A Whole-Body Control Framework Based on the Operational Space Formulation Under Inequality Constraints via Task-Oriented Optimization

Abstract

This paper presents practical enhancements of the operational space formulation (OSF) to exploit inequality constraints for whole-body control of a high degree of freedom robot with a floating base and multiple contacts, such as humanoids. A task-oriented optimisation method is developed to obtain a feasible torque resolution solely for task variables based on the OSF, which effectively reduces the number of optimisation variables. Interestingly, the proposed scheme amends assigned tasks on demand of satisfying inequality conditions, while dynamic consistency among contact-constrained tasks is preserved. In addition, we propose an efficient algorithm structure ameliorating real-time control capability which has been a major hurdle to transplant optimisation methods into the OSF-based whole-body control framework. Control performance, the feasibility of the optimised solution, and the computation time of the proposed control framework are verified through realistic dynamic simulations of a humanoid. We also clarify the pros and cons of the proposed method compared with existing optimisation-based ones, which may offer an insight for practical control engineers to select whole-body controllers necessitated from the desired application.