's existed since the '90s but has recently taken off with some of the exciting possibilities it has in the game industry.

For a brief history, it's important to understand how a game's basic art and displays work. As anyone interested in joining the game industry should know, a game consists of a series of code that basically sends data back and forth between one another, different streams of data illicit different types of reactions and responses from other parts of the code. How they react is based on the engine, both the game engine and the physics engine which can either be the same or two separate things. Usually the game engine is just how the game world interprets the code around it, however.

In the days of Doom, animations were very elementary. Using basic .gif images, enemy creatures were simply two or three cycled images that occasionally flashed and played a short midi file based on certain parameters. For example, what would appear to be a creature walking towards you was actually the creature following a pre-scripted "path" based on its limited A.I. - the "movements" of its legs were simply a cycle of images in very archaic animating styles, usually just flashing an image of the creature with its leg up, and one with its leg down.

Death animations were even more limited - usually it just played a small, three or four frame animation, before the slain enemy appeared as a corpse. The earliest limitations were obvious when a player would kill an enemy, see it hit the floor, and would see the same angle of the body no matter what side they were on. Therefore, in Doom, if an enemy werewolf was killed, it would show its feet and lay on its back. If you tried to go on the other side of your victim, it would appear as though the dead body is turning 180 degree angles so that you could not see the head. That was because what existed was now a simple image, almost as though it were pasted on the screen.

Things have come a long way since Doom however - simple collision coding has given way to realistic physics engines, accounting for things like wind, speed of the bullet, bullet drop, penetration of enemy armor, you name it. But just as exciting are ragdoll physics. Whereas before, the enemies were simple .gif images, now they're fully constructed models that exist inside the game world with a pre-defined skeleton. Using a series of algorithms, the NPC has a seemingly infinite number of potential "death animations" based on how and where it is hit.

In essence, rather than the old days of simply cycling a death animation, which was far less CPU intensive, the model persists, no animation is run, and it "reacts" based on where you hit it and how, with what force, with what kind of object, and really any other parameter that the designers have put in.

One of the most recent games to showcase this ability is Elder Scrolls IV: Oblivion, which departed rapidly from Elder Scrolls III: Morrowind's rather dated form of simply cycling a death animation. Though not using simple .gif images like Doom, it did however have a limited one cycled animation per creature slain, therefore clearly inhibiting realism and enjoyment.

How might this effect you in the game industry though? As a game designer, you must be thinking, "By gosh, let's use it right now on all our games!" It seems like an appealing idea, what could be more enjoyable than any game that uses these ragdoll physics to help emulate actual, real world reactions to physical conditions?

Well there are two major problems to consider. The first is how remarkably difficult to test these are. In addition to having to test all the prescribed parameters, which could take at least a solid day of testing for an entire team on a single model, there are a number of negative tests that need to be carried out. For example, what if an NPC collides with two other NPCs of varying heights, after getting hit by the player's weapon/bullet? This may seem simple, but with the wrong parameters put in, any number of things could happen, including all three enemies simply getting stuck inside of one another, or getting lodged into strange parts of the zone.

Therefore the entire process of implementing these physics is very time consuming - but potentially worth it, based on the amount of "freedom" it allows. Because animations all have to be hand drawn, an NPC could not be asked to fall off a high point and hit every object coming down, reacting believably without an animator personally putting in all of those possible parameters. However, with ragdoll physics, the body/skeleton will react as realistically as it can, without having to have the animations pre-defined.

The other major problem comes from the fact that ragdoll physics are still relatively unrefined. That is, there is still a large deal of "obtuseness" and rigidness to the way they move. Often times, joints will lock up, and the pelvis in particular will appear to be an "anchor" point of the body. It will fly backwards, and the rest of the body will bend with it lifelessly. The joints, rather than having their partial rotating ability they have in real life, will remain largely locked or only wave up and down rather than side to side.

Sometimes, with weight not proportioned correctly, the NPC will flail about even unrealistically, its only real advantage being its unpredictability in thrashing.

Nevertheless, ragdoll physics have the opportunity to change the way video games are treated, made, and expected - so they are a powerful tool to be considered in the game industry.