|
Post by Mouthwash on Sept 7, 2016 23:08:57 GMT
I have some worries about this. Spore's ability to animate characters without a predetermined morphology was cutting-edge, and it couldn't even incorporate collision. Has procedural animation really advanced so far in a decade that a team of nonprofessional volunteers (not to disparage them whatsoever; they're clearly talented) can do better?
|
|
|
Post by GRODOG on Sept 7, 2016 23:35:13 GMT
well in the behavior editor ur gonna get to animate how they move, mate, eat, etc...
|
|
|
Post by Mouthwash on Sept 7, 2016 23:54:42 GMT
Manually, you mean?
|
|
|
Post by ATP Kraken on Sept 8, 2016 0:16:34 GMT
There will probably be automatic animations, but have a little bit of gait adjustable manually. Just don't take it too far, and we get this:
|
|
|
Post by Moopli on Sept 8, 2016 18:52:07 GMT
Sweet topic. I'm currently working on prototyping exactly this, working out the math, and the approach I'm using is actually kinda novel.
Most animation systems work in terms of joint angles -- posing a rig will set the angles of all joints (in all the degrees of freedom each joint has), and animations are interpolations between those joint angles. For foot, hand, and head placement, modern systems use inverse kinematics to solve, analytically, for the exact joint angles needed to place the given body part in the correct position.
Problem is, there aren't general analytical solutions for IK; so most state of the art IK systems (some in animation, but most in mechatronics) use specific analytical techniques for specific classes of problem.
We're going to have to be more general, however we don't have to be as precise; so what I'm working on is an approximate solution that doesn't keep limb length constant (allowing for some error so solutions are more stable) and also doesn't provide exact collision constraints or joint angle constraints. In essence, by allowing bones to flex a teeny bit, and allowing the body to be a bit squishy, you can get away with cheaper approximations for much more general limb configurations.
Given my experiences with verlet systems, I think it could be cheap enough that we could even run your organism's animations in real time rather than bake animations. Possibly all animations; if we parallelize the computation, but we'd obviously want to do some LOD so far-away organisms use static animations without good foot placement etc.
|
|
|
Post by StealthStyleL on Sept 8, 2016 18:53:30 GMT
What do you mean bake animations? Excuse my question.
|
|
|
Post by Moopli on Sept 8, 2016 19:18:51 GMT
Baking is a general term when talking about assets, where you convert an asset into a form ready for reading directly into game -- eg, from an animation in Maya to whatever format the game reads, or from a psd of an art asset to the image format your game will read, etc. When I used it above I referred to the idea of taking an animation that's generated procedurally in game, and making it an animation that could be written to file and read directly by ogre; rather than an animation that's used by moving every joint of the body in Thrive code.
|
|
|
Post by StealthStyleL on Sept 8, 2016 19:28:39 GMT
Ok, thanks.
|
|