Stable walking of a bipedal humanoid robot involving three-dimensional upper body motion

This paper presents a method of designing a natural human-like walking pattern for a bipedal humanoid robot. Motivated by biomechanical studies on human walking, we model the walking pattern with continuous and differentiable mathematical functions. For stable walking of the robot, we design a pattern generator based on the ZMP (Zero-Moment Point) criterion. The proposed walking pattern involves three-dimensional motion of the upper body instead of restricting the upper body motion to a flat surface. With the proposed walking pattern the robot can walk with almost stretched knees. Due to environment uncertainty such as unevenness and inclination of the surface the walk may become unstable. Therefore, this paper also presents a body posture control method. The controller is based on sensory feedback and modifies the reference walking pattern in real time in order to stabilize the robot. We have implemented the proposed walking pattern and control method on the humanoid robot NAO. Dynamic walking experiments have verified that the proposed scheme improves the walking stability of the robot.