A novel calibration method of parallel kinematic manipulators based on multi-population coevolutionary neural network

Kinematic calibration is an indispensable procedure for the predefined parallel kinematic manipulators to improve their operational accuracy with/without external loads. However, different with the calibration of simple devices such as force/torque sensors, as a rather complex mechanical system, the calibration of parallel kinematic manipulator is complicated and time-consuming. In this research, the concept of optimal robust calibration is developed as an effective approach to largely reduce various errors of the predefined parallel mechanism. A multi-population coevolutionary neural network is designed to establish the complex nonlinear relationship between joint variables and the related deviation with respect to the measured pose of the end-effector. With this algorithm, the pseudo-error in arbitrary joint configuration is obtained and thus the control parameters can be adjusted accordingly. The results are validated through the case studies about a unique parallel robotic machine tool.