Sorry for the confusing title, this seems to be a difficult problem to concisely describe. I am working on a 50-DOF (yes, 50) robot that is comprised of "elbow macaroni" links that rotate around their end normals, as shown below
So far, I have implemented inverse kinematics using CCD, which works very well and solves fast even for the large amount of links. However, I am still having a problem trying to get the robot to "stay on" it's previous path. Let me explain:
The robot is continuously "fed-out" by discrete units of 0.1" or so. As it feeds out, the desired behavior is to move the end effector in a certain path, while the links behind the end effector track the previous path in space of the end effector. So, if the end effector moved forward in 4 steps of 0.1" to points A, B, C, and D - the centerline of each of the links would travel along (or as close as possible to) the interpolated path as possible.
Something like fixing the actuator angles wouldn't work since the position of each link change. And if I fix the position of each point to some distance down the path, I have no guarantee that the configuration of the robot would be valid.
Any thoughts on solving this problem?