Pose error prediction and real-time compensation of a 5-DOF hybrid robot

This paper proposes a new calibration method for a 5-DOF hybrid robot, concentrating particularly on addressing the contradiction between measurement efficiency and calibration accuracy and the real-time compensation with high precision. The approach involves two successive steps: (1) an error prediction model based on a back propagation neural network (BPNN) and the Denavit-Hartenberg (D-H) method is established by the strategy of pose error decomposition; (2) an embedded joint error compensator based on a BPNN is designed to achieve real-time compensation with high precision. Experimental verification shows that the maximum position/orientation errors can be reduced by 87.05%/85.62% over the entire workspace of the robot after calibration.