A Theoretically Motivated Reduced Order Model for the Control of Dynamic Biped Locomotion

A new reduced order model of a biped locomotion system and its theoretical basis are presented based upon the concept of local feedback. It has been reported from physiological studies that local feedback at each articular joint exists in human motor control, and the authors have already studied from the viewpoint of control engineering that local feedback makes the total system robust. A theorem is presented by which two dominant modes are obtained by the application of local feedback at each articular joint in the biped locomotion machine and it is shown that these two modes correspond to inverted-pendulum modes. The reduced order model derived by using these two modes contains the effect of the motion of body and swing leg which is indispensable for the control of biped locomotion. It is shown that this model can approximate very well the original higher model in almost all walking phases containing start-up and stopping motions.