So, i''ve encounted an intresting problem which iv''e not encountered before. It involves the animiation system which im building. For the game im working on, I''ve decided to decouple the the character animiations from any particualar state of a character, rather a proxy object acts to translate a particualr character state into an approriate animiation, or set of animiations. This proxy object, is basically the visualy represntation of the internal state of a character. It''s set up as a series of bitflags describing the characters movements, moods, etc.. from an external observers viewpoint. So generating this act vector, as i call it, is no problem. The issue is generating the mapping from act vector to animiation sets. Since the number of bits involved are large on the order of easily 32 bits, the number of possible states are also very large. So a one to one mapping is out of the question. So instead i''ve im using a distance algorihtim to find the best fitting animiation for a given act vector. Now the issue is the distance formula will work most of the time, but there are constraints within the transitions between animiation sets and also wihtin the animiation sets ( with respect to bits which must be present or mutually exculisive bits which cannot be both active etc.. ). The question is how do i codify these constratints, within a expandable and maintainbale framework. I want to add onto these aniimation sets in the future and they should be plug and play with respect to the characters. Im tenativly exploring bloom filters linked in a hierachy manner to account for the enourmous state space. Any ideas? -ddn

Okay, so you have roughly 232 possible states (from 32 binary variables) for the proxy. How many animation variables do you have and how many states can they take on? This knowledge is necessary if you''re to find a reasonable mapping from proxy states to animation states.

I have a thought about how to solve this, but it is dependent on your ability to produce a suitable set of training instances. Can you, for a given proxy bit string, determine the approximate animation state you want to display? I presume the answer is yes, but that your problem is determining the mapping for all states.

If you can supply more information, I might be able to help you out with a solution.

Cheers,

Timkin

Currently there are only a small set of animiations. I have :

stand
walk
run
crouch
crawl
attack with weapon
attack with fist
fall down
die
cower

That''s just the inital set. The idea is as we have more animiations we just add them into the set of animiations, with the approriatte mapping descriptors and the system uses them correctly. No need to maintain or create code to use any new animiations.

There are constraints like :

run can only transiton from walk
crawl can only transition from crouch
attack must face attack target
etc...

I want to codify these constratins into a cohesive framework as well.

-ddn