MODELING AND SIMULATION OF QUADRUPED ROBOT LOCOMOTION

Abstract

This paper presents a comprehensive analysis and implementation of a model predictive control framework designed to compute ground reaction forces for joint-level control of quadruped robots. The robot’s rigid-body dynamics are systematically simplified, allowing the control problem to be formulated as a convex quadratic optimization. Simulation results, obtained using a reduced-order dynamic model, confirm the reliability and consistency of the proposed control strategy. These results are further validated in a high-fidelity environment using the MuJoCo simulator, in which the Unitree-Go1 quadruped robot model is employed. The model predictive controller successfully maintains dynamic stability. The ground forces are then mapped to joint torques through inverse dynamics computations, ensuring that the generated torque commands remain physically realizable. The controller demonstrates robust tracking performance of desired center-of-mass trajectories. The outcomes suggest that the simplified model retains sufficient dynamic fidelity to capture the core behaviors essential for real-time gait generation and control of the robot. These results establish a strong foundation for future developments in advanced control strategies.