Quasi Real-time Walking Control of a Bipedal Humanoid Robot Based on Walking Pattern Synthesis

This research is aimed at the development of bipedal humanoid robots working in human living space, with a focus on its physical construction and motion control method. At the initial stage, we developed a 35 DOF bipedal humanoid robot WABIAN (WAseda BIpedal humANoid)[1] in 1996, and proposed a control method for dynamic cooperative biped walking. By this control method, we have realized a various number of patterns, such as dynamic forward or backward walking, marching in place, dancing, and carrying a load. In this paper, we presented a quasi real-time walking control of a bipedal humanoid based on those walking pattern syntheses. Concretely, we developed a follow-walking control method with a patterns switching technique for a bipedal humanoid robot to follow human motion through hand contact, as an application to a human-robot physical interaction. At the same time, we adopted a 6-axis force-torque sensor system to WABIAN’s hand. By using the control methods, WABIAN has been able to perform dynamic marching in place, walking forward and backward during a continuous time while someone is pushing or pulling its hand in such a way.