关键词: 四足机器人, 单腿运动控制, 动力学参数辨识, 前馈-反馈复合控制, 非线性扰动, 轨迹跟踪

Abstract: During high-speed dynamic locomotion,the single leg system of a quadruped robot is subjected to strongly coupled nonlinear disturbances,which cause significant tracking errors in traditional Proportional-Derivative (PD) control and constrain the robot's overall dynamic performance. To enhance motion precision under such conditions,a feedforward-feedback composite control method based on dynamic parameter identification is proposed and implemented. Firstly,a three-degree-of-freedom single leg dynamic model incorporating joint friction is established based on the Lagrange equation. The system's minimum inertial parameter set is then extracted through symbolic analysis with the SymPyBotics library for the elimination of redundant terms. Secondly,an excitation trajectory,formulated as a finite Fourier series,is designed to satisfy the persistent excitation condition,and key dynamic parameters are identified offline using a regularized weighted least-squares method with prior parameter information. Then,a feedforward torque is calculated in real-time based on the identified dynamic model and desired motion state to actively compensate for nonlinear disturbances caused by Coriolis forces,centrifugal forces,gravity,and joint friction. Concurrently,model uncertainties and external disturbances are suppressed by a PD feedback controller,thus establishing a synergistic composite control architecture of feedforward prediction and feedback correction. The method's effectiveness is validated through comparative experiments on a single leg test platform,where the tracking performances of traditional PD and the proposed composite control are evaluated. Results indicate that the identified dynamic model can accurately predict joint torques, the root mean square error of torque prediction under the verified trajectory is relatively low. Compared to traditional PD control,the proposed composite control method reduces the average three-dimensional foot-end trajectory tracking error by exceeding 35 %,the maximum reduction in joint angle tracking error is 50.34 %. This research confirms that the proposed composite control method can effectively compensate for nonlinear disturbances during high-speed motion,leading to a significant improvement in the dynamic trajectory tracking accuracy of a quadruped robot's single leg.

Key words: quadruped robot, single-leg motion control, dynamic parameter identification, feedforward-feedback composite control, nonlinear disturbance, trajectory tracking