This paper analyzes the physical correctness of interpolated motion based on physical properties, and show that with some modifications to interpolation techniques previously propsed they prove that their constraints are satisfied, providing that the motions being interpolated do not violate certain conditions.

Contributions / Key ideas of the paper

 * Analysis of interpolated motion by linear and angular momentum, the phase when the feet are in contact with the ground, and the continuity of the transitions between phases. 
 * Changes to how interpolation is done to better satisfy physical correctness.

Questions that this paper brought up in my mind
 * Has anyone quantified the case where 2 natural motions produce an awkward motion that may still be physically correct, but which will look unnatural.
 * Has any progress been made in the case of highly dynamic motion.