Selectively Damped Least Squares for Inverse Kinematics

We introduce a method of selectively damped least squares (SDLS) for inverse kinematics, designed for multibodies with multiple end effectors. This method is compared with Jacobian transpose and damped least squares methods. SDLS has advantages in converging in fewer iterations, in not requiring ad hoc damping constants, and in converging quickly to reachable targets. Experiments show the importance of clamping the distance from end effectors to unreachable target positions.