I have no knowledge of what kind of logic is used in their FSD code, but I don't see how they can derive much of a motion planning algorithm for a humanoid robo

I have no knowledge of what kind of logic is used in their FSD code, but I don't see how they can derive much of a motion planning algorithm for a humanoid robot from it. Does a car have to deal with self-collision (accidentally hitting yourself in the face with your hand) or singularities (aka "gimbal lock"/"wrist flip")? Does motion planning for a nonholonomic system like a car easily translate into a holonomic system like the human body (maybe it's easier to go in one direction vs the other, like holo->nonholo or vice-versa)?