Humans robustly adhere to dynamic walking principles by harnessing motor abundance to control forces.

Human walking dynamics are typically framed in the context of mechanics and energetics rather than in the context of neuromuscular control. Dynamic walking principles describe one helpful theoretical approach to characterize efficient human walking mechanics over many steps. These principles do not, however, address how such walking is controlled step-by-step despite small perturbations from natural variability. Our purpose was to identify neuromechanical control strategies used to achieve consistent and robust locomotion despite natural step-to-step force variability. We used the uncontrolled manifold concept to test whether human walkers select combinations of leading and trailing leg-forces that generate equivalent net-force trajectories during step-to-step transitions. Subjects selected leading and trailing leg-force combinations that generated consistent vertical net-force during step-to-step transitions. We conclude that vertical net-force is an implicit neuromechanical goal of human walking whose trajectory is stabilized for consistent step-to-step transitions, which agrees with the principles of dynamic walking. In contrast, inter-leg-force combinations modulated anterior-posterior net-force trajectories with each step to maintain constant walking speed, indicating that a consistent anterior-posterior net-force trajectory is not an implicit goal of walking. For a more complete picture of hierarchical locomotor control, we also tested whether each individual leg-force trajectory was stabilized through the selection of leg-force equivalent joint-torque combinations. The observed consistent vertical net-force trajectory was achieved primarily through the selection of joint-torque combinations that modulated trailing leg-force during step-to-step transitions. We conclude that humans achieve robust walking by harnessing inherent motor abundance of the joints and legs to maintain consistent step-by-step walking performance.