Feasible Wrench Set Computation for Legged Robots

Implementation and Tuning of Momentum-Based Controller for Standing Balance in a Lower-limb Exoskeleton with Paraplegic User

“…One expected challenge during walking in the Symbitron LLE is the lack of actuation in the ankle in-/eversion. In [24] we proved that, due to underactuation, the device cannot reject an external push that generates sideways motion and whole-body rotations in the frontal plane during single stance. Therefore, the control of momentum in this direction during dynamic walking will require precise and robust scheduling of changes in contact configuration, and possibly account for the ankle passive dynamics even when using reduced models.…”

“…Gains and weights for each task have been adjusted per experiment and the used values are reported in the corresponding sections. Based on our previous findings [24], we know that during double stance (parallel stance in particular) our device cannot apply a yawing torque around the CoM. Thus, we do not control the heading of the torso and feet and the angular momentum around the vertical and focus on maintaining balance relative to the device heading.…”

“…The found region is robust in the sense that bounded forces at the CoM can theoretically be rejected in all directions without changing configuration, provided sufficient actuation. The whole-body pose influences how (and to what extent) actuation and contact forces can generate a (recovery) centroidal wrench [24] and, therefore, adding joint references with the lowest priority can assist in attaining and returning to robust poses after perturbation. Furthermore, in [29] it was proven that, to achieve stable zerodynamics, on top of controlling angular momentum, CoM, and joint positions (in the nullspace of the centroidal dynamics), the angular momentum loop must include an integral term, like torso orientation control.…”

Recovering linear and angular momentum during walking

“…We have shown some aspects of human balance recovery strategies that are important for the implementation of balance assistance in exoskeletons. These aspect require the use of complex and 3D models as used in Pratt et al [169] and Vallinas Prieto et al [170] in order to predict human like balance recovery (discussed in section: 'Simulation and validation of human-like balance recovery').…”