Inverse Kinematics using Jacobians, CCD and other methods.
Updated: May 28, 2020
This week we learnt about inverse kinematics problem, which is basically a problem of any linked system to move from one pose to another meaningfully.
Jacobian meaning: Jacobians come from robotics, so in robotics a rotational system is turned into a linear system, simply by breaking the circular arc into small parts till we reach a target. Jacobians do the same function and linearize the 3d rig system.
Following are the types of Jacobian Techniques:
1. Jacobian pseudoinverse
2. Jacobian Transpose Methods
3. Null space
4. Damped Psuedoinverse.
CCD: Cyclic co-ordinate descent
This is another technique to solve the inverse kinematics problem. CCD deals with each joint individually and tries to rotate each joint by an angle to reach the target position. Although it is not as mathematically grounded as Jacobian, it's much simpler to implement.
Other Methods worth mentioning:
1. Parallel Inverse Kinematics - Here parallel computing is performed, on each of the factors that affect inverse kinematics solution like rest pose ref, closeness to target position, dampness factor.
2. Fabrik - Fabrik is a solution which looks at the joint system as a point on a line to get closer to the target position. It follows two passes forward and backward pass.