Most of the walking motion for humanoid robot is done by off-line playback. It means that the motion is pre-programmed and robot moves as the data already calculated. However this method is weak for disturbance. In order to overcome the disadvantage, there must be an on-line stabilization algorithm for walking.
In this study, posture stabilization for one leg is treated. Walking motion can be thought as a combination of one leg support phase and two legs support phase. However mostly one leg support and weight shifting are thought as essential element of walking. Therefore study about one leg stabilization is the fundamental research in walking.
Considering the algorithm of walking for human and sensing organs, the minimum sensor information in order to stabilize the robot is discovered.
One leg robot is represented by a double inverted pendulum. For the case of human, one rotates the hip joint right and left so as to stabilize in one leg. Similarly, the leg and body was thought as a link respectively and actuator for hip joint and free joint for the ankle which almost no torque is generated.
Double inverted pendulum was controlled to stand still by LQ controller. First, system is controlled with the degree information of ankle by using encoder. Second, system is controlled with the angular velocity information of ankle. Although using only the angular velocity information, it was possible to stand opposite to the gravity regardless of ground inclination. Considering that human also do not have the exact angle measuring sensor, but angular velocity sensor in ear, this result tells that posture stabilization mechanism which is very similar to human has been developed.
Stabilization algorithm is suggested and experiments are preformed.