search

CMM-22 / a2

CMM Assignment 2 - Kinematic walking controller
1 stars 0 forks source link

[Remark] Use estimated Jacobian (computed by FD) in your IK solver. #18

Open eastskykang opened 2 years ago

eastskykang commented 2 years ago

You may not get the full point if you use analytic Jacobian in your IK solver implementation, since our auto-grading system will use numerical Jacobian for evaluation.

Ynossy commented 2 years ago

I suppose that also means we should not use a regularisation term for the pseudo inverse, right? (i.e. to avoid the legs going crazy when we lift the body so high, the feet can't reach the floor anymore)

eastskykang commented 2 years ago

@Ynossy No. You don't need a regularizer for Gauss-Newton (in our use case).