Tentacle CCD IK

Had trouble implementing this one so I created some supporting visuals. I added debugLineSegments and stepped through individual tentacle segments to get a better understanding.

The pink lines point from a segment’s pivot point to the endpoint and to the target (the mouse click). Starting with the last segment, the blue line will become the new angle of the segment. The second last segment with a blue line indicating how it should be rotated to get the endpoint to touch the target. Same for the third last segment.

It turned out my problem was that I had calculated the angle between the two pink vectors (pivot to endpoint and pivot to target) but I did not have a direction of rotation. So sometimes the segment would rotate the endpoint toward the target, other times away from the target. This depended on the relative positions of the pivot, endpoint, and target.

The fix came from StackOverflow: https://gamedev.stackexchange.com/a/45434

enum {clockwise = 1, anticlockwise = -1};

inline int Vector2D::Sign(const Vector2D& v2)const
{
  if (y*v2.x > x*v2.y)
  {
    return anticlockwise;
  }
  else
  {
    return clockwise;
  }
}

Now it’s working.

CCD IK Results

#

At first the behaviour was horrible. These results are really weird… until I realized I was processing from the base to the tip instead of vice versa. I had changed my for loop to accommodate highlighting the correct line segments from my debugging above. Correcting that gets me much better results: These would be better as GIFs, reckon.

Wow, ImageMagick supports this now. Before with Diagrammar I had to use a tool called “Gifsicle.” Or maybe that was for quality or something, I dunno. But for this I can:

convert -delay 50 -loop 0 set01/*.png set01.gif

And Evernote supports them, thankfully. Cool. Here are the others: It doesn’t move like I would expect a tentacle to, which is pretty interesting. Reckon the goals of the algorithms I’m using are different from mine. Also, I should keep in mind the intermediate frames wouldn’t necessarily be seen by the user. Once I know the solution I could have the tentacle move toward that solution in some other way. So the question is whether the end frames look like what I want. The first and last animation yes. The second one isn’t great but maybe if I have longer segment lengths closer to the base and shorter ones near the tip it looks okay.