Stretched leg walking for a humanoid robot

Abstract

This thesis describes the realization of the stretched leg walking of a biped humanoid robot. Several humanoid robots such as ASIMO, HRP and HUBO have imitated the walking of a human, but the walking of those robots differs from that of a human in that the robots bend the knee excessively when walking. There are several reasons why humanoid robots walk with bent a knee. Bent-knee walking has several advantages; it is easy to generate a walking pattern with this style of walking, and this type is more stable compared to stretched leg walking. However, stretched leg walking is more efficient and appears more natural than bent-knee walking. Thus, this paper suggests the generation of a stretched leg walking pattern and a stabilization algorithm for walking. In the beginning of this thesis, the advantages of stretched leg walking compared to conventional bent-knee walking in addition to the generation of the walking pattern are described. Though stretched leg walking has several drawbacks, including its short double support time, large landing impact and a singular problem when generating the walking pattern, it is more efficient walking in a view point of energy because there is no excessive load on the knee joint. Moreover, it appears more natural than bent-knee walking. In order to verify the proposed walking pattern, a platform called as the Walking Guide Platform (WGP) was developed. The WGP helps a robot move in the sagittal plane without disturbances. With the aid of the WGP, it is possible for a robot to walk without stabilization. In a comparison of the walking energy of both walking patterns in which the current was measured, it was found that stretched leg walking is more efficient than bent-knee walking. Later in this thesis, a stabilization algorithm that can maintain the stability of a humanoid robot is described. Stabilization algorithm is divided into two types of controller. The first is a dynamic balancing controller that maintains the posture...