A dual-stage complementary filter for dead reckoning of a biped robot via estimated contact point

This paper proposes a novel technique of dead reckoning for biped robots. The trunk position of a robot with respect to the inertial frame is estimated only using internal sensors in a short interval, so that it is available for high-rate feedback control. It is a complementary filter in which (1) the low-frequency component of the motion of the trunk is inversely computed from the relative motion of the contact point with respect to the trunk, where the rotation and rolling of the support foot about the contact point are taken into account, (2) the high-frequency component of that is estimated by double-integral of acceleration, and (3) the crossover frequency to combine those estimations is automatically adjusted based on the ground reaction force. The efficacy of the proposed method was verified through some simulations.