Leg Kinematic Parameter Identification of Quadruped Robots Based on a Virtual Fixed Base

The leg kinematic parameters of quadruped robots, including connecting rod length and joint angle, often deviate significantly due to machining and assembly errors. This inaccurate knowledge can lead to foot-end position errors and ultimately affect precise state estimation and control. Directing against this issue, this paper introduces the concept of a virtual fixed base and designs a non-redundant parameter identification method. The virtual fixed base is established by 4 marking points on the body, determining a hypothetical coordinate system, which is used to solve the hand-eye calibration problem of quadruped robots. This study provides the overall process of the kinematic parameter identification method. Simulation and experiment results show that the identified kinematic parameters converge to their actual value and the foot-end position accuracy is improved by over 80% after parameter compensation.