Emergent Humanoid Robot Motion Synergies Derived From the Momentum Equilibrium Principle and the Distribution of Momentum

Abstract

The momentum equilibrium principle reveals the relative character of the spatial momentum equation in floating-base robotics. This work clarifies how to take advantage of this character in motion generation and the design of a multitask controller for a humanoid robot. We focus especially on the angular momentum of the robot and the inherent redundancy resolution problem referred to as the “momentum distribution problem.” It is shown that with a proper momentum distribution, emergent behaviors (motion synergies) can be obtained that resemble those used by humans. A real-time controller for position- and torque-controlled humanoid robots is proposed, which has a simple structure. The performance of the controller is confirmed in a simulated environment with a number of tasks, including reactive motion control to accommodate external disturbances (both continuous-force and impacts), acrobatic tasks such as somersaults and jumps, dynamic walking with various gaits and variable center of mass height, blind stepping on an unknown obstacle, and balancing on a wobble board under external disturbances.