Push Recovery of a Biped in Different Stance Scenarios

Abstract Push recovery of a biped robot is challenging because of the complexity of biped locomotion contributed by its straints. The present work proposes a novel optimized control methodology for push recovery against external disturbances. bances utilizing “Centroidal Momentum” and “Momentum Jacobian Matrix” (MJM). The novelty of this work lies in the optimization framework of the control problem, which includes an eigenanalysis of MJM. The optimization considers various constraints associated with the biped locomotion, i.e., no-slip, no-tipping on the ground, and input joint torques to remain within permissible actuator limits while minimizing the necessary cost function. Successful push recoveries of a biped from different stance scenarios (single leg stance, double leg stance, and double leg stance at different heights) are demonstrated using the unique methodology proposed in this work. Such a unique methodology using eigenanalysis of MJM for push recovery under diverse stance scenarios where the controller does not know the push force's direction, line of action, or terrain geometry has yet to be described in the literature. The balance stability of the robot after being pushed is evaluated by its Center of Mass (CoM) motion and Zero Moment Point (ZMP) criteria. It is also demonstrated that the proposed methodology in the present work can recover a biped from a greater impulse force normalized by biped weight compared to most other existing push recovery methods. To perform this comparison, a new term called ‘Effective Disturbance Ratio’ (EDR) is introduced.