Arbitrary jumps on the curve

Along the curve, the catheter moves with the capsule in front using the Curve modifier. The capsule is parented to a yellow empty, and the empty is attached to the catheter using vertices. This is the only easy way to move the capsule in sync with the catheter. But trouble happens. At some moments, the capsule makes a leap to the side. This happens even between the far-spaced vertices of the curve. That is

, the curvature of the path in theory doesn’t affect what I initially thought. I can’t find the cause of jumps. Such an illustration is attached. Thank you in advance.

Empty is vertex parented to the capsule? It can depend on the deformation of the verts to which it’s parented. Spacing is important. For a curve, you’d probably want closely spaced vert parents. (For other kinds of more local deformation, you sometimes want three distantly spaced verts.)

Might be helpful if you linked a .blend somewhere.

Here is the .blend. Frame N350
and many other frames.

I’m not sure if I can figure it out. I already find 2.8 interface frustrating, and that file is more frustrating than most due to occlusion, vast differences in scale, and origin points way off the mesh (which is probably unavoidable, given the curve modifier.)

Here’s what I’d recommend:

  1. Since the capsule extends further than the end of the catheter, parent to a non-rendering (transparent or whatever) face that extends the length of the capsule, rather than to whatever verts you have it parented to. This is basically the technique I brought up two threads ago, which was too complicated for you.

  2. You need more loops on your catheter and high resolution on your curve if you want to work at that scale. Every time the the catheter tip passes through 1/12th (since you have a resolution of 12) of the distance between two control points, it’s going to jump. Increasing the resolution will increase the number of jumps but reduce the magnitude of each individual jump.

  3. That curve is way too noisy. That many control points is going to lead to a really shaky curve at the scale you’re looking at. And for smoothness’s sake, control points should be roughly evenly spaced (the resolution is per control point, not evenly spaced along the length of the curve.)

But, I’m not sure that all of that will solve your problems. I got to the point where I was too frustrated with 2.8 to play further with the file.

Thank you anyway. But I try to understand why is jumps on DIRECT segment. What is the logic?

I found a solution. If someone faces the same problem. I created a triangle and joined it with a catheter. Created a cube and parented to it a capsule empty. Then, the cube parented to three vertices of the triangle. Now the connections are placed correctly, this was the problem.

Yup, that’s what I was saying with 1. above.

Not really. It is very important to connect to the right vertices, otherwise it will not work.