Abstract
This paper proposes a novel concept called COG impedance matrix of a humanoid robot, which provides an explicit relationship with the joint impedance or joint PD controller. Using the COG impedance matrix, we can calculate the macroscopic feedback gain which approximates the joint impedance, and compute the maximal output admissible (MOA) set for falling detection. First-lag-order COP dynamics is also proposed to compensate the nonlinearity in the whole body dynamics.
