Perturbation recovery of biped walking by updating the footstep

Biped walking is sensitive to large perturbation because of the limited foothold and unstable nature. This paper proposed a strategy of dynamic updating the footstep for humanoid robots to recover their balance from an unexpected perturbation with one step. The proposed strategy consists of a desired footstep calculator and a swing leg controller. The desired footstep calculator consists of three stages: perturbation detection, rapid adaption, and capturabililty inspection. In the perturbation detection stage, the acceleration and the jerk of the pelvis are used to detect whether the perturbation occurs. In the rapid adaption stage, the robot modifies the desired step location quickly according to the integration of the changed acceleration during the perturbation. In the capturability inspection stage, capture region is calculated to inspect whether the desired foot step is inside the capture region. The swing leg controller is to generate the joint trajectories and torques of the swing leg based on the desired footstep online. Simulations of walking of a 12-DOF humanoid robot show that the proposed method is effective in recovering its balance even when it is perturbed from different directions by an impulse of 20 [Ns].