Multi-step limit cycle generation for Rabbit’s walking based on a nonlinear low dimensional predictive control scheme

Abstract

In this paper, a new nonlinear predictive control scheme is proposed for a five-link planar under-actuated biped walking robot. The basic feature in the proposed strategy is to use on-line optimization to update the tracked trajectories in the completely controlled variables (actuated coordinates) in order to enhance the behavior and the stability of the remaining indirectly controlled ones (unactuated coordinates). The stability issue is discussed using the Poincaré’s section tool leading to a computable criterion that enables the stability of the overall scheme to be investigated as well as the computation of a candidate region of attraction. The whole framework is illustrated through simulation case-studies. To attest the efficiency of the proposed scheme, robustness against model uncertainties and ground irregularities are investigated by simulation studies.