On the Design of a Nonlinear Model Predictive Controller based on Enhanced Disturbance Observer for Dynamic Walking of Biped Robots

Abstract

In this paper a nonlinear model predictive control strategy based on enhanced nonlinear disturbance observer is proposed to control the dynamic walking of biped robots on the smooth surface considering double support phase, single support phase, and impact. Optimal tracking of reference trajectories via optimal joint torque is established via a nonlinear predictive controller with well-defined cost functions and associated constraints. The implementation of a conventional disturbance observer encounters numerous challenges due to the joint acceleration requirements. The proposed nonlinear disturbance observer here, which only requires the position and angular velocity, helps to estimate the disturbances introduced on the robot and reduce the complications. The simulation results performed on the dynamic walking of a 5-DOF biped robot on flat surface shows the merits of the proposed method in tracking arbitrary trajectories despite the disturbances.