This study describes the human-like walking ability of a humanoid robot, consisted of upper and lower arms, as well as thighs, shanks, and feet, total 11 segments. To imitate the human gait, the profile of each segment of a human body while walking is recorded. These profiles were used to train the humanoid robot to achieve human-like walking ability. In
order to build a more realistic robot walking control, perturbations that are typically encountered by humans while walking are considered in the robust design of the walking control. These perturbations include
the ground reaction force, the impulse disturbance, and internal biological electromyo noise interference. These perturbations are learned and then
cancelled by adaptive neural fuzzy networks and cancellation residue with external disturbance is also attenuated by H∞ tracking again. With nonlinear adaptive H∞ tracking control including neural fuzzy networks, the humanoid robot is able to emulate the human walking gait. The robust tracking control algorithm proposed in this paper achieves human-like
walking control ability, enabling humanoid robots to walk like real human beings.