Ankle Torque Control for Steady Walking of Humanoid Robot

Abstract

A humanoid robot operated with position control has the characteristic of high trajectory tracking accuracy. However, its ability to adapt to complex environments is insufficient during walking due to the rigidity of the joint. More complex methods of control are needed when walking on uneven ground. A humanoid robot operated using torque control has strong robustness to complex environments or disturbance, but the flexibility of the joint makes its operation imprecise. The study of human motion reveals that the flexibility of the ankle joint allows people to walk in complex environments, such as uneven ground, with great stability. This paper presents a stable walking strategy for a humanoid robot based on ankle joint torque control. In this method, a humanoid ankle joint with two degrees of freedom is used for joint torque control. We describe a feed-forward torque calculation method and the setting of controller gain parameters. An algorithm compliant in the z-direction was designed for swinging legs. Finally, the BHR-6P model was used to simulate and verify the proposed algorithm.