Time-Efficient Contact Consistent Whole-Body Control Framework via Reduced-Dimension Dynamics Construction

Whole-body control frameworks for highly redundant, floating-base robots like humanoids are challenging due to the increased computational complexity arising from their high degrees of freedom. In this paper, the concept of reduced-dimension dynamics previously applied to whole-body control frameworks is extended by integrating it into the operational space formulation. The reduced-dimension dynamics simplifies the dimension of the dynamic model of humanoid robots by partitioning the kinematic chain into constrained and unconstrained chains. The unconstrained chain is reduced using centroidal dynamics, enabling efficient computational modeling in the operational space. By applying the reduced-dimension dynamics to operational space whole-body control, the computational burden caused by the high degrees of freedom in the whole-body dynamics is reduced, leading to shorter computation times. Unlike existing methods that utilize full-dimensional whole-body dynamics, the proposed method constructs a more efficient framework by performing optimization on reduced-dimensional dynamics, combined with operational space formulation. Comparative simulation results demonstrate a significant reduction in computation time, specifically a 41% decrease compared to existing methods, which provides a new approach to contact-consistent control in humanoid robots.