Implementation of the model predictive control for on-line trajectory planning of a walking robot subjected to external disturbances

Abstract

In everyday life environments, various disturbances like pushing or pulling forces may cause humanoids to fall. Thus, they should be able to quickly adapt to disturbances. Model predictive control is a general control scheme which can generate the online motions need to be realized to act efficiently against disturbances. In this paper the model predictive control is used to address the problem of online compensation of the external disturbances exerted to a biped robot. This is accomplished using a three dimensional linear inverted model of the robot in combination with model predictive control scheme. Several simulations have been done considering wide range of the push forces exerted in the Single Support Phase (SSP) and Double Support Phase (DSP) of the walking cycle. Design parameters of the MPC such as control and prediction horizon and constraints on the manipulating variable are adjusted to improve the results. Simulation results demonstrate the effectiveness of the method to compensate the disturbances both in SSP and DSP.