RISS 검색 - 국내학술지논문

This paper proposes a human-inspired walking controller consisting of a feedforward optimal term and a feedback stabilizing term. The feedforward term is the optimal control input as calculated from the dynamics using trajectory optimization via direct collocation. The feedforward optimal control is analogous to walking muscle synergies in humans, which are the learned muscle contraction patterns that humans use during walking. The feedback term involves tracking the deviation from the optimal joint trajectories with a PD controller, and this term can be seen as approximating the stability which comes from co-contraction and natural stiffness and damping in human joints. This hybrid controller mimics human motor control in the hopes of obtaining some of the robustness of human walking, and it aims to offset the disadvantages of using feedforward or feedback alone. Feedback mitigates the sensitivity of feedforward to minor perturbations. Feedforward combats the vulnerability of feedback to sensor noise and allows lower feedback gains to be used, since the motion of the biped is no longer entirely reliant on the feedback term. Preliminary results using this controller in the simulation of a five-link biped are presented, in which the proposed controller is compared to a feedback linearizing controller when subjected to an impulse perturbation.

상세검색

RISS 보유자료

상세검색

해외전자자료

연관 검색어 추천