Human-like Walking of Humanoid Robot Based on Biped Kinematics and Captured Motion of Human

摘要

Human-like walking for humanoid robot is still an open research topic. This article proposes a method to generate real human-like walking pattern for humanoid robot. As walking patterns generated by existing methods differ a lot from these of human, especially for the traditional walking pattern with knee bent and pelvis lower down with a constant height, we started from the modeling of kinematics of the biped mechanism and analyzed the constrains that restrict the flexible motion of the robot legs. Based on the study of captured walking motions of human, we pre-determine the rotation motion of the waist. The two knee joints are planned corresponding with each other to avoid restriction of pelvis workspace. A human-like walking pattern with knee-stretch, high level waist height and smooth waist movement is generated. The walking pattern is verified in a dynamic simulation environment. A stable human-like walking with 0.3m step length and 0.7s step time is realized.