Hierarchical fall avoidance strategy for small-scale humanoid robots

Abstract

Ankle, momentum, and/or take-a-step strategies constitute different schemes that a humanoid robot may take to avoid falling, where each strategy has a different energy overhead associated to it. To minimize energy consumption it is important to know when each of these strategies can be applied and yet be effective at preventing a fall. This paper is a continuation of our previous work on the development of a hierarchical fall avoidance approach for humanoid robots. While ankle and hip strategies were previously developed, here we develop a decision surface for a stepping strategy that determines at the onset of a disturbance if by taking a step falling can be avoided. Experiments are conducted on the Webots simulator to validate the theory.