Spline IK Joints "Floating" above curve #81707

New Issue

System Information
Kubuntu 20.04 LTS
GTX 1080 driver 450

Blender Version
Broken: 2.91 (current master), 2.63, 2.79b

Short description of error
Spline IK on a chain of bones may have joints that do not lie exactly on the curve, but "float" just above:

See the 3D Cursor -- this joint is not on the curve! This bug occurs on curves that have sharp, acute angles.

Exact steps for others to reproduce the error
Spline_IK_floating_joint.blend
Create a curve with a sharp angle, then create a few bones and make a Spline IK constraint. Likely as not, you'll see the error, too.

I don't even know if this is a bug, or if the IK algorithm does this on purpose as an approximation... but I think it's unexpected to users that expect the bones to follow the spline exactly.

**System Information** Kubuntu 20.04 LTS GTX 1080 driver 450 **Blender Version** Broken: 2.91 (current master), 2.63, 2.79b **Short description of error** Spline IK on a chain of bones may have joints that do not lie exactly on the curve, but "float" just above: ![image.png](https://archive.blender.org/developer/F8990423/image.png) See the 3D Cursor -- this joint is not on the curve! This bug occurs on curves that have sharp, acute angles. **Exact steps for others to reproduce the error** [Spline_IK_floating_joint.blend](https://archive.blender.org/developer/F8990429/Spline_IK_floating_joint.blend) Create a curve with a sharp angle, then create a few bones and make a Spline IK constraint. Likely as not, you'll see the error, too. I don't even know if this is a bug, or if the IK algorithm does this on purpose as an approximation... but I think it's unexpected to users that expect the bones to follow the spline exactly.

Added subscriber: @Josephbburg

Added subscriber: @lichtwerk

Changed status from 'Needs Triage' to: 'Confirmed'

I can confirm this.

Might be indeed the way it is approximated, but will confirm for now.

Observations:

Increasing the curve resolution helps (Resolution Preview U)
seems especially problematic with Even Divisions in the Spline Constraint being turned OFF

I can confirm this. Might be indeed the way it is approximated, but will confirm for now. Observations: - Increasing the curve resolution helps (`Resolution Preview U`) - seems especially problematic with `Even Divisions` in the Spline Constraint being turned OFF

Yeah! And in fact it seems like one of the best ways to get rock-solid performance from a spline IK is to line each joint up exactly with the evaluated curves' vertices... which is annoying to do because they aren't usually even divisions, and even then the bones won't follow them exactly.

I love Spline IK ❤, so I want to help make it better!

Yeah! And in fact it seems like one of the best ways to get rock-solid performance from a spline IK is to line each joint up exactly with the evaluated curves' vertices... which is annoying to do because they aren't usually even divisions, and even then the bones won't follow them exactly. I love Spline IK ❤, so I want to help make it better!

Added subscriber: @dr.sybren

Please test this with a few older versions of Blender. Knowing whether a bug was recently introduced or has been in Blender for longer will help developers in finding the root cause of the issue.

Please test this with a few [older versions of Blender](https://download.blender.org/release/). Knowing whether a bug was recently introduced or has been in Blender for longer will help developers in finding the root cause of the issue.

Here's an image of 2.79b:

I was also able to confirm in 2.63, so I think it's safe to say this behavior has been around for a while.

Here's an image of 2.79b: ![image.png](https://archive.blender.org/developer/F8994631/image.png) I was also able to confirm in 2.63, so I think it's safe to say this behavior has been around for a while.

Please see my comment in #81704 -- I think I know why this bug occurs.

Here's an extreme example:

You can see that in this case the difference in the World-Space distance from point to point, and the space along the curve is exacerbated by the low curve resolution and the high chain-resolution. I think that's what's causing the problem. However, I could be wrong... in my Python Snap Pose Bones To Curve function in my addon, I haven't corrected for this and it still looks right... so I could be wrong that that is the cause of the problem! Maybe the factor along the curve is actually being calculated incorrectly somehow.

Here's the .blend file I'm using for this example:
spline_ik_floating_joint_bug_play_with_me.blend

Use the shape key on the curve to make it straight and then curved, and use the resolution of the curve to see how the problem seems to right itself when the curve is high-resolution.

Please see my comment in #81704 -- I think I know why this bug occurs. Here's an extreme example: ![image.png](https://archive.blender.org/developer/F9000348/image.png) ![image.png](https://archive.blender.org/developer/F9000384/image.png) You can see that in this case the difference in the World-Space distance from point to point, and the space along the curve is exacerbated by the low curve resolution and the high chain-resolution. I think that's what's causing the problem. However, I could be wrong... in my Python Snap Pose Bones To Curve function in [my addon](https://gitlab.com/josephbburg/blender-chaintools/tree/development), I haven't corrected for this and it still looks right... so I could be wrong that that is the cause of the problem! Maybe the factor along the curve is actually being calculated incorrectly somehow. Here's the .blend file I'm using for this example: [spline_ik_floating_joint_bug_play_with_me.blend](https://archive.blender.org/developer/F9000401/spline_ik_floating_joint_bug_play_with_me.blend) Use the shape key on the curve to make it straight and then curved, and use the resolution of the curve to see how the problem *seems* to right itself when the curve is high-resolution.

Broken: 2.91 (current master)

Please always include the Git hash when you're not referring to an official release. That way it's less ambiguous which version you're referring to. It's happened that people just hadn't pulled changes for a while, and were a week or two behind on updates and were reporting a bug that was already fixed. Given that you've just shown that it was in 2.63 already chances of that are small, but it's good to do anyway. It also helps to keep the task description up to date, so that people who want to fix it don't have to read through all the comments to get the info. I've done that just now, this is more of a heads-up for future reports ;-)

Here's an extreme example:

Do you mean that it's caused by the difference between the distance from A to B in a straight line vs. along the curve?

> Broken: 2.91 (current master) Please always include the Git hash when you're not referring to an official release. That way it's less ambiguous which version you're referring to. It's happened that people just hadn't pulled changes for a while, and were a week or two behind on updates and were reporting a bug that was already fixed. Given that you've just shown that it was in 2.63 already chances of that are small, but it's good to do anyway. It also helps to keep the task description up to date, so that people who want to fix it don't have to read through all the comments to get the info. I've done that just now, this is more of a heads-up for future reports ;-) > Here's an extreme example: Do you mean that it's caused by the difference between the distance from A to B in a straight line vs. along the curve?

Yes, that's what I meant, but now I am not so sure... since I've been poking through the code and fixing #81704, I think it's actually probably a bug in where_on_path in anim_path.c rather than Spline IK. Aside from #81704, I think Spline IK is doing the right thing. For evidence-- it looks like the Follow Path constraint, which also uses where_on_path has the same behavior. Here's a .blend file:
where_on_path_bug_maybe.blend
Still, the fact that it seems to have the right number of segments, but the wrong length and rotation for the segments (as seen in the pictures in my last comment), this is very strange.

Now, where_on_path https://developer.blender.org/diffusion/B/browse/master/source/blender/blenkernel/intern/anim_path.c$206 is complicated and deals with all kinds of weird math. So I don't think I can fix this.

Anyways, I think that if the Spline IK were doing calculations on a mesh representation of the curve, it could find the points by doing sphere-intersections of the radius of the bone. That would solve this and #81704, although I have no idea how to do it, attempting to get a mesh from the curves' Runtime resulted in a crash, so I've been trying to solve #81704 by getting a better point from where_on_point.

I'm OK with leaving this as a Known Issue... it's annoying but I can just use a Poly spline if I wanted exact follow on a low-res spline.

Thanks 😁

Yes, that's what I meant, but now I am not so sure... since I've been poking through the code and fixing #81704, I think it's actually probably a bug in `where_on_path` in anim_path.c rather than Spline IK. Aside from #81704, I think Spline IK is doing the right thing. For evidence-- it looks like the Follow Path constraint, which also uses `where_on_path` has the same behavior. Here's a .blend file: [where_on_path_bug_maybe.blend](https://archive.blender.org/developer/F9051927/where_on_path_bug_maybe.blend) Still, the fact that it seems to have the right number of segments, but the wrong length and rotation for the segments (as seen in the pictures in my last comment), this is very strange. Now, `where_on_path` https://developer.blender.org/diffusion/B/browse/master/source/blender/blenkernel/intern/anim_path.c$206 is complicated and deals with all kinds of weird math. So I don't think I can fix this. Anyways, I think that if the Spline IK were doing calculations on a mesh representation of the curve, it could find the points by doing sphere-intersections of the radius of the bone. That would solve this and #81704, although I have no idea how to do it, attempting to get a mesh from the curves' Runtime resulted in a crash, so I've been trying to solve #81704 by getting a better point from `where_on_point`. I'm OK with leaving this as a Known Issue... it's annoying but I can just use a Poly spline if I wanted exact follow on a low-res spline. Thanks 😁

In #81707#1040313, @Josephbburg wrote:
I'm OK with leaving this as a Known Issue... it's annoying but I can just use a Poly spline if I wanted exact follow on a low-res spline.

Ok, then I'll mark it as known issue, that way it's at least classified.

> In #81707#1040313, @Josephbburg wrote: > I'm OK with leaving this as a Known Issue... it's annoying but I can just use a Poly spline if I wanted exact follow on a low-res spline. Ok, then I'll mark it as known issue, that way it's at least classified.

This issue was referenced by cf2baa585c

This issue was referenced by cf2baa585cc8788b29147d6e34fa8c46609e5bf9

Changed status from 'Confirmed' to: 'Resolved'

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

Spline IK Joints "Floating" above curve #81707