Predictive Control of a Spined Quadrupedal Robot Based on a Dual Rigid-Body Model

The spines of quadrupedal animals play crucial roles in their locomotion, as the spinal structure provides flexibility and ensures stability for necessary posture adjustment. However, for a quadrupedal robot with a similar multi-degree-of-freedom (multi-DOF) spine structure, the corresponding control problem poses great challenges for real-time application, especially with limited computation resources onboard. Besides, the single rigid-body model developed in previous research is not competent any more. This paper proposes a dual rigid-body model for a customized quadrupedal robot with a 2-DOF spine. A modified model predictive control framework is also developed to take spine torques as optimization inputs for dynamic locomotion. The resultant leg and spine coordination helps the robot in simulation trot and gallop at maximum speeds of 3.8 m/s and 4.9 m/s, respectively, with varying spine configuration separately as the robot is moving possible. However, the robot's performance in experiments lags behind due to the motor torque limitation in simulation is doubled. The physical robot can trot at a maximum speed of 1 m/s before the knee motor protection is triggered. The hardware limitation will be addressed in the near future to unleash the spine's potential in dynamic locomotion.