Recent searches
Search options
I'm working on revamping the inverse kinematics algorithm for my procedural animation with focus on handling a variety of foot roll scenarios better. Results so far are promising, though far from done yet.
Only, I'm having an issue with snapping when the heel was the best target but can't reach anymore, so it tries to reach with the toe instead. It's not just a technical issue; I'm unsure how I even want it to work.
#ProcGen #GameDev
@sinbad Haha, yeah :)
I think I'll keep getting comments comparing to Ministry of Silly Walks for as long as I keep working on this procedural animation...
@runevision @sinbad I hope you lean into it and add a silly mode
@runevision detect this case and impart an unbalanced impulse that may make the character trip and fall over?
That, or make desired planting/location anchor be an animated bone by itself, which moves hell/toe during the cycle.
@StompyRobot I'm trying to get away from determining the foot contact specifics based on timing, as it's brittle. If the footstep location is in a bit different location than the cycle "expects" or the animal is transitioning from one gait style to another, timing based poses can quickly break down. My current thinking is to make as much as possible inferred without timing.
@runevision I actually love that approach, because I think canned non-physical animations always come off unnatural!
Natural Motion did some work on this, but more for body pose not foot contact.
Run a local collision against the ground (as an expanded foot aligned box or capsule) to see where the purchase would be, and make that part of the foot be the target?
If I had to pick just one point, I'd pick always toes, but you end up walking a bit as if you're always sneaking around...
@StompyRobot I do know where the ground is. The problem is that which part of the foot is lowest (heel or toe) can change based on how bent the leg is, which again depends on which height the ground is at (relative to the hips), so this can vary in non-trivial ways based on walking on a flat surface or up or down slopes or steps. That's what I'm trying to solve with the new IK. I'll find a way to solve the snapping issue, it was just a little insight in what I'm currently tinkering with. :)
@runevision hmmm wouldn't that depend on the momentum of the character as a whole? In motion, they surely want to catch themselves with the heel, especially if the target is where the foot is supposed to land, and push off with the toe, especially if the target is still where the foot was planted. When standing, hm, tricky!
Just random thoughts, surely not of much help : )
@Naam Well, if a human is walking backwards, I’d say they set off with the heel and catch the ground with the toe. So my theory is that it’s more about which foot orientation is more natural in front of versus behind the body.
@Naam And most natural here seems related to just assuming the least strained ankle joint rotation.
@runevision ah yes I didn't consider walking backwards ^^ more: maybe it would behave differently when standing still. But then, you _are_ building a procedural walkcycle generator, so I guess momentum is a given!