How to get backwards ik?

#5

I disagree, this is somewhat tougher, as the model is mechanical, and proper joint placement is crucial for those objects to animate properly. By placing the Joints where you need them to be mechanically, you get a backwards chain, but you need the IK to evaluate forward like a knee. This obviously doesn’t work due to the limits of the IK solver itself. You’ll have to get clever, by using the rotations of the forward IK chain to rotate the bind chain that’s essentially backwards. It’s not a complete solution, but should get you in the right direction.

btw, please enlighten me if I’m mistaken here.

Best

#6

Oh you mean that even though the limbs are bending backwards they should still bend “forward” like a standard knee (or elbow)?

I didn’t get that part if that is the case. Yeah that’s slightly more involved.

#7

I happen to agree this doesn’t really make any sense with out a diagram for what you are attempting to achieve.

In Maya even you set up the IK one way you can get it to move the other direction by setting the Perfered angle and setting up pole vectors correctly. For the most part your pole vectors will control the direction of the bend. All the IK is doing is handling the calculation for keeping your end joint in the correct position.

If you draw out or sketch out what you are attempting to do, I can throw together a quick rig that should be able to handle what you are looking to achieve. Just show how you want it to bend.

If you can not do that, you may not understand what it is that you are attempting to do.

#8

everyFrame is correct and yes, I’ve basically been trying to cook up how another chain could drive this chain but I can’t figure out the relationship.

#9

The problem with a design like that is , in order for the character to squat, it first has to rise up. If you had some suspension in the lower leg, it would be easier to achieve as the suspension would buffer the downward movement during the bend but because it is rigid, you wont be able to get around that issue.
Its possible to make this work, but I think you will need more than IK. I think you;ll need a regular IK that is forward of the knee, and have expressions that link to the rotations in the front chain to ride the upward movement (of the whole body) before the drop.
Or you give him some suspension at the ankle.

#10

Long time no see! I’m stuck with the design, unfortunately (it’s the Korean answer to Asimo - Hubo). I just haven’t been able to wrap my head around how the other chain would drive this chain. :banghead:

#11

Yeah its been ages… I hope you are doing well

I just did this in XSI to see how hard it was

Make an Upside down bone chain so the goal is at the top

have an FK chain that constrains the orientation to one chain, then blend it to the other during the blend. (I constained to parented Nulls to correct the orientation on the upside down chain)

The FK chains end will ride upward during the blend, but there is no way around that without adding suspension. Its not really a design that makes a great deal of physical sense (no offense meant. Thats the fun of fantasy)

#12

Very clever. Nice pooby!

#13

So a couple notes.

Your rotation points are off on the arm and legs, that is if the object is skinned. You can get the same result keeping the illusion that the object is rotating from the inner pivot.

If you absolutely need to place the joint at that location bring the wrist joint up to the forearm area raise it up to be level with the inside joint pivot. Then create a second set of joints to run from the forearm to the wrist to recenter on the wrist. I’ll do a mock up in a moment.

The real problem that you will encounter when over extended a joint like that arm is the pop that you will get when animating. That “Pop” will visually distract from most any animation that you do.

Most times people set joint limits to keep from ever attending a flat or straight arm and this creates the pop. I have also use a double elbow joint to stop the pop from happening.

#14

Attached is a sample file where I used the IK to Bend at the required joint with no skinning.

This was a Fast 5 minute sample. But should illustrate how it would work.

#15

I’m not quite sure this is a solution, even with skinning. Yes the IK is moving the knee where it’s required to bend. however, the extra joint at the top isn’t being evaluated in the IK solution, and therefor doesn’t receive any rotations (am i wrong here?) Skinning to it won’t do any good, since it’s there to simulate the upper leg joint, but doesn’t get any rotations. Maybe I missed something, please enlighten.

Perhaps it could be a start, But overall the issue Pooby touched on is the real road block here. The physics of this design -require- the feet to drop, or the hips to raise for the backwards chain to evaluate forward without squashing of some sort.

#16

I’m afraid I’m not following your solution either, Darksuit. I can’t see how I can make Pooby’s work, either. The best I can come up with is changing the rest positions so that I get the proper ik bend, then when either limb has to be straight I’d have to switch to FK. Not really a solution, though.

#17

Here is the arm animated. Also note that I did not skin the two parts. This is just a rough rig. Done to illustrate that the part can move and pivot from those locations.

#18

Ah, it’s that locked joint that does it. Thank you for your help. :beer:

#19

Sorry dood. I applaud your effort, But I don’t think your quite getting the issue. Also, are you drawing joints in the perspective view? Nothing seems to line up on any axis…?

Maybe it’s how it’s animated, but It doesn’t look like the IK is evaluating appropriately, and still does not quite seem like a solution.

It’s really not the limitations of IK, the joints, or anything software specific. It’s the inherit design that’s flawed (no offense obviously) Looking at Asimo, It’s obvious that each the hip, knee, and ankle are in a direct vertical line with each other. Of course this is solvable. But with his design, even if the IK supported a backwards chain the rigid objects would still go through the floor, or raise the hips when animated, without some sort of squashing.

#20

The robot design sucks, no doubt. Anyway, I made a rough test with parts of the arm.

HuboArmtest001

Now the shoulder…

#21

or just silly simple

#22

Lol, of course! How we love to get complicated sometimes ")

#23

Wont that make the ankle have the rotation at the back?

#24

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.