GESO-based Force Control for Dynamic Locomotion of Quadruped Robot with Uncertainties

Abstract

Quadruped robot control is drawing increasing research interest but is challenging due to various uncertainties caused by unmodeled dynamics and parameter variations. This paper presents a generalized extended state observer (GESO)-based force control method for dynamic quadruped locomotion under large parameter uncertainties. The uncertainties in robot dynamics are treated as unknown disturbances, which are estimated by the GESO alongside linear and angular velocities. By augmenting a classical quadratic programming (QP)-based force control framework with feedforward compensation of the disturbance estimates, a robust composite controller is developed for better system robustness. Comparative simulation results in high-fidelity ROS/Gazebo 11 environments demonstrate that the proposed approach effectively maintains locomotion performance despite parameter uncertainties. Video demonstration is available at https://liangbingyun.github.io/geso-based-force-control.