Abstract:Recent developed humanoid robots have had the capability to demonstrate fast and stable walking, while it is still a challenge for those robots to walk stably and flexibly in complex environments. A common solution is to search for appropriate footsteps either from a predefined motion set or not. However, the inaccuracy in models and external perturbation during implementation will make the robot fail to follow such footsteps. This paper proposes a solution to replan motions online, which promises the robot closely follow the given footsteps as well as keep stable. With analysis of the dynamics and kinematics of the humanoid robot, the proposed method replans the desired ZMP (Zero Moment Point) for the preview control based stable biped walking by combining the analysis on CoM (Center of Mass) state transition and the adjustment of supporting time. Then the desired ZMP is slightly regulated within its stable region to achieve a smooth CoM trajectory. Through the dynamic planning process in each control cycle, the robot can perform flexible motions with balance maintained. The simulation results indicate that the method proposed can effectively overcome external disturbances and provide better support during biped walking. The time cost also meets demands of online control.

A Zmp and Com Online Replanning Method for Stable Walking of Bipedal Robots