This paper addresses the development of versatile control algorithms for quadruped robots, with a specific focus on the Unitree Go2. Traditional control algorithms often lack generalizability and require custom adaptations based on the robot's hardware or specific movement tasks. In response, we present a control algorithm for the Unitree Go2, which integrates a joint-free centroid dynamics model and a comprehensive body dynamics model. By prioritizing variables such as the linear and angular acceleration of the centroid and the swing leg trajectory, we utilize an open-source switching system optimal control toolbox to construct and solve a nonlinear model predictive controller. This controller effectively plans the movement trajectories for the robot's centroid and limbs, translating these into control signals for the joints. Simulation results demonstrate that the proposed control scheme not only enables smooth movement on flat terrain but also enhances stability and robustness on uneven surfaces, preventing falls.
Liu et al. (Wed,) studied this question.