Laser-Based Kinematic Calibration of Robot Manipulator Using Differential Kinematics

This paper proposes a novel systematic technique to estimate entire kinematic parameter errors of robot manipulator. Small errors always exist in link length and link twist for physical manipulators, which affect the precision in kinematic equations leading to calculate wrong joint angle values in inverse kinematic equations. In order to solve these problems, the proposed technique employs a structured laser module (SLM), a stationary camera, the Jacobian matrices, and an extended Kalman filter (EKF). The SLM is attached to the end-effector of the manipulator arm and the stationary camera is used to determine an accurate position where the laser comes out. Variances between actual and measured positions of laser beams are represented by the Jacobian matrices formulated from differential kinematics. Then, the EKF is used to estimate kinematic parameters. Effectiveness of the proposed technique is verified with 7 DOF humanoid manipulator arm by computer simulation and 4 DOF manipulator by actual experiment.

I. Introduction

The geometry of a robot manipulator commonly does not match the designed goals, which reduces the accuracy of the end-effector location significantly. Inaccuracies in kinematic parameters of the robot manipulator arise from two types of errors: geometric and nongeometric errors. The geometric errors are due to inaccurate knowledge of the kinematic model, and the nongeometric errors are caused by joint and link compliances, thermal, wear, etc. In order to operate the humanoid robot [1] [2] or to move the manipulator arm [3] and robot finger [4] to a specified location consistently and accurately, a kinematic calibration is required because the location of the manipulator is based on a precise description of the kinematic parameters.