Whole-body Motion Planning and Control of Angular Momentum Tasks in Humanoid Robots

Abstract

Biomechanical studies have determined that the angular momentum is a privileged quantity in human motion. During walking the angular momentum is regulated close to zero through coordinated limb movement, whereas in force-disturbed scenarios it can be actively generated to remain in balance. Inspired by these investigations, a new method for planning and control of angular momentum in humanoid robots is presented. The Angular Momentum Optimizer (AMO) takes whole-body reference trajectories provided by a planner as input and optimizes them with respect to an angular momentum objective. The AMO is executed online utilizing the full system model. Special emphasis is placed on the compliance with kinematic and dynamic constraints of the resulting trajectories, which serve as input for the passivity-based whole-body controller that has an explicit angular momentum tracking task. A novel concept for generating angular momentum trajectories for balancing and walking applications is introduced. Moreover, the utilization of actively generated angular momentum in force-disturbed scenarios is investigated. The presented approaches are validated in simulations and experiments with the humanoid robot TORO developed at the German Aerospace Center (DLR).