On the stable dynamic walking of biped humanoid robots

Abstract

This thesis describes about biped walking principle, gait generation and establishment of generalized walking control algorithm for biped humanoid robots. Biped walking is the locomotive way using two legs only, which shows excellent mobility with the simplest mechanism among many kinds of legged locomotion. However, biped walking shows unstable natural dynamic characteristics from the successive single support period. Therefore, to walk without losing the stability, high-level control technique and precise sensory organs are needed considering even humans learn biped walking through ceaseless practice for one or two years. To make biped humanoid robot walks, it is necessary to copy the human gait and understand its dynamics first. In this thesis, the characteristics of static walking and dynamic walking, COM(Center of Mass), ZMP(Zero Moment Point) as the criterion of biped walking stability and understanding of simple physical model are presented in the beginning part. Next, design philosophy, hardware and software architectures of KHR (KAIST Humanoid Robot) series as humanoid robot platforms are introduced. In the middle of the thesis, the motion control architecture is built first. To generate gait motion, three essential factors are defined by considering human walking process, dynamics of simple inverted pendulum model, and solution of inverse kinematics. Next, standard walking patterns of several walking types are designed properly with three essential factors. For the realization of control algorithm for dynamic biped walking, firstly, we summarize essential sensory devices for walking control and their effective uses. We define walking stages by dividing the standard waling patterns into five stages according to the movement of the robot. Secondly, three kinds of walking strategies (Walking pattern control strategy, Real-time balance control strategy and Predicted motion control strategy) are proposed with their objectives. In each control strateg...