A computational approach for push recovery in case of multiple noncoplanar contacts

This paper presents a new computational approach for humanoid push recovery in a generalized noncoplanar multicontact context. Our approach is based on a simplified model and consists of two main steps. The first one predicts whether the perturbed system can be stopped in a final static state while maintaining fixed contacts with the environment. This prediction can be seen as a dynamic stability indicator. The second one consists of a strategy that stabilizes the system, when stability cannot be reached, through a contact change chosen as the best one among a set of a priori possible contact changes.