A Compensation-Based Model Predictive Control of Quadruped Robot for Loco-Manipulation

Quadruped robots have already achieved high levels of dynamic mobility through extensive research. However, their application scenarios are significantly limited due to deficiencies in their manipulation abilities. Consequently, integrating a robotic arm into a quadruped robot has emerged as a crucial approach to extend its manipulation capabilities. Nevertheless, the addition of a robotic arm increases the overall degree of freedom and alters the mass distribution, thereby exacerbating control challenges. Therefore, this paper proposes a compensation-based model predictive control method to achieve collaborative operation between the quadruped robot and the robotic arm. This method models the disturbances introduced by the robotic arm upon the body, further incorporates them into the whole-body dynamics and model predictive control framework to compensate for the disturbances using ground reaction forces. The effectiveness of the proposed method was validated through simulations, where a quadruped robot performed the trot gait and the robotic arm executed large-scale movements collaboratively.