For inverse kinematics, aren't there often many different ways to move the claw to the specified position? Which one is preferred? Do we try to minimize the amount each joint needs to move?
I think this really depends on how you define the optimization problem and the constraints. For cases like this I assume Generative Adversarial Imitation Learning (GAIL) will work very good.
@HelloWorld Right—there are lots of ways to do it. See some comments here.
@anonymous_panda Learning is probably overkill here; you can set up some simple objectives for problems like this that will do quite well. In fact, this animation was just numerical descent on a regularized objective.
Can all inverse kinematics be solved with convex optimization?