Estimation of the center of mass of humanoid robot

A closed-loop observer is suggested which can extract more reliable estimates of humanoid CoM than the CoM conversion method of using just joint angle measurements. The discrete Kalman filter is adopted as the estimation framework, where the two heterogeneous measurements of ZMP and joint angles are fused to produce optimal outputs in the sense of minimum variance of estimation error. Simulation results show that the developed closed-loop CoM observer is more beneficial than the popularly used open-loop type CoM conversion equation, especially when the robot has severe parametric uncertainties and external disturbances.