Fast Walking with Rhythmic Sway of Torso in a 2D Passive Ankle Walker

There is a category of biped robots that are equipped with passive or un-actuated ankles, which we call Passive-Ankle Walkers (PAWs). Lack of actuation at ankles is a disadvantage in the fast walking of PAWs. We started this study with an intuitive hypothesis that rhythmic sway of torso may enable faster walking in PAWs. To test this hypothesis, firstly, we optimized the rhythmic sway of torso of a simulated PAW model for fast walking speed, and analyzed the robustness of the optimal trajectories. Then we implemented the optimal trajectories on a real robot. Both the simulation analysis and the experimental results indicated that optimized torso-swaying can greatly increase the walking speed by 40%. By analyzing the walking patterns of the simulated model and the real robot, we identified the reason for the faster walking with swaying-torso: The rhythmic sway of torso enables the robot to walk with a relatively large step-length while still keeninu a hizh sten-frenuencv.