Problem with IK legs, too snappy

[aaron_johnson]

i've put some time into making a good character rig, and have whipped up a walk cycle (see link below for movie), however i'm having trouble getting realism using maya's IK on his legs.

for the legs i tried using IK without any FK blending - as a result as the IK chain reaches "full extension" it tends to snap straight quickly, resulting in legs which are too "snappy". i thought a solution would be to rig with knee joints bent (like i did) but model the character with straight legs, then the IK would never go completely straight and the leg wouldn't snap, even though the leg model would look straight. obvious problem there is the knees wouldn't bend in the right place if the joint was in front of the kneecap.

what are people doing to solve this? i've tried a custom leg rig that automates scaling on the leg bones:

http://www.cipher.co.nz/live/catwalk\_side.mov

this is better than the one i made earlier with IK only - but i still think leg snap is noticeable when the grounded leg is extended back, just before it leaves the ground. also the scale routine is a bad workaround i think - it makes it hard to get that fully extended look.

[drblacksnake]

For me it seems you have rather animation problem. Humans, when the walk, do not move this way.

[joojaa]

well actualy youve misundertood things a bit.

IK has a unlimitted amount of soulutions, but animation progs tell the cvoputer om wich plane the solution should be. Even so theres still 2 solutions. inorder for the solver to know wich of the solutions you want to use you will need to either build the knees bent or leater bend them and tell its the prefeered angle.

Now if you neverever want the knee to go completely atraight either put limits to how the individual bones can bend. Or better never trech the ik beyond what the bone can reach, becaue it will ALLWAYS cause you to over extend yourself, in amnner that will give results that look much like your anim.

[joojaa]

well actualy youve misundertood things a bit.

IK has a unlimitted amount of soulutions, but animation progs tell the cvoputer om wich plane the solution should be. Even so theres still 2 solutions. inorder for the solver to know wich of the solutions you want to use you will need to either build the knees bent or leater bend them and tell its the prefeered angle.

Now if you neverever want the knee to go completely straight either put limits to how the individual bones can bend. Or better never trech the ik beyond what the bone can reach, becaue it will ALLWAYS cause you to over extend yourself, in amnner that will give results that look much like your anim.

[aaron_johnson]

do not move what way? be more specific when giving feedback on a clip, what aspect of 'movement' are you referring to? Or were you giving reference to the rig setup i was describing?

i don't think you understood the post i made - perhaps english is not your first language?

By not being able to give accurate feedback it shows your major lack of experience and professional courtesy.

[joojaa]

Well interestingly i know a lot of people who seem to understand english intuitively better then people who actualy have it as first language. go fugure.

[aaron_johnson]

actually joojar i don't think you understood the post either - what i was saying was that it snaps as the IK approaches full extension, it has never been over extended, and as i said in my post - the leg bones were bent when i made the rig.

i do want the leg model to go completely straight (like normal legs do), is what you're saying that the IK line (the line that runs between start and end effector) is never meant to be straight?

i'd love to see some of your human character animations for reference.

[drblacksnake]

So then You didn't understand what Joojaa have said.
Put leg bones under angle on straight legs and limit their rotation in knee joint so they never go straight. Then bind skin.
Otherwise make alternative skeletal setup to which current bones are constrained.

And accordingly Your claim of more detailed answer and valuing me as lame - actually I had written more text than posted, just erased because I was too less motivated to go in basics of animation. Your animation has lack of phisics, thus moves snappy and unnatural, like there's no gravity and no mass. Just You didn't ask about it. You've asked about charsetup, where knee snapping is least of woes. And actually legs rarely go "completely straight" unless upper leg's frontal muscle is additionally stressed, like when soldiers do march.

[aaron_johnson]

joojar never gave any reference to the model - this was a question i asked in my first post which was never answred until now (ie, using bent bones but a straight legged model). i had concerns that the knee wouldn't bend in the right place as a result, but i'll try it out.

RE lack of physics, if you watch male catwalk models they are are trained to walk with a very rigid upper body, and not bounce very much at all.

i agree its not perfect, i 've spent a bit of time fighting with rig issues. i'm not really concerned with accurate muscle simulation at this stage, although i think the model i made looks pretty good.

i apologise if i've offended anyone, but i think i'm owed more than the feedback i recieved, because i have never seen any animation work posted on highend3d.com, or any CG website, that makes a good attempt at realism, its all been over-emphasised cartoon style animation which is not what this is.

only an animator would know this - but its infinitely more difficult to make subtle realistic animation than just overweight everything.

[joojaa]

you never wnat to overextend the efllector! it does not mean you cant get the leg to go straight anyway. You just dont want to overextend the ik efflector further than what the leg can reach.

WHY? well the answer is a bit techy, it can be solved but can end up beign realy tedious. Well its because you use a a worldspace efflector that interpolates its move. now when you reach very close to the edge of your reach or beyond you dont have controll anymore whan this extenzssion happens, because it gets straigfht the second your efflector is further than your bone can reach (and thats not neccesery the monment you key for it to straighten). Also theres no easing this unless you key so that the key is exactly where your full extension is.

NOw for the problem because you interpolate the samples its realy easy to place thekeys so that the actual distance maximum does not happen ON the key but rather off it, on the span connecting the next key. now what thid means is that the delta movement of the leg suddenly is out of controll. Usualy visualizing as a very fast snap.

It took me a very long time to figure all of this out on my own tough so its not something readily apparent.

Then on the bone structure. Well human knees are actualy very tricky. IK can never realy confrtably bend like the human knestructure does because it can not solve over turning. To this end one can however use bonstructures taht are actualy a bit skewed worng. If you place extra satic bone in midbone you can compensate for this... but thats a dirty trick.

Anyway straughtening your leg is not something people usualy (some people do this agreed especialy parading russian military personell) do untill they plan to actualy stop, they almost straighten but not quite (this is actualay a tragic mistake most people do when they happen in a accident shatrtering the leg in thousands of pieces, personaly i succeded to do this with my right arm a long time ago, damn locking joints... it healed but the experience of having your elbow where your shoulder should be is something that could have been avoided with a few hours of ukemi training!).

[aaron_johnson]

hey joojar - thanks for taking the time to post that explanation, very interesting. i think i might get away with the rig i have but i'll go over the keys again and look a bit closer.

ouch that accident sounds painful!

[joojaa]

Not realy didnt feel a thing... It wasnt untill they tried to xaray it whan the pain kicked in, i reflexively dropped the attending x-ray doctor with my left hand. and was pumpped with painkillers so that i still to this day can uderstand how a square room can have total sum of angles greater than 360 but not in the normal perception of things

[wrapnode]

Hi. I thing Joojaa is right. In other words, what I use to solve this problem is using longer bones for character's skeleton standing in rest pose. It means longer femur and tibia bone. The knee joint can be sometimes(in special cases) placed in front of geometry. Skinning is in this case little bit challenging, but you can solve it with influence objects. But results can be achievable in good measure. Joints will never reach their limits(not limits in rotation, but limits in straight line they can get to, causing snappy movement) in Rotate Plane Solver solution. In case of limits: I never use it. It is extremmely conraproductive. You don't give the animator freedom of play. Look at this little walkcycle: http://www.thefilebin.com/userfiles/Nictuk...ic/OgreWalk.avi
This animation is done without method I have mentioned above, so the snappy movement of legs is little bit vissible. But this character is not concepted for walk with straight legs and bwfore I tuned this movement it last long time.

[kubba]

ok limb hyperextension is something that no ik system can handle natively right now (AFIK sure) though as human knees and arms are capable of it..
explanation is pretty simple, when IK controled limb reaches its limit even small change in ik translation creates large angle change and than it sudennly stops..(im not sure if I wrote it clearly, but result is this snappy knee)
so solutions are :
a)refine you animatinon of root TY movement, so leg never gets totaly extended..
set rot limits so knees are slightely bended all the time (results in "sht in pants walk")
c) create bone structure that allows knee to go to hyperextension: one leg with knee slightely extended forward, than via ik's connect bind leg to this leg create IK's for longer leg and animate longer leg original goes to HE.. theres problem that leg doesnt stay on ground - so 2 solutions create another leg of same length for rot values bigger than say 5deg and create conditon node/expresion based on angle to switch between them (good for games or export to other packages that doesnt support bone scaling), count distance between shorter knee and longer ankle and use it to scale shorter tibia (no one will notice in result)

if you need it desperately I can send you script though it expects exact joint axis to work and is not for public yet..

I think I'll write some tutorial for alias site as it seems to be one things ppl have troubles with..

[joojaa]

hmm intteresting I usualy only mede the bones pivot in a slightly different pplace so i dint have to manage this. BUt noew taht i think of it this si a great solution