Abstract
This paper shows two kinds of walking methods applied to a quadrupedal robot which has mammal-shaped legs such as those of horses. One of the methods is static balanced walking which has a limitation o its walking speed because of the movement of the center of gravity. The other is a new method of walking, that is, semidynamical walking with two-legged support, as do horses, for example. For the realization of this kind of walking, it is necessary that each ankle of the robot's legs has a control mechanism on the roll angle, because they maintain the posture of the robot when walking. In this paper, the walking pattern is studied by analyzing the dynamic stability in walking, and developing walking control programs on a software servocontrol system which in turn controls the microprocessors by decentralized processing. The experiment for a quadrupedal walking robot shows that semidynamical walking is 4 times faster than the static balance method.
