Kinematic Self-calibration of a 3-DOF Parallel Mechanism with Ill-conditioned Identification Matrix.

Kinematics calibration is an effective means to improve the accuracy of the mechanism. Parallel mechanisms with redundant actuation have potential for kinematic self-calibration. In this paper, the self-calibration algorithms of a parallel mechanism with 3-DOF are studied. The Jacobian matrix for self-calibration is ill-posed/ill-condition caused by multicollinearity between kinematics parameter, and the calculation with least square method diverges. Truncated singular value decomposition (TSVD), ridge regression (RR) and Liu estimation algorithm are utilized to address this problem, and the identification performances under different error parameters are compared.