Quadruped Robot Control Through Model Predictive Control with PD Compensator

Chang Xu,Ma Hongxu,An Honglei

DOI: https://doi.org/10.1007/s12555-020-0734-9

IF: 2.964

2021-01-01

International Journal of Control Automation and Systems

Abstract:This paper proposes a control method when legs mass of a quadruped robot cannot be ignored or the robot carries heavy goods of unknown weight. The core part of the method is to solve the ground reaction forces on the foot ends. Expected ground reaction force is composed of two parts. The first part is given by the uniform force MPC controller. It uses a simplified robot dynamics equation to solve the ground reaction force received at each foot end. By unifying the input force at each time step, the solution time is shortened. The second part is given by a PD compensator, which is composed of force compensator and torque compensator. In order to verify the proposed method, a 12-degree-of-freedom quadruped robot model with a leg mass to body mass ratio of approximately 1:2 is established in the Gazebo simulation environment. The simulation results demonstrate that the method can realize the quadruped robot to traverse uneven terrain when carrying heavy objects and achieve various gait movements. In addition, the control method is robust and can be easily extended to legged robots.