Optimized kinematic parameter identification and torque compensation methods for bipedal robots

To accurately model bipedal robots, this paper proposes a model parameter identification method that integrates optimized kinematic parameter identification and torque feedforward compensation via an improved Stribeck friction model. This approach effectively compensates for motion errors caused by machining and assembly inaccuracies while addressing joint friction torque under various operating conditions, thereby enhancing the precision and compliance of bipedal robot motion control. A fundamental open-closed chain mechanism is designed and developed, transforming the kinematic parameter identification problem into a constrained nonlinear multivariable optimization problem based on structural characteristics. Furthermore, a robust transformation relationship between output torque and current is established across various working conditions. To validate the proposed method, a motion capture platform and a motor drag testing platform are constructed for the CASIA-BBR bipedal robot. A series of experiments, including motion capture, motor drag tests, and standing and squatting tests, is conducted to assess the effectiveness of the proposed approach. The experimental results reveal that the correction of the robot pole length exceeds 14%, whereas, the joint torque tracking error is reduced by more than 40% compared to the conventional friction models. These findings demonstrate that the proposed approach can remarkably enhance the accuracy of robot modeling, thereby narrowing the gap between simulations and real-world performance and facilitating the sim-to-real transfer of locomotion control algorithms.